Evaluation of Digital Transformation and Upgrading in Emerging Industry Innovation Ecosystems: A Hybrid Model Approach

Abstract

In order to scientifically and reasonably evaluate the digital transformation and upgrading level of “emerging industry” innovation ecosystems, this paper firstly uses the grounded theory to extract the factors influencing the digital transformation and upgrading of the emerging industry innovation ecosystems. Secondly, a cloud model is introduced to evaluate the importance of the influencing factors, select the important factors, and construct an evaluation index system. Thirdly, the projection pursuit model based on the quantum genetic algorithm is used to search for the optimal projection direction and determine the weight and comprehensive evaluation value of each index. Finally, the digital transformation and upgrading levels of 506 innovation subjects are divided into a budding level (I), growth level (II), and mature level (III) based on K-means and the SVM—most of which are at a medium–low level. Therefore, countermeasures and suggestions for promoting the digital transformation and upgrading of the emerging industry innovation ecosystems are put forward. This paper provides a systematic and complete method for the evaluation of digital transformation and upgrading of the emerging industry innovation ecosystems. Further, this paper promotes the combination of qualitative and quantitative analysis and realizes the effective integration of the overall logic chain of theoretical demonstrations, method design, and data analysis.

1. Introduction

In China, high quality is the current dominant development idea. Emerging industries refer to industries that are related to the optimization and upgrading of industrial structure and high-quality economic development, which have a high technology content and high added value and have become the main force of innovation in China. In the digital economy era, scattered and independent industrial innovation has been unable to adapt to the requirements of complex and dynamic environmental change. The innovation ecosystem of collaboration, dependence, and symbiosis has become the core of innovation in emerging industry. Digitalization changes the technological innovation behavior of the innovation subject in the innovation ecosystem, enhances the synergy between system resources and innovation subjects, and improves the system’s perception accuracy and response speed with respect to the external ecological environment. The digital transformation and upgrading of the emerging industry innovation ecosystem has become an inevitable choice to improve innovation efficiency, innovation quality, and innovation adaptability. The scientific cognition of the digital transformation and upgrading of the emerging industry innovation ecosystem is the premise of promoting the construction of industrial digitalization, which includes two aspects: Firstly, the digital behavior orientation of the innovation ecosystem is clearly defined, that is, the scientific subject behavior direction, the association mode, and the friendly ecological technology innovation environment are determined so as to provide specific behavior guidance support for the digital transformation and upgrading of the innovation ecosystem. Secondly, the development status of digital transformation and upgrading of the innovation ecosystem should be clarified, and the digital transformation and upgrading level should be mastered through behavior judgment, so as to provide an effective decision-making basis for subsequent digital transformation and upgrading behavior. The effectiveness of behavior orientation mainly depends on the evaluation index system design of digital transformation and upgrading, and the objectivity of grading status mainly depends on the evaluation of digital upgrading development level.

At present, scholars’ research on the evaluation of the digital transformation and upgrading level mainly focuses on the construction of the evaluation index system, the selection of an evaluation method for regional and enterprise digital capability. There are few studies on the evaluation of the digital transformation and upgrading level of the emerging industry innovation ecosystem. Therefore, the academic research on the evaluation of the digital transformation and upgrading level of the emerging industry innovation ecosystem is still in the initial stage, and the research on the factors influencing the digital transformation and upgrading has not yet formed a systematic theoretical framework.

Based on the above analysis, this paper aims to establish an evaluation index system for the digital transformation and upgrading level of the emerging industry innovation ecosystem, choose objective and accurate methods to identify the digital transformation and upgrading level, and further formulate improvement strategies. Firstly, this paper makes use of the grounded theory to completely and systematically extract the factors influencing the digital transformation and upgrading of the emerging industry innovation ecosystem. In order to overcome the randomness, ambiguity, and difficulty of quantification of factors, the cloud model is used to screen out important indexes and build an evaluation index system for the digital transformation and upgrading level of the emerging industry innovation ecosystem; secondly, the projection pursuit model is introduced, which is solved based on the quantum genetic algorithm (QGA), and then the optimal weight and comprehensive evaluation value of each index are determined, which avoids the disadvantage of subjectivity in determining the weights of indexes and evaluating a single index feature; finally, based on K-means and the Support Vector Machine (SVM), the accuracy and scientificity of the index system are quantified and verified, and the index system is applied to measure the digital transformation and upgrading level of the emerging industry innovation ecosystem in China.

The rest of this paper is structured as follows. Section 2 presents a systematic literature review. Section 3 describes the methods for evaluating the digital transformation and upgrading level of the emerging industry innovation ecosystem. In Section 4, we use the proposed method to evaluate the digital transformation and upgrading level of the emerging industry innovation ecosystem and compare it with the random forest (RF) and the back-propagation artificial neural network (ANN). Finally, we present our discussion and conclusions in Section 5.

2. Literature Review

2.1. Innovation Ecosystems and Sustainable Development

Inspired by nature’s ecosystems, the ecosystem analogy in the business field was first proposed in the 1990s, and since then research on the innovation ecosystem has become popular in a rapidly growing literature [1,2]. The innovation ecosystem has produced a new, more complex innovation logic involving a variety of different stakeholders and its continuous development, and it has also brought new advantages in the development of emerging industries [3]. The innovation capacity of a multi-layered innovation ecosystem involves science, technology, and business sub-ecosystems, which include six research streams: industry platform, innovation ecosystem strategy, innovation management, managing partners, the innovation ecosystem lifecycle, and new venture creation [4]. And the ability of the innovation ecosystem to solve practical problems is being raised. Facing criticisms, some other scholars have tried to use cases to demonstrate the problem-solving ability of the innovation ecosystem [5,6]. Meanwhile, sustainable development has also become one of the laws of innovation ecosystem evolution. Sustainability refers to a process or state that can be maintained over a long period of time. Sustainability, in essence, is a characteristic of the innovation ecosystem rather than a fixed state. It reflects the inherent capacity of the system to maintain benign evolution, continuous innovation output, and adaptive responses to environmental changes over the long term. The aforementioned “state” merely describes the manifestation of this characteristic at specific stages, emphasizing the dynamic nature of sustainable operation. A benign evolution and sustainable operation of the innovation ecosystem is the key to improving the level of innovation and achieving high-quality development. Some scholars have conducted in-depth research on sustainable development [7,8] and proposed solutions to practical problems [9,10]. A sustainable innovation ecosystem can continuously promote the generation of innovation. On the one hand, ecosystems can meet sustainability requirements, and on the other hand, sustainable innovation processes are essential for the existence and sustainability of innovation ecosystems [11]. Digitalization has significantly reshaped human and social life worldwide, serving as a powerful enabler of a sustainable economy [12], and China is actively exploring new development models to promote the development of the sustainable economy through digital transformation [13].

2.2. The Design of Evaluation Index Systems

As summarized in

Table 1. Summary of literature on evaluation indexes for digital transformation and upgrading in emerging industry innovation ecosystems. [Source: own elaboration].

Research Theme	Main Indicators/Considerations	Reference No.
The evaluation index system for innovation ecosystems	Innovation subjects	[14]
	Innovation elements	[15,16]
	Synergistic interaction	[17,18]
	Innovation environment	[14,19]
The evaluation index system for digital transformation and upgrading	Digital infrastructure	[20,21,22]
	Business processes of innovation subjects	[21,22]
	Government R&D subsidies	[23,24]
	Connection between innovation ecosystems and platforms	[25]
	Organizational structure	[22]
	Human capital	[26]
	Capital investment	[22,27,28]
	Industrial innovation synergy	[20,22]
	Digital innovation environment	[27]

, the construction of evaluation index systems for innovation ecosystems mainly focuses on innovation subjects [14], innovation elements [15,16], synergistic interactions [17,18], and the innovation environment [14,19]. However, studies specifically addressing the digital transformation of emerging industry innovation ecosystems remain scarce. Existing digital transformation evaluation index systems primarily include factors such as digital infrastructure [20,21,22], innovation subject business processes [21,22], government R&D subsidies [23,24], ecosystem and platform connectivity [25], organizational structure [22], human capital [26], capital investment [22,27,28], industry innovation synergy [20,22], and the digital innovation environment [27]. While these prior works provide substantial theoretical foundations for constructing evaluation systems of digital transformation and upgrading levels in emerging industry innovation ecosystems, they do not effectively integrate the ecological characteristics of such ecosystems with the core aspects of digital transformation. As a result, a scientifically rigorous and accurate evaluation of digital transformation and upgrading levels in emerging industry innovation ecosystems remains lacking.

2.3. The Evaluation Methods for Digital Transformation and Upgrading

The existing evaluation methods for digital transformation and upgrading, such as the grey relational model [29], the analytic hierarchy process and fuzzy comprehensive evaluation [30], the entropy weight top-SIS evaluation method [22], and so on, are essentially subjective and fuzzy. There are also other evaluation methods that use the self-characteristics of data, such as the entropy value method [27], the improved entropy weight method [31], the factor analysis method [22], and so on. However, these methods lack self-learning and simulation classification and cannot objectively evaluate the digital transformation and upgrading level of the emerging industry innovation ecosystem.

Meanwhile, in terms of specific research objects and the construction of evaluation systems, the current research on digital transformation includes regional digitalization degree [25] and capability [27,32,33], enterprise digitalization capability [30,33] and transformation maturity [34,35,36], and manufacturing digitalization transformation and upgrading evaluation [19,29,37], while reasonable, objective, and effective index systems for digital transformation and upgrading are rarely discussed in the literature, such that the objectivity and accuracy of the evaluation results are affected.

3. Methods

3.1. Grounded Theory

The grounded theory is a typical qualitative method, which is mainly used to explore the inner logic of the research object [38]. At present, there are relatively few theoretical studies and results for direct reference on the factors influencing the digital transformation and upgrading of the emerging industry innovation ecosystem, and the indexes cannot be quantitatively studied. Therefore, this paper adopts the grounded theory to extract the factors influencing the digital transformation and upgrading of the emerging industry innovation ecosystem according to the steps of open coding, spindle coding, and selective coding. The specific steps are as follows:

(1)

Open coding and spindle coding. Open coding is used to analyze and summarize original data one by one, so as to find intrinsic relevance and similarities in text information and conceptualize and categorize original data, while remaining objective at all times in the process; spindle coding is a deep coding of initial concepts with the same or similar meanings on the basis of open coding.

(2)

Selective coding and saturation test. After repeated deliberation and scrutiny, the selective coding finally obtains a complete framework of factors; the saturation test is mainly used to ensure the scientificity and rigor of the theoretical model.

3.2. Cloud Model

The digital transformation and upgrading level evaluation of the emerging industry innovation ecosystem is an evaluation process involving multiple complex indexes. This section uses the cloud model to evaluate the importance of the influencing factors extracted above and select important indexes.

A cloud model is a model that realizes the conversion between qualitative concepts and quantitative representations by using specific algorithms on the basis of fuzzy set theory and probability theory. Because it overcomes the defects of randomness and ambiguity, it improves the scientificity and rationality of the conversion and also reveals the inherent relationship between randomness and ambiguity [39]. The cloud is composed of a large number of cloud droplets. Let Z be the quantitative universe represented by an exact numerical value indicating the value range of qualitative index attributes in the scheme; C be the natural language value on Z, if x ∈ Z; and x be a random realization of Z. Then, the deterministic function C_Z(x) ∈ [0, 1] of x on C is a random number with a tendency to be stable. A cloud model is mainly determined by three parameters, namely, expectation (Ex), entropy (En), and super entropy (He). Ex reflects the center of gravity of the cloud and represents the mathematical expectation of cloud droplet distribution in the solution evaluation domain. En reflects the degree of dispersion of cloud droplets in the cloud map and represents the measurable granularity and ambiguity of the evaluation concept; the larger En is, the greater the randomness. He reflects the thickness of cloud droplets and represents the entropy of entropy, which determines the degree of dispersion of the entire cloud; the larger the He value, the thicker the cloud, the more discrete the cloud droplets, and the greater the randomness of the degree of certainty, C_Z(x).

The Forward Cloud Generator (FCG) is a mapping from qualitative concepts to quantitative representations which realizes the quantification of qualitative concepts. It generates cloud droplets and conceptual certainty based on three numerical features (E_x, E_n, and H_e). The specific calculation steps are as follows:

(1)

Determining three numerical features (E_x, E_n, and H_e). There are various calculation methods for the three parameters of the cloud model. This paper uses Formula (1) to calculate them according to the actual evaluation situation of the digital transformation ability of the innovation subject:

$$\left\{\begin{array}{c}{E}_x=\left(c_{\max }+c_{\min }\right)/2\\ E_n=\left(c_{\max }-c_{\min }\right)/6\\ H_e=k\end{array}\right.$$

(1)

where $c_{\max }$ and $c_{\min }$ are the maximum and minimum values within the range of evaluation grades, respectively [40].

(2)

Generating cloud droplets. This paper selects the cloud model with better universality. k is a hyperparameter that can be adjusted according to the actual situation. After the parameters are determined, firstly, a normal random number, E_n′, with expected value E_n is generated, and its standard deviation is H_e; secondly, a normal random number, x, with expected value Ex and standard deviation E_n′ is generated, and the certainty, y_i = exp[−(x − E_x)²/2E_n′²], of x_i on the qualitative concept Z is calculated; finally, the above steps are repeated to generate enough cloud droplets.

(3)

Computerizing expert evaluation scores. Firstly, the natural language evaluations of m experts on the influencing factors for the digital transformation capabilities of n innovation subjects are collected, and the number of experts in different divisions of the i-th factor can be obtained. Secondly, the FCG is used to calculate the certainty.

(4)

Filtering indexes. For the indexes based on the grounded theory above, multiple experts give natural language opinions on their importance, and the number of experts in each interval for each index is counted. Due to the inherent randomness of the certainty, each certainty value and its corresponding score were calculated 10 times to ensure reliability, and the mean of these repetitions was taken as the final certainty and score.

3.3. Projection Pursuit Model

The digital transformation and upgrading of the emerging industry innovation ecosystem is a typical complex problem affected by multiple factors and has a nonlinear optimal solution. To avoid subjective weighting, the projection pursuit model was applied to determine the optimal weight of each index by searching for the optimal projection direction. The indices obtained from the above cloud model were then projected and aggregated to obtain their comprehensive evaluation value.

The projection pursuit model is an exploratory or deterministic analysis for some nonlinear problems. It can effectively project a combination of high-dimensional data onto a low-dimensional subspace and find the optimal projection value for the projection index function. It can also effectively avoid information distortion, inaccurate determination of index weights, and poor adaptability to high-dimensional nonlinear problems [41]. Since the commonly used particle swarm algorithm and traditional genetic algorithm have problems such as slow convergence speed, weak global optimization ability, and premature convergence, the quantum genetic algorithm (QGA) is used to solve the projection pursuit, which has superior diversity [42]. Based on the principle of quantum computing, QGA encodes through quantum bits, uses quantum revolving gates to update populations, and calculates fitness to obtain the optimal solution.

The main steps are as follows:

Step 1: Data sources and filtering.

The paper uses a Likert scale to design the questionnaire and conducts surveys on relevant personnel. After the reliability and validity analysis using SPSS, the reliable questionnaires are selected.

Step 2: Index data preprocessing.

There are many evaluation indexes, the dimensions of which are obviously different. In order to unify the dimensions of an index system, the normalization method is used to preprocess the index data, so that the numeric size remains between 0 and 1. Let the index sample set be $\left\{\left.x^{\ast }\left(i,j\right)\right|i=1\cdots n,j=1\cdots p\right\}$, where n is the number of samples and p is the number of indexes [38]. The indexes are preprocessed using Formulas (2) and (3):

$$x\left(i,j\right)={\displaystyle \frac{[x^{\ast }(i,j)-x_{\min }(j)]}{[x_{\max }(j)-x_{\min }(j)]}}$$

(2)

$$x\left(i,j\right)={\displaystyle \frac{[x_{\max }(j)-x^{\ast }(i,j)]}{[x_{\max }(j)-x_{\min }(j)]}}$$

(3)

where the positive index is normalized by Formula (2) and the inverse index is normalized by Formula (3). x(i, j) in Formulas (2) and (3) is the index after normalization; x*(i, j) is the j-th index of the i-th sample; and x_max(j) and x_min(j) are the maximum and minimum values of the j-th index, respectively.

Step 3: Projection objective function construction.

The projection direction is defined as a unit-length vector, A = [a(1), a(2), …, a(n)]. Each sample, x*(i, j), is then projected using Formula (4) to obtain a one-dimensional projection value, z(i), where a(j) represents the weight of the j-th evaluation index.

$$z\left(i\right)={\displaystyle \sum _{j=1}^{p}a(j)x(i,j)}$$

(4)

In order to better reflect the differences of sample data, the standard deviation, s_z, of the projection value, z(i), should be as large as possible, and the local density, d_z, of z(i) should also be as large as possible. Based on the above constraints, the projection objective function can be constructed as follows:

$$q(a)=s_z{d}_z$$

(5)

$$d_z={\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n(t-r(i,j))u}}(t-r(i,j))$$

(6)

$$r(i,j)=\left|z(i)-z(j)\right|$$

(7)

where t is the local density window width, r_ij is the distance between two samples, and u is the unit step function.

Step 4: Projection objective function optimization.

The value of the projection objective function, q(a), changes with the change in the projection direction, a, and Formula (8) is maximized to obtain the optimal projection direction.

$$\begin{array}{l}\max q(a)=s_z{d}_z\\ s.t.{\displaystyle \sum _{j=1}^p{a}^2(j)}=1\end{array}$$

(8)

The algorithm flowchart is shown in Figure 1:

3.4. K-Means and SVM

3.4.1. K-Means

The K-means algorithm is a prototype-based unsupervised clustering algorithm that divides groups according to distance [43]; it is a simple, fast algorithm. For a given sample set, according to the distance between samples, the samples are divided into k compact, spherical clusters, so that the points in the clusters are as closely connected as possible and the point distance between clusters is as large as possible. The algorithm steps are as follows:

(1)

Randomly select k center points;

(2)

Assign each data point to its closest center point;

(3)

Recalculate the average distance between each data point and its corresponding center point in each class;

(4)

Reassign each point to its nearest center point;

(5)

Repeat steps (3) and (4) until all data points are no longer assigned or the maximum number of iterations is reached.

3.4.2. SVM

SVM is a classification technique proposed by the AT&T Bell Labs research group led by Vanpik in 1963 [44]. It is a pattern recognition method based on statistical learning theory. SVM has many unique advantages in solving small-sample, nonlinear, and high-dimensional pattern recognition problems. The basic idea of SVM is as follows: The training data set is nonlinearly mapped to a high-dimensional feature space. The purpose is to map the linearly inseparable data set in the input space to the high-dimensional feature space, yielding a linearly separable data set. Then, an optimal separating hyperplane with the largest isolation distance is established in the feature space, which is also equivalent to generating an optimal nonlinear decision boundary in the input space.

The traditional SVM is a typical binary classification algorithm, but in many practical applications there is no way to use a simple binary classifier to complete the classification. Therefore, in order to expand the application scenarios of the SVM, the algorithm implementation in the case of multi-classification must be considered.

Multi-class SVMs are divided into two types: one-vs.-one (OVO) and one-vs.-the-rest (OVR).

OVO algorithm: For a k-class problem, two classes of samples are selected at a time as positive and negative samples to construct k(k−1)/2k(k-1)/2k(k−1)/2 binary classifiers, which helps avoid imbalanced training data. Each classifier distinguishes between a specific pair of classes (e.g., class 1 vs. class 2, class 1 vs. class 3, etc.). When classifying an unknown sample, the class with the most votes across all classifiers is assigned to the sample.

OVR algorithm: For a k-class problem, the i-th classifier treats the i-th class as the positive class and all remaining classes as negative, resulting in k binary classifiers. For an unknown sample, the classifier that outputs a positive result determines the sample’s class.

4. Results

4.1. Index Extraction

Appendix A (

Table A1. Open coding for factors influencing digital transformation and upgrading of the emerging industry innovation ecosystem. [Source: own elaboration].

Category	Original Concept and Source Material	Literature
Procurement Digitalization, A1	Supplier digital management system application level, network connection status, procurement activity automation intelligence, and controllability	[37,47]
R&D Digitalization, A2	Digital R&D platform construction level and R&D process visualization level	[21,30,35]
Production Digitalization, A3	Enterprise-owned digital processing equipment and technology platforms, automated production lines and intelligent workshops, and intelligent management of production processes	[21,22,47]
Marketing Digitalization, A4	Enterprises build e-commerce platforms to achieve online sales and use new media such as live broadcasts, e-commerce, and Douyin to promote marketing activities	[36,47]
Logistics Digitalization, A5	Enterprises build a digital warehouse to realize the digitalization of inventory management, visualization of warehousing, and traceability of logistics distribution processes	[36,37]
Organizational Digitalization, A6	Senior managers attach importance to and participate in the formulation process of digital strategic planning, and support the digital transformation of enterprises from the top strategy; daily management patterns, such as talent management, financial management, and customer relations, have also been transformed digitally	[22,48]
Cultivation of Digital Transformation Professionals, A7	Cultivation of professionals for digital transformation, such as big data and artificial intelligence	[49,50]
Research Direction Setting of Digital Technology such as Big Data and Artificial Intelligence, A8	Research direction setting of digital technology such as big data and artificial intelligence	[50,51]
Digital Related Research Team A9	Setting up a dedicated digital technology research team	[51]
Research Digital Repository and Research Digital Tools, A10	Technical reserves and specialties of universities and research institutes, such as building and improving digital resource bases for scientific research and purchasing digital tools	[51]
Establishment of the Special Promotion Agency for Digital Achievement Transformation, A11	Establishing relevant departments and institutions dedicated to promoting the transformation of digital achievements	[52]
Personalized Needs of Customers, A12	Customers follow the trend of digitalization and generate personalized digital needs	[53]
Customer Experience Digitalization, A13	Customers participate in digital production and service processes	[53,54]
E-Government, A14	Using modern information technology means, such as computers, networks, and communication systems, to realize the optimization and reorganization of government organizational structure and work flow	[55]
Information Sharing, A15	Using digital technology to improve the ability and level of government information sharing	[55]
Digital Governance, A16	The ability of the government to carry out digital governance and the level at which it does so	[55]
Digital Trading Platform Construction Level, A17	The number of digital trading platforms and the development of digital trading markets	[56]
Development Level of Data Marketization Standards, A18	The promotion and construction of market-based standards for data elements	[56]
Digital Transformation Consulting Services, A19	Digital intermediary service agencies and platforms provide enterprises with specialized digital transformation consulting services	[52]
Digital Technology Knowledge Mobility, A20	Digital technology knowledge and information flow freely and smoothly among innovators	[57]
Digital Interface Extension and Compatibility, A21	The interface of digital products is extensible and compatible	[58]
Complementary Innovations of Digital Technologies, A22	Each innovation subject contributes its own resources and capabilities to interact and realize complementary advantages and collaborative innovation	[59,60]
Digital Level of Industry–University–Research Cooperation, A23	The degree of digital cooperation among enterprises, universities, and research institutes	[57]
Digital Market Information Matching, A24	Intermediaries or governments guide multilateral markets to participate in digital trading platforms for effective communication, exchange, trading, and value co-creation, and each achieves free communication on digital trading platforms	[61]
Modular Design, A25	The digital platform divides the entire business process or service according to certain standards, designed as several modules that can be freely combined to reduce communication barriers between users and improve the efficiency of digital transactions	[62,63]
Data Collaboration, A26	The digital platform provides data assistance for innovative subjects on the platform, such as searching for business information for consumers, providing consumer demand and future trend forecasts for businesses, and coordinating data suppliers and demanders	[55,64]
Digital Empowerment, A27	Providing support for digital resources and capabilities to businesses or subjects of the ecosystem, enabling these businesses or subjects to have a certain capability or nature	[55]
Fiscal Policy Support for Digital Transformation, A28	The government formulates preferential tax policies and financial subsidy policies for digital transformation to solve the financial obstacles of digital transformation and enhance the confidence of digital transformation	[65,66]
Financial Support for Digital Transformation, A29	Emerging industries are technology-intensive and capital-intensive industries. The government supports digital transformation with preferential loans and special funds to encourage enterprises to actively explore digital transformation	[66,67]
Policy Guidance for Digital Transformation, A30	The government introduces publicity and guidance for the emerging industry digital transformation, promotes the construction of 5G facilities, builds the industrial Internet platform, and provides digital basic technology training	[28]
Consumption Digitalization Level, A31	Per capita e-commerce transaction volume (sales + purchases)	[56]
Digital Market Competition Level, A32	Competitors build competitive advantage through digital transformation	[49]
Digital Consumption Scenarios, A33	New digital marketing scenarios, such as the new crown epidemic, the Internet wave, smart stores, and the live broadcast economy, are forcing digital transformation of enterprises	[68]
Digital Infrastructure, A34	Digital technologies, such as big data, artificial intelligence, and 5G, are important foundations and prerequisites for digital transformation	[22]
Digital Technology Innovation Capability, A35	The ability to leverage digital technologies and digital infrastructure for digital innovation	[46]
Digital Technology Innovation Diffusion Capability, A36	The transfer, spillover, and network externalities of digital technology promote the diffusion of digital technology innovation	[57]
Digital Innovation Culture Atmosphere, A37	Increasing tolerance for innovation failure by fostering the digital culture	[35,36]
Digital Innovation Initiative, A38	Positivity and enthusiasm for using digital technologies to innovate	[36]
Digital Human Resources, A39	The digital operation, digital communication ability, and digital literacy of employees are the guarantee for implementing digital transformation	[30,57]
Digital Investment, A40	Increasing research and development of the latest digital technologies, such as cloud computing, big data, smart logistics, and artificial intelligence, and building smart core technologies	[67]

) presents the 40 initial categories identified using grounded theory. Subsequently,

Table 2. The value ranges corresponding to the evaluation levels. [Source: own elaboration].

Natural Language	Very Important	Relatively Important	Generally Important	Secondarily Important	Unimportant
Value Range	0.8~1	0.6~0.8	0.4~0.6	0.2~0.4	0~0.2

and Appendix A (

Table A2. Spindle coding and explanation. [Source: own elaboration].

Main Category	Subcategory	Category	Interpretation of Relationship
Innovation Subject, C1	Enterprise Digitalization Level, B1	Procurement Digitalization (A1), R&D Digitalization (A2), Production Digitalization (A3), Marketing Digitalization (A4), Logistics Digitalization (A5), Organizational Digitalization (A6)	The digitalization level of core business and organizational structure reflects the capability and level of enterprise digital transformation
	Universities and Research Institutes Digitalization Level, B2	Cultivation of Digital Transformation Professionals (A7), Research Direction Setting of Digital Technology such as Big Data and Artificial Intelligence (A8), Digital Related Research Team (A9), Research Digital Repository and Research Digital Tools (A10), Establishment of the Special Promotion Agency for Digital Achievement Transformation (A11)	The cultivation of digital talents and the setting of research directions such as big data and artificial intelligence reflect the level of digitalization in universities. The establishment of digitalization-related research teams, the acquisition of digital scientific research resources and digital research tools, and the promotion of digital achievements are key to measuring the digital transformation capabilities of scientific research institutes
	Digitalization Level of Customers, B3	Personalized Needs of Customers (A12), Customer Experience Digitalization (A13)	The personalized digital experience needs of customers, driven by their pursuit of digital trends, reflect the level of their digital transformation capability.
	Digitalization Level of Government, B4	E-government (A14), Information Sharing (A15), Digital Governance (A16)	The capability and level of e-government and digital governance reflect the capability and level of government digital transformation
	Digitalization Level of Intermediaries, B5	Digital Trading Platform Construction Level (A17), Development Level of Data Marketization Standards (A18), Digital Transformation Consulting Services (A19)	The establishment of digital trading platforms and trading markets, the construction of data marketization standards, the promotion of data cross-border flow, and the level of digital transformation consulting services reflect the digitalization levels of intermediaries
Connection Relation, C2	Digitalization Level of Innovation Chain, B6	Digital Technology Knowledge Mobility (A20), Digital Interface Extension and Compatibility (A21), Complementary Innovation of Digital Technology (A22), Digital Level of Industry–University–Research Cooperation (A23), Digital Market Information Matching (A24)	The mobility of digital technology knowledge, the scalability and compatibility of digital interfaces, and the complementary innovation of digital technology reflect the level of digital connection in the enterprise innovation chain; the digitalization level of industry–university–research cooperation reflects the digital connection level of the industry–university–research innovation chain; the matching level of digital market information allows enterprises to participate in the purchase of data through market transactions and realizes the digital connection between enterprises and intermediaries or the government, which is a digital connection that assists the innovation chain; a small connection between two or three innovation subjects connects the original scattered and isolated innovation subjects
	Application Level of Digital Platform, B7	Modular Design (A25), Data Collaboration (A26), Digital Empowerment (A27)	The digital platform divides business processes or services into modules that can be freely combined according to certain standards and provides data assistance and capability support for innovation subjects on the platform, so that innovation subjects or business activities have digital capabilities and characteristics and realize the digital upgrade of the ecosystem
Innovation Environment, C3	Policy Environment, B8	Fiscal Policy Support for Digital Transformation (A28), Financial Support for Digital Transformation (A29), Policy Guidance for Digital Transformation (A30)	The government provides fiscal policy support through tax cuts and financial subsidies, provides financial support through loan support and the establishment of special funds for digital transformation, introduces policies to promote digital transformation, increases demonstration and publicity of successful cases, actively provides training and coaching for enterprise digital transformation, and guides enterprises to effect digital transformation
	Market Environment, B9	Consumption Digitalization Level (A31), Digital Market Competition Level (A32), Digital Consumption Scenarios (A33)	The consumption digitalization level has been continuously improved, and competing companies are implementing digital transformation strategies; the rise of new digital consumption scenarios has led to increasingly fierce market competition and forced digital transformation of enterprises
	Technical Environment, B10	Digital Infrastructure (A34), Digital Technology Innovation Capability (A35), Digital Technology Innovation Diffusion Capability (A36)	The innovation level of emerging industry is improving globally. Digital technologies such as big data, artificial intelligence and 5G are developing continuously, and the transformation and upgrading of emerging industries are accelerating. The stronger the innovation and diffusion capacity of digital technologies, the more effective they will be in promoting the digital transformation and upgrading of the innovation ecosystem
	Cultural Environment, B11	Digital Innovation Culture Atmosphere (A37), Digital Innovation Initiative (A38)	The cultural atmosphere and enthusiasm for digital innovation in the innovation ecosystem affect the positivity of enterprises with respect to participating in digital innovation and increase the motivation for digital transformation of enterprises
	Resource Environment, B12	Digital Human Resources (A39), Digital Investment (A40)	Digital investment and digital talents are the resource guarantee for enterprises to implement digital transformation

) illustrate the final coding results, including the core categories and 15 subcategories derived through iterative analysis. Notably, the digitalization of innovation subjects is categorized based on their primary functions, activities, and behaviors. For example, enterprises realize digitalization from major business links such as procurement, R&D, production, marketing, and logistics; universities and research institutes train digital transformation professionals, set up digital technology research directions such as big data and artificial intelligence, set up digital-related research teams, build scientific research digital resource libraries, purchase scientific research digital tools, and establish digital achievement transformation promotion agencies and other means to achieve digital transformation; customers mainly adapt to the development of the digital market economy by experiencing and using digital products; the government realizes digital transformation by building a digital government, improving the capabilities of e-government and digital governance; intermediaries go digital by building digital trading platforms, promoting the development of data marketization standards and providing digital transformation consulting services. Some scholars consider the process, interface, and platform of digital innovation and study the connection relationships in the digitalization of innovation chains, industry–university–research cooperation chains, auxiliary innovation chains, and application platforms. The innovation environment is an important part of the digitization of the industrial innovation ecosystem, including soft environments, such as cultural and policy environments, and hard environments, such as digital infrastructure [19].

To ensure the representativeness of expert opinions for the 506 organizations, the 10 selected experts were deliberately chosen to cover all major categories of innovation subjects, namely, enterprises, universities, research institutions, government departments, and intermediary organizations. These experts were affiliated with six types of entities corresponding to the composition of the 506 organizations: three were from leading enterprises (including manufacturing and digital services), two were from universities specializing in innovation ecosystem research, two were from government regulatory agencies responsible for industrial digital policies, one was from a research institute focused on new energy technology, and two were from intermediary organizations engaged in digital transformation consulting. A summary expert information table is provided in Appendix A (

Table A3. Summary table of expert information. [Source: own elaboration].

Expert ID	Institution Type	Institution Name	Position/Title
E1	Enterprise	Jinan Bresee Co., Ltd.	Digital Transformation Manager
E2	Enterprise	Shandong Inspur Group	Senior Engineer
E3	Enterprise	Weihai Cyberguard Technologies	CTO
E4	University	Shandong University	Professor
E5	University	Harbin University of Science and Technology	Associate Professor
E6	University	Shandong Jiaotong University	Professor
E7	University	Shandong Technology and Business University	Professor
E8	Government	Qiqihar Municipal Government	Officer of the Division of Science, Technology, and Electronic Information
E9	Government	Weihai Municipal Government	Director of the Municipal Big Data Center
E10	Intermediary	Weihai Wisdom Valley	Strategy Consultant

) to enhance transparency regarding the expert profiles. All experts had more than ten years of experience in digital transformation or innovation ecosystem management, with deep knowledge of industry dynamics. Their insights were refined through three rounds of Delphi consultation and validated against actual operational data from the 506 organizations. Furthermore, this expert pool was sufficient to achieve conceptual saturation in the grounded theory analysis, as confirmed through a saturation test. These measures ensured that, despite the small number, the selected experts could effectively represent the broader characteristics of the sample population.

The original data were coded, and 10 experts from enterprises, universities, research institutions, and other stakeholders engaged in research and management of industrial digital transformation and upgrading were selected. The authors discussed the completeness of the refined concepts and categories, and finally confirmed that the categories in Appendix A (Table A2) had achieved saturation.

4.2. Screening of Evaluation Indexes

As shown in Table 2, this paper divides the importance of influencing factors into five levels, which are “Very Important”, “Relatively Important”, “Generally Important”, “Secondarily Important”, and “Unimportant”, and their corresponding values are “[0.8, 1]”, “[0.6, 0.8)”, “[0.4, 0.6)”, “[0.2, 0.4)”, and “[0, 0.2)”. Bilateral constraints are selected for the middle interval, and unilateral constraints are selected at both ends. k is a hyperparameter that can be adjusted according to the actual situation. This paper selects k = 0.05 in the middle interval and k = 0.01 in the two end intervals through the experiments. The specific parameters are shown in

Table 3. Characteristics of Cloud Model Parameters. [Source: own elaboration].

Natural Language	Very Important	Relatively Important	Generally Important	Secondarily Important	Unimportant
Ex	1	0.7	0.5	0.3	0
En	0.017	0.033	0.033	0.033	0.017
He	0.01	0.05	0.05	0.05	0.01

.

Figure 2 shows the cloud map of the evaluation criteria for the digital transformation capability levels of the innovation subjects. Ten experts in the field of digital innovation, digital transformation, and digital policy from enterprises (Jinan Bresee Co., Ltd.; Shandong Inspur Group; and Weihai Cyberguard Technologies Co., Ltd.), universities, and research institutes (Shandong University, Harbin University of Science and Technology, Shandong Jiaotong University, and Shandong Technology and Business University), the government (Qiqihar City and Weihai City), and intermediaries (Weihai Wisdom Valley), as well as other innovation subjects involved in the emerging industry innovation ecosystem, were selected. For the 40 indexes based on the grounded theory above, 10 experts gave natural language opinions on their importance, and the number of experts in each interval of each index was counted. Taking the index A₅ with relatively balanced expert scores as an example, the numbers of experts corresponding to “Very Important”, “Relatively Important”, “Generally Important”, and “Secondarily Important” indexes were 1, 4, 4, and 1, respectively, and the corresponding score ranges were [0.8, 1], [0.6, 0.8), [0.4, 0.6), and [0.2, 0.4). Their corresponding certainty values are shown in Figure 3. According to the method in step (4), the certainty of each score interval in A₅ was calculated as 0.983, 0.707, 0.502, and 0.305, respectively, and the final evaluation score was 0.615, the “Generally Important” score interval being [0.4, 0.6], the cloud parameter being (0.7, 0.0033, and 0.05), and the calculated coordinates of the four points being [0.709, 0.641], [0.678, 0.683], [0.730, 0.802], and [0.713, 0.840], respectively. Therefore, the maximum certainty was 0.840, and its corresponding value, A_5,2, was 0.713. The values for other score intervals were A_5,1 = 0.983, A_5,3 = 0.512, and A_5,4 = 0.269. The final evaluation values of the 40 indexes are shown in Appendix A (

Table A4. Index evaluation values of cloud model. [Source: own elaboration].

Category	Importance					Score
	[0, 0.2)	[0.2, 0.4)	[0.4, 0.6)	[0.6, 0.8)	[0.8, 1]
Procurement Digitalization, A1	1	4	5	0	0	0.632
R&D Digitalization, A2	7	3	0	0	0	0.912
Production Digitalization, A3	6	3	1	0	0	0.861
Marketing Digitalization, A4	3	4	3	0	0	0.730
Logistics Digitalization, A5	1	4	4	1	0	0.615
Organizational Digitalization, A6	2	4	2	2	0	0.650
Cultivation of Digital Transformation Professionals, A7	3	2	2	3	0	0.635
Research Direction Setting of Digital Technology such as Big Data and Artificial Intelligence, A8	0	3	4	3	0	0.502
Digital Related Research Team, A9	0	6	2	1	1	0.553
Research Digital Repository and Research Digital Tools, A10	2	6	2	0	0	0.719
Establishment of the Special Promotion Agency for Digital Achievement Transformation, A11	0	3	3	3	1	0.451
Personalized Needs of Customers, A12	0	5	3	1	1	0.534
Customer Experience Digitalization, A13	0	2	5	2	1	0.445
E-Government, A14	4	3	3	0	0	0.757
Information Sharing, A15	2	3	2	2	1	0.501
Digital Governance, A16	1	3	3	2	1	0.520
Digital Trading Platform Construction Level, A17	3	4	1	1	1	0.664
Development Level of Data Marketization Standards, A18	3	6	1	0	0	0.771
Digital Transformation Consulting Services, A19	1	3	5	1	0	0.586
Digital Technology Knowledge Mobility, A20	2	3	2	3	0	0.604
Digital Interface Extension and Compatibility, A21	4	5	1	0	0	0.796
Complementary Innovation of Digital Technology, A22	6	4	0	0	0	0.880
Digital Level of Industry–University–Research Cooperation, A23	5	5	0	0	0	0.848
Digital Market Information Matching, A24	1	8	1	0	0	0.711
Modular Design, A25	2	6	2	0	0	0.717
Data Collaboration, A26	4	5	0	1	0	0.779
Digital Empowerment, A27	2	6	1	1	0	0.696
Fiscal Policy Support for Digital Transformation, A28	1	4	4	1	0	0.617
Financial Support for Digital Transformation, A29	6	4	0	0	0	0.884
Policy Guidance for Digital Transformation, A30	6	3	1	0	0	0.863
Consumption Digitalization Level, A31	0	2	3	4	1	0.407
Digital Market Competition Level, A32	4	6	0	0	0	0.813
Digital Consumption Scenarios, A33	0	4	2	3	1	0.467
Digital Infrastructure, A34	8	2	0	0	0	0.947
Digital Technology Innovation Capability, A35	7	3	0	0	0	0.913
Digital Technology Innovation Diffusion Capability, A36	1	6	1	1	1	0.597
Digital Innovation Culture Atmosphere, A37	2	5	1	2	0	0.663
Digital Innovation Initiative, A38	0	3	2	4	1	0.430
Digital Human Resources, A39	4	4	2	0	0	0.775
Digital Investment, A40	9	1	0	0	0	0.970

).

This paper takes the upper quartile of 0.75 as the screening standard according to the general sample screening criteria [1], which is representative but not too broad and can comprehensively cover research samples and content.

As shown in Appendix A (Table A4), this paper adopts the upper quartile (0.75) as the screening standard, following the general sample screening criteria [45]. This threshold is representative and can comprehensively cover the research samples. Therefore, this paper used the upper quartile to screen out 15 influencing factors with scores higher than 0.75, namely, R&D Digitalization (A₂), Digital Interface Extension and Compatibility (A₂₁), Complementary Innovations of Digital Technologies (A₂₂), Digital Technology Innovation Capability (A₃₅), Digital Investment (A₄₀), Policy Guidance for Digital Transformation (A₃₀), Financial Support for Digital Transformation (A₂₉), Digital Market Competition Level (A₃₂), Digital Infrastructure (A₃₄), Development Level of Data Marketization Standards (A₁₈), Production Digitalization (A₃), Digital Human Resources (A₃₉), E-Government (A₁₄), Data Collaboration (A₂₆), and Digital Level of Industry–University–Research Cooperation (A₂₃). Based on these 15 influencing factors, this paper will construct an evaluation index system for the digital transformation and upgrading level of the emerging industry innovation ecosystem.

4.3. Digital Transformation and Upgrading Level Evaluation

This paper used a Likert scale to design a questionnaire based on the above 15 indexes and conducted surveys on relevant personnel in different innovation areas, such as enterprises, universities and research institutes, governments, and intermediaries in the emerging industry innovation ecosystem. A total of 550 questionnaires were distributed and 535 were recovered. To address missing or inconsistent data, we implemented two-step screening:

Missing data handling: Questionnaires with more than 10% missing items (i.e., ≥2 missing values among the 15 indexes) were excluded, as incomplete responses could bias the evaluation results. For questionnaires with ≤1 missing item, we used median imputation based on the corresponding index’s distribution in the valid sample to retain valuable data while minimizing distortion.

Inconsistent data handling: We first checked for logical inconsistencies (e.g., contradictory responses between “digital infrastructure investment” and “digital production level”) and excluded 12 such samples. Secondly, we used Cronbach’s alpha to test internal consistency (as reported later), and samples with item-total correlation <0.3 were further excluded to ensure data reliability.

After the above screening, 506 valid questionnaires were left, and the effective recovery rate was 92%. These samples were collected from innovation subjects spanning multiple key sectors within China’s emerging industries—such as advanced manufacturing, digital services, information technology, and new materials—which are characterized by rapid technological advancement and high growth potential. While the present study aims to construct a generalized evaluation framework for the digital transformation and upgrading level of these emerging industries, sectoral differences are not explicitly analyzed in this work. Addressing such sector-specific variations remains an important direction for future research.

We used IBM SPSS Statistics (Version 31) to analyze the reliability and validity of the questionnaires. Due to the large number of 15 items in the scale, Cronbach’s alpha may have lacked robustness. Therefore, this paper divided these 15 projects according to main categories. A₂, A_3, A₁₄, and A₁₈ belong to C₁, and the Cronbach’s alpha is 0.865; A₂₁, A₂₂, A₂₃, and A₂₆ belong to C₂, and the Cronbach’s alpha is 0.937; and A₂₉, A₃₀, A₃₂, A₃₄, A₃₅, A₃₉, and A₄₀ belong to C₃, and the Cronbach’s alpha is 0.943. The results indicate that they have good internal stability and consistency.Therefore, the questionnaire items were well-designed, and the instrument demonstrated adequate reliability and validity. QGA was used to optimize the projection index function. The sample dimension, which represents the number of indexes selected by the cloud model as described above, was set to 15. The number of samples was 506, and the index data were standardized. Python (Version 3.7.13) was used to program and run the algorithm. To evaluate the robustness of the QGA-based projection pursuit model, we conducted a sensitivity analysis of key parameters, including population size and the maximum number of generations. Specifically, population sizes of 200, 400, 600, 800, and 1000 and generation numbers of 100, 200, 300, 400, and 500 were tested. As shown in

Table 4. Sensitivity analysis of QGA-based projection pursuit model parameters. [Source: own elaboration].

Population Size	Generation	Evaluation Score	Standard Deviation	SVM Accuracy (%)
200	100	1.532	0.061	95.02
400	200	1.524	0.045	96.15
600	300	1.521	0.039	96.92
800	400	1.519	0.037	97.37
1000	500	1.518	0.038	97.24

, the results demonstrate that the model’s evaluation score and classification performance remained stable across the tested parameter space, with variations in the final objective value within 5% and fluctuations in SVM classification accuracy below 2%. This indicates that the model is not overly sensitive to changes in population size and generation number and thus exhibits good robustness. The bold values in the table indicate the best results for each measure. Based on these results and considering the trade-off between computational efficiency and model performance, we selected a population size of 800 and a maximum number of generations of 400 as the final parameter configuration in our experiments.

Finally, the optimal projection direction vector of the digital capability index of the innovation ecosystem is a = [0.37624874, 0.12151165, 0.03925526, 0.1467647, 0.16771306, 0.23791851, 0.16357041, 0.0379746, 0.50011435, 0.02659977, 0.48218915, 0.26340933, 0.38924595, 0.02636343, 0.04462719], so A₃₄ > A₃ > A₁₄ > A₂ > A₃₉ > A₃₀ > A₄₀ > A₂₉ > A₃₅ > A₂₁ > A₂₃ > A₂₂ > A₃₂ > A₁₈ > A₂₆, where a represents the weight of the corresponding indexes. The obtained optimal projection direction is brought into the formula, and then the comprehensive evaluation value of 506 innovation subjects can be calculated.Due to space limitations, only a subset of the comprehensive evaluation values is presented in

Table 5. Some of the comprehensive evaluation values. [Source: own elaboration].

Questionnaire	Comprehensive Evaluation Value	Questionnaire	Comprehensive Evaluation Value
1	2.20106941	502	1.63285729
2	1.74326225	503	1.65309318
3	2.05240864	504	1.57287097
4	2.1652659	505	1.88745441
5	1.92698998	506	1.73927888
……	……	……	……

.

4.4. Classification of Digital Transformation and Upgrading Levels

Based on the comprehensive evaluation value of the projection pursuit obtained above, this paper firstly used the “k-means” function in Python’s “sklearn” package to cluster the above 506 innovative subjects and divided their digital transformation and upgrading levels into three levels: budding level (I), growth level (II), and maturity level (III); secondly, the “SVC” function in python’s “sklearn” package was used to perform multi-classification training on the 15 influencing factors of 506 innovation subjects extracted from the cloud model, where the hyperparameters were grid-searched through “GridsearchCV” (the search range parameter is shown in

Table 6. “GridsearchCV” hyperparameter search range. [Source: own elaboration].

Parameter	Decision_Function_Shape	Kernel	C
Range	OVO	RBF	[1, 10, 100, 1000]
		Linear	[1, 10, 100, 1000]
	OVR	RBF	[1, 10, 100, 1000]
		Linear	[1, 10, 100, 1000]

), and the final selected hyperparameters were as follows: the penalty coefficient of the objective function was “C = ‘100’”; the kernel function was “kernel = ‘RBF’”, namely, the Gaussian kernel function; the multi-classification strategy was “decision_function_shape = ‘ovo’”; and the other parameters were set according to default settings. The data in the above 506 valid questionnaires were divided into two parts: the training data accounted for 70%, and the test data accounted for 30%. As shown in Figure 4 and Figure 5, the final training accuracy was 99.72%, and the test accuracy was 97.37%. Therefore, the classification result for “OVO” meets the experimental requirement.

In the final classification results of the 506 innovation subjects, 191 innovation subjects were at the budding level (I), 285 innovation subjects were at the growth level (II), and 30 innovation subjects were at the maturity level (III). It can be seen from the classified data that only 5.9% of the innovation subjects have digital transformation capabilities at the maturity level (III), and 94.1% of the innovation subjects are at the budding level (I) and growth level (II). Therefore, the digital transformation and upgrading level of most of China’s emerging industry innovation ecosystems is still at a relatively low level. At present, the digital infrastructure construction in China’s emerging industry innovation ecosystem is basically complete, but some innovation subjects have not yet completed their own digital upgrade. Only a small number of innovation subjects have achieved digital single connection and small connection, but the three-dimensional linkage capability is still insufficient. Therefore, it is necessary to take targeted measures to actively promote the development of data marketization standards, build various big data innovation platforms to realize the cross-border circulation and connection of data, and improve the digital upgrade speed and efficiency of China’s emerging industry innovation ecosystem.

4.5. Comparison with Other Methods

In this paper, to fairly assess our method for evaluating digital transformation and upgrading level in the emerging industry innovation ecosystem, we compared our method with two other typical prediction methods, namely, RF [41] and ANN [46]. We set the RF parameters and ANN parameters similarly to those described in [41] and [46], respectively.

As can be seen from

Table 7. Evaluation results comparison. [Source: own elaboration].

Method	Train Accuracy	Test Accuracy
RF	98.53%	94.12%
ANN	90.45%	89.73%
Our Method	99.72%	97.37%

, our method has the highest accuracy (highlighted in bold in the table) on both training data and test data, so the results confirm that the SVM is a reliable method when evaluating digital transformation and upgrading level in the emerging industry innovation ecosystem. We argue that the reason is that SVM can obtain much better results than other algorithms on small-sample training sets [44], while RF and ANN require more data.

5. Discussion and Conclusions

5.1. Discussion

This paper refers to research conclusions in the existing literature and combines qualitative and quantitative methods for analysis.

Firstly, this paper builds a digital transformation and upgrading level evaluation index system for the emerging industry innovation ecosystem. At present, the research on digital level evaluation is mainly concentrated on the construction of regional and corporate digital evaluation index systems. There are few studies on the digital transformation and upgrading level of the emerging industry innovation ecosystem. In this paper, we use a combination of quantitative and qualitative methods to build the digital transformation and upgrading level evaluation index system for the emerging industry innovation ecosystem.

Secondly, this paper adopts a quantitative method to determine indicator weights, addressing the limitations of traditional subjective or heuristic approaches such as the gray-associated model, analytic hierarchy process (AHP), and entropy weight method (EWM). Specifically, a projection pursuit model optimized by a quantum genetic algorithm (QGA) is employed to determine weights and calculate comprehensive evaluation values. This ensures greater objectivity and accuracy in the weighting process.

Thirdly, the proposed hybrid model offers practical advantages in terms of implementation cost and scalability compared to traditional methods. For example, AHP relies heavily on expert judgment, making it resource-intensive and less practical as the number of indicators grows. EWM automates weighting but depends on large amounts of high-quality data, which can be challenging to obtain in regions with limited digital infrastructure, thus limiting its applicability and adaptability. In contrast, our model employs grounded theory and the cloud model to streamline indicator selection, reducing the need for extensive expert involvement and mitigating data-quality dependency. Additionally, the modular framework combined with SVM-based classification allows efficient handling of complex and large-scale data sets, making the model scalable and adaptable to various regional contexts, including those with constrained digital infrastructure.

Fourthly, the classification and prediction methods used in this paper outperform traditional techniques. A combination of K-means and SVM was applied to classify and predict the digital transformation level of China’s emerging industry innovation ecosystem. Compared to random forest (RF) and artificial neural networks (ANNs), the proposed method achieved higher accuracy.

Therefore, this paper provides a theoretical basis and effective ways for evaluating the digital transformation and upgrading level of the emerging industry innovation ecosystem with the help of qualitative and quantitative methods and puts forward effective countermeasures and suggestions.

Given that only 5.9% of innovation subjects have reached a mature level of digital transformation, this underscores an urgent need for differentiated and stage-specific policy support. Firstly, policymakers should implement stratified guidance, focusing on targeted digital empowerment for entities in the nascent and growth phases through subsidies, tax incentives, capacity-building training, and enhanced platform access. Secondly, the innovation ecosystem should actively promote benchmark cases of mature digital transformation to create demonstration effects and facilitate learning spillovers. Finally, stronger collaboration between national and local governments is essential to improve cross-sector digital infrastructure and establish inter-organizational data-sharing platforms, which collectively accelerate digital maturity across the ecosystem. These policy priorities will help close the digital maturity gap, especially in early-stage ecosystems, by addressing both capability building and enabling environments.

Based on the above understanding, this paper puts forward the following specific countermeasures and suggestions to promote the digital transformation and upgrading of the emerging industry innovation ecosystem:

(1) Accelerate the digital transformation and upgrade process of various innovation subjects. The digitalization of the entire innovation ecosystem can be accelerated only if innovation subjects complete their own digital transformations and upgrades. Specifically:

Enterprises: improve awareness and capabilities regarding digital upgrades; increase digital innovation investment in procurement, R&D, production, marketing, logistics, and other business links; establish an open learning organization model; and realize digital office and digital R&D processes to meet the needs of the digital market.

Universities and research institutes: act as digital innovation think tanks by cultivating digital transformation professionals; set up digital technology research directions (e.g., big data and artificial intelligence); build digital resource banks; purchase digital research tools; and establish specialized promotion agencies for digital achievement transformation.

Customers: actively participate in the digital production and service process to meet the digital transformation and upgrade needs of the supply side.

Government: play a catalytic role in industrial digital innovation; actively build a digital government; and improve the ability and level of e-government and digital governance.

Intermediaries: build digital trading platforms and markets; promote the construction of data marketization standards and the cross-border flow of data; and provide consulting services for digital transformation to promote their own digital upgrading.

(2) Strengthen the construction of digital R&D innovation cooperation platforms. Innovation subjects should actively engage in digital collaboration to establish horizontal and vertical alliances for digital technology research and development. Specifically:

Enterprises: leverage their initiative to participate in digital R&D alliances; adopt big data platforms, digital industry–university–research cooperation platforms, and blockchain technology platforms; and promote digital empowerment to enhance innovation efficiency.

Universities and research institutes: build upon existing R&D platforms to conduct digital technology research; share resources and knowledge; and facilitate collaboration with enterprises and laboratories.

Key laboratories and leading enterprises: provide infrastructure and technical support for digital R&D platforms; open up “data silos”; and enable the free circulation and effective matching of data among innovation subjects.

(3) Improve the digitalization level of the innovation environment and leverage its digital empowerment role. Specifically:

Government: actively introduce policies to support digital transformation and create a favorable policy environment for innovation.

Supply side: implement digital reforms that increase consumers’ demand for digital innovation and enhance the level of digital consumption.

Cultural environment: strengthen the cultural atmosphere of digital innovation to increase the enthusiasm and engagement of innovation subjects.

Innovation subjects and environment integration: achieve organic integration, positive interaction, and collaborative symbiosis between innovation subjects and the innovation environment; promote the digital transformation and upgrading of the emerging industry innovation ecosystem; and realize multi-dimensional development of digital innovation.

Moreover, although the evaluation model is constructed based on empirical data from Chinese innovation subjects, the methodological framework—including grounded theory, the cloud model, the projection pursuit model, and the combination of K-means and SVM—is generalizable and data-driven. Therefore, it can be transferred to other regions with different industrial environments, provided that localized data are collected to recalibrate the index system. Future research can further explore the cross-regional applicability of the model through empirical validation in different national or regional contexts.

5.2. Conclusions

This paper provides a theoretical basis and effective method for evaluating the digital transformation and upgrading level of the emerging industry innovation ecosystem with the help of qualitative and quantitative methods.

First, this paper studies the digital transformation and upgrading level of emerging industry from the perspective of the innovation ecosystem. It is a new attempt to measure and analyze the digital transformation and upgrading level of the emerging industry innovation ecosystem and enriches and expands the theoretical research on digital transformation and upgrading. The existing research on digital transformation and upgrading mainly focuses on micro-enterprises and traditional industry. However, the global emerging industry competition paradigm has shifted from competition among enterprises to competition among innovation ecosystems. The digital transformation of a single enterprise is inseparable from its interaction with other innovation subjects. Therefore, the innovation ecosystem is a suitable perspective from which to study the digital transformation and upgrading of the emerging industry.

Secondly, this paper refines and expands the factors influencing the digital transformation and upgrading of the emerging industry innovation ecosystem and constructs a corresponding evaluation index system. Based on grounded theory and the cloud model, key factors affecting digital transformation and upgrading are identified. According to the projection pursuit results, the importance of these indexes, ranked from high to low, is as follows: digital infrastructure, production digitalization, e-government, R&D digitalization, digital human resources, policy guidance for digital transformation, digital investment, financial support, digital technology innovation capability, digital interface extension and compatibility, industry–university–research cooperation level, complementary innovation of digital technology, digital market competition, development of data marketization standards, and data collaboration.

Finally, this paper classifies the digital transformation and upgrading levels of China’s emerging industry innovation ecosystems and puts forward corresponding countermeasures and suggestions. The analysis results based on K-means and SVM indicate that most innovation ecosystems in China remain at the embryonic and growth stages. Challenges persist in areas such as innovation subject digitalization, ecosystem connectivity, and the innovation environment. Targeted improvement measures should be implemented in accordance with the evaluation results to effectively guide digital transformation and upgrading. Meanwhile, the proposed models and methods have been shown to effectively help emerging industries identify their digital maturity levels and provide a valuable reference tool for evaluating digital development and planning digital upgrade pathways.

Furthermore, although this study does not focus on a specific named innovation ecosystem, the comprehensive evaluation of 506 innovation subjects in China offers empirical evidence for assessing the sustainability of emerging industry innovation ecosystems. The results show that only 5.9% of these subjects are at a mature stage, while the remaining 94.1% are still in the budding or growth stages. This suggests that the overall digital transformation and upgrading of China’s emerging industry innovation ecosystems remain at an early to intermediate stage. Insufficient interconnection among innovation actors, limited diffusion of digital technologies, and underdeveloped innovation environments hinder the ecosystems’ ability to evolve sustainably. Therefore, targeted measures are urgently needed to enhance system-level collaboration, strengthen digital infrastructure, and promote cross-actor data sharing to support long-term sustainability. Building on the hybrid evaluation model developed in this study, future research will apply this framework to specific real-world cases—such as regional innovation clusters or the new energy vehicle industry—to conduct more targeted assessments and generate practical insights for enhancing ecosystem sustainability and performance.

5.3. Limitations

Several limitations are recognized in this study, which include the following three aspects:

Firstly, although the grounded theory and cloud model support the scientific selection of influencing factors, there are still theoretical limitations in applying grounded theory in this context. The process of open coding and categorization is heavily dependent on expert interpretation, which may introduce subjectivity, despite efforts such as expert validation and saturation testing. Furthermore, grounded theory is inherently context-specific and may not fully capture dynamic, latent, or interdisciplinary aspects of digital transformation—such as digital strategy, patent activity, innovation efficiency, and data security—which are also crucial to the development of emerging industry innovation ecosystems.

Secondly, the conversion from qualitative insights to quantitative indexes may involve a degree of abstraction and simplification. While the integration of grounded theory with the cloud model and projection pursuit enhances the analytical rigor, the methodological transition from qualitative constructs to computable metrics may result in partial loss of nuance or misalignment between theory and measurement. Future studies could consider combining grounded theory with data-driven empirical models or machine learning techniques to further bridge this gap.

Thirdly, the scope of research samples selected in this paper is limited. Although the study includes innovation subjects from enterprises, universities, research institutes, governments, and intermediaries, the sample size and geographic coverage remain limited.

Future research could improve the generalizability and robustness of the evaluation results by expanding the sample pool and diversifying data sources, such as integrating regional statistical indexes or conducting cross-country comparisons.