Enhancing Real Estate Valuation in Kazakhstan: Integrating Machine Learning and Adaptive Neuro-Fuzzy Inference System for Improved Precision

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article is interesting. It concerns real estate valuation. Real estate plays an important role in every developed country's economy and has a significant impact on its functioning. This article examines the forecasting of real estate prices using various methodologies. Statistical models, data visualization, regression analysis, neural networks, and adaptive neuro-fuzzy inference system (ANFIS) were used. The study was based on a dataset of apartment sales collected from websites in Kazakhstan. More than 145 thousand observations were collected. The data collected included various characteristics of apartments, including price per square meter, number of rooms, total price, location, living area in square meters, main construction material, floor number, total number of floors in the building, name of the residential complex, year of commissioning, condition of the apartment, mortgage status, presence of balcony, type of balcony (glazed or not), type of door, type of floor, ceiling height, whether the property was a former hostel, possibility of exchange, type of internet access, parking, amenities, landline telephone connection, type of toilet, security features, others.

The article is quite well written, but it can be improved. The summary contains concise information about the study. The introduction correctly describes previous studies on various analyses of real estate, in different countries, using different methods based on different indicators, factors, and measures. The introduction lacks what is presented in the individual points and sub-points of the article to prepare the reader to follow the authors' approach. I recommend supplementing it. In the reviewer's opinion, some materials and methods also require supplementing. It should be explained why this data set was studied. It seems that it should be grouped in some way. It is known that the price of real estate depends most on the location. Apartment prices in Astana will be completely different than in smaller cities. Probably, apartment prices in individual districts of Astana will also differ. The further from the center, the prices will decrease. In the case of apartment square footage, it seems that there should also be a significant difference. Small apartments will be more expensive than larger ones. This should be explained in the study. In point 3, the results estimated using regression, neural networks, using an adaptive neuro-fuzzy inference system are correctly described. The results were presented correctly in a descriptive form and using graphs. The discussion part was also presented correctly. According to the Reviewer, a few sentences about the limitations of this study were missing. In the Reviewer's opinion, a point about possible errors in the study could be added in this part. The authors of this article should also indicate directions for future research in the context of this article. The conclusions were formulated correctly.

The study is interesting, includes relevant literature, and addresses an engaging problem, but requires supplementation. I propose accepting the article for publication after the necessary revisions are made.

Author Response

Comments:

The article is interesting. It concerns real estate valuation. Real estate plays an important role in every developed country's economy and has a significant impact on its functioning. This article examines the forecasting of real estate prices using various methodologies. Statistical models, data visualization, regression analysis, neural networks, and adaptive neuro-fuzzy inference system (ANFIS) were used. The study was based on a dataset of apartment sales collected from websites in Kazakhstan. More than 145 thousand observations were collected. The data collected included various characteristics of apartments, including price per square meter, number of rooms, total price, location, living area in square meters, main construction material, floor number, total number of floors in the building, name of the residential complex, year of commissioning, condition of the apartment, mortgage status, presence of balcony, type of balcony (glazed or not), type of door, type of floor, ceiling height, whether the property was a former hostel, possibility of exchange, type of internet access, parking, amenities, landline telephone connection, type of toilet, security features, others.

The article is quite well written, but it can be improved. The summary contains concise information about the study. The introduction correctly describes previous studies on various analyses of real estate, in different countries, using different methods based on different indicators, factors, and measures. The introduction lacks what is presented in the individual points and sub-points of the article to prepare the reader to follow the authors' approach. I recommend supplementing it. In the reviewer's opinion, some materials and methods also require supplementing. It should be explained why this data set was studied. It seems that it should be grouped in some way. It is known that the price of real estate depends most on the location. Apartment prices in Astana will be completely different than in smaller cities. Probably, apartment prices in individual districts of Astana will also differ. The further from the center, the prices will decrease. In the case of apartment square footage, it seems that there should also be a significant difference. Small apartments will be more expensive than larger ones. This should be explained in the study. In point 3, the results estimated using regression, neural networks, using an adaptive neuro-fuzzy inference system are correctly described. The results were presented correctly in a descriptive form and using graphs. The discussion part was also presented correctly. According to the Reviewer, a few sentences about the limitations of this study were missing. In the Reviewer's opinion, a point about possible errors in the study could be added in this part. The authors of this article should also indicate directions for future research in the context of this article. The conclusions were formulated correctly.

The study is interesting, includes relevant literature, and addresses an engaging problem, but requires supplementation. I propose accepting the article for publication after the necessary revisions are made.

Response:

1) We add some information about individual points and sub-points of the article in introduction section:

In this study [35], the authors employed several machine learning 123 algorithms to predict property values such as: Linear Regression, Decision Trees, Random 124 Forests, Support Vector Machines.

2) We also explain why this data set was studied:

The examination of this comprehensive dataset of apartment sales in Kazakhstan serves multiple critical objectives. Primarily, it facilitates a deeper understanding of the dynamics within the real estate market in Kazakhstan, encompassing price trends, demand patterns, and the influence of diverse apartment features on pricing. This granular analysis aids stakeholders, including investors, developers, and policymakers, in making well-informed decisions. Furthermore, real estate prices and sales volumes are indicative of economic vitality, thus, analyzing such data provides insights into the region's economic status and growth prospects. Detailed insights into buyer preferences—such as the importance of location, building materials, and features like balconies and parking facilities—inform marketing strategies and development plans. The development of predictive models for apartment pricing, which incorporate multiple variables, proves beneficial for buyers, sellers, and real estate agents in understanding fair market values and anticipating future price trends. This study also provides empirical insights to local governments and urban planners, supporting the enhancement of urban planning, housing policies, and regulations based on the current housing conditions, offered amenities, and building states. The employment of a custom scraper and sophisticated data analysis techniques underscores the role of technology in the collection and processing of real estate data, potentially setting a benchmark for analogous studies in different locales or sectors. By gathering and scrutinizing a vast array of characteristics across a significant number of observations, the research aims to provide a detailed portrait of Kazakhstan's housing market, thus delivering invaluable knowledge to various stakeholders engaged in the real estate domain.

3) We add some information about paper’s limitations, possible errors, and future research:

The analysis of the dataset employed for investigating apartment sales in Kazakhstan is ostensibly thorough, however, certain potential limitations were not explicitly articulated within the document. The data was extracted at a singular temporal point, which may not encapsulate seasonal fluctuations or longitudinal trends within the real estate market. Although the dataset encompasses a broad spectrum of apartment characteristics, it potentially excludes critical factors such as the economic vitality of the neighborhood, prospective development initiatives, and additional socio-economic indicators that might significantly influence apartment prices. A comprehensive understanding of these limitations is imperative for accurate interpretation of the study's findings and could provide a foundation for future research aimed at addressing these deficiencies.

Future research endeavors predicated on the analysis of the Kazakhstan real estate market ought to contemplate several advanced and nuanced avenues to enrich our comprehension and augment the precision of market forecasts. Prospective studies should endeavor to collect data across multiple temporal junctures to scrutinize seasonal and cyclical effects on the real estate market, thereby facilitating the development of more dynamic and robust forecasting models. Moreover, broadening the dataset to encompass additional regions within Kazakhstan or facilitating comparisons with international markets could yield deeper insights into regional disparities and the influence of global economic trends on local real estate dynamics. Furthermore, the integration of socio-economic indicators, such as income levels, demographic transitions, and economic policies, could significantly enhance the model’s predictive capabilities by correlating real estate trends with broader macroeconomic conditions. Pursuing these research directions could profoundly augment our understanding of real estate dynamics, consequently leading to more informed decision-making and policy development.

Reviewer 2 Report

Comments and Suggestions for Authors

This paper investigates the Kazakhstan real estate sector by employing various forecasting methodologies, particularly focusing on the use of Adaptive Neuro-Fuzzy Inference Systems (ANFIS). It explores how fair value principles, as defined by International Financial Reporting Standards (IFRS) and US Generally Accepted Accounting Principles (GAAP), are applied to real estate asset valuation. The study highlights the challenges of real estate price forecasting, considering its critical role in both manufacturing and development sectors. By integrating statistical models, regression analysis, neural networks, and ANFIS, this paper aims to provide insights into more accurate real estate management practices.

 The real estate sector remains a crucial aspect of both corporate and development industries, and this paper provides a timely analysis of how fair value principles are applied in the Kazakhstan market. The focus on the ongoing challenges in real estate valuation is relevant for both academic and professional audiences. The inclusion of advanced forecasting techniques such as Adaptive Neuro-Fuzzy Inference Systems (ANFIS) sets this paper apart. ANFIS is a hybrid approach that combines neural networks and fuzzy logic to handle uncertainties and complexities, making it a suitable method for real estate price forecasting.

Here are some suggestions for improvement:

- While the methodologies are well-discussed, the paper could benefit from a more detailed explanation of the data sources used for analysis, especially the Kazakhstan real estate sector data. Providing insights into the size, range, and specifics of the data (e.g., geographical scope, time period) would strengthen the validity of the conclusions.

- The paper highlights ANFIS as the main method, but the relative advantages of ANFIS compared to traditional statistical models and regression analyses should be discussed in greater detail. How much improvement does ANFIS bring compared to these other methods? A clearer comparison of accuracy metrics (such as Mean Absolute Error or Root Mean Square Error) across the methods would be helpful.

- Fuzzy logic approach is a very useful algorithm, more relevant work application of this method for prediction should be introduced, such as: https://doi.org/10.1007/s00170-020-06394-4

- While the theoretical analysis of forecasting methodologies is thorough, it would be beneficial to delve more deeply into the practical implications of this research. For example, how can real estate managers or investors in Kazakhstan apply these insights in day-to-day operations or long-term decision-making?

- Real estate prices are influenced by various macroeconomic factors (e.g., inflation, interest rates, government policies). While the paper focuses on predictive models, adding a discussion on the role of these economic factors in the Kazakhstan real estate sector could provide context to the price movements and enhance the comprehensiveness of the research.

- The paper briefly mentions the application of IFRS and GAAP to real estate asset valuation. A deeper discussion on the adaptability and potential challenges of implementing these standards in Kazakhstan’s real estate market would enrich the analysis, considering the specificities of emerging markets.

Overall, the paper presents an innovative approach to analyzing real estate prices in Kazakhstan through a combination of advanced forecasting techniques, with a focus on ANFIS. It provides valuable insights into how fair value principles (IFRS, GAAP) apply to real estate asset management. The comprehensive use of multiple forecasting methodologies, particularly the integration of ANFIS, sets the paper apart as a novel contribution to real estate forecasting research.

Author Response

Comments: 

This paper investigates the Kazakhstan real estate sector by employing various forecasting methodologies, particularly focusing on the use of Adaptive Neuro-Fuzzy Inference Systems (ANFIS). It explores how fair value principles, as defined by International Financial Reporting Standards (IFRS) and US Generally Accepted Accounting Principles (GAAP), are applied to real estate asset valuation. The study highlights the challenges of real estate price forecasting, considering its critical role in both manufacturing and development sectors. By integrating statistical models, regression analysis, neural networks, and ANFIS, this paper aims to provide insights into more accurate real estate management practices.

 The real estate sector remains a crucial aspect of both corporate and development industries, and this paper provides a timely analysis of how fair value principles are applied in the Kazakhstan market. The focus on the ongoing challenges in real estate valuation is relevant for both academic and professional audiences. The inclusion of advanced forecasting techniques such as Adaptive Neuro-Fuzzy Inference Systems (ANFIS) sets this paper apart. ANFIS is a hybrid approach that combines neural networks and fuzzy logic to handle uncertainties and complexities, making it a suitable method for real estate price forecasting.

Here are some suggestions for improvement:

- While the methodologies are well-discussed, the paper could benefit from a more detailed explanation of the data sources used for analysis, especially the Kazakhstan real estate sector data. Providing insights into the size, range, and specifics of the data (e.g., geographical scope, time period) would strengthen the validity of the conclusions.

- The paper highlights ANFIS as the main method, but the relative advantages of ANFIS compared to traditional statistical models and regression analyses should be discussed in greater detail. How much improvement does ANFIS bring compared to these other methods? A clearer comparison of accuracy metrics (such as Mean Absolute Error or Root Mean Square Error) across the methods would be helpful.

- Fuzzy logic approach is a very useful algorithm, more relevant work application of this method for prediction should be introduced, such as: https://doi.org/10.1007/s00170-020-06394-4

- While the theoretical analysis of forecasting methodologies is thorough, it would be beneficial to delve more deeply into the practical implications of this research. For example, how can real estate managers or investors in Kazakhstan apply these insights in day-to-day operations or long-term decision-making?

- Real estate prices are influenced by various macroeconomic factors (e.g., inflation, interest rates, government policies). While the paper focuses on predictive models, adding a discussion on the role of these economic factors in the Kazakhstan real estate sector could provide context to the price movements and enhance the comprehensiveness of the research.

- The paper briefly mentions the application of IFRS and GAAP to real estate asset valuation. A deeper discussion on the adaptability and potential challenges of implementing these standards in Kazakhstan’s real estate market would enrich the analysis, considering the specificities of emerging markets.

Overall, the paper presents an innovative approach to analyzing real estate prices in Kazakhstan through a combination of advanced forecasting techniques, with a focus on ANFIS. It provides valuable insights into how fair value principles (IFRS, GAAP) apply to real estate asset management. The comprehensive use of multiple forecasting methodologies, particularly the integration of ANFIS, sets the paper apart as a novel contribution to real estate forecasting research.

Response:

We add some text to the paper:

1) The dataset utilized for studying apartment sales in Kazakhstan is notably expansive and detailed, covering a broad geographical scope and a wide range of property characteristics. The dataset comprises a significant total of 145,188 observations. It includes data from various locations across Kazakhstan, providing a comprehensive view of the real estate market dynamics within the country.

2) We adding 40th reference to paper:

Li, W.; Zhang, L.; Chen, X.; Wu, C.; Cui, Z.; Niu, C. Quality evaluation Predicting the evolution of sheet metal surface scratching by the technique of artificial intelligence. International Journal of Advanced Manufacturing Technology 2021, 112, 853--865.

Reviewer 3 Report

Comments and Suggestions for Authors

 

First of all, according to the main research content and objectives of the article, the title is not very accurate, and it is recommended to modify it. At the same time, the writing of the abstract does not conform to the general norms of scientific papers, so it is suggested to rewrite according to the general paradigm.

In the introduction part, the paper presents the research background to a certain extent, but it is not detailed enough and the problem statement is not clear enough. It is suggested that the research questions and objectives should be put forward directly. Of course, the author has made a detailed review of relevant research in this field, but the sorting and refining of rap research is not clear enough. It is suggested that the research gap should be put forward clearly.

In the part of materials and methods, this paper mainly introduces the data collection and cleaning process, as well as the corresponding descriptive statistics and correlation analysis. But the methods section is missing. It is recommended to separate data source and methods into two separate sections, which will be clearer and easier to read. The effects of Figures 1 and 2, in my opinion, are somewhat repetitive and reflect similar information.

For the results section, what is the relationship between the findings from the regression learner models, neural networks, and the ANFIS model? Meanwhile, the discussion part is too general and fails to correspond with the results part. It is suggested to improve this part. In addition, the authors could further enhance the discussion by comparing the ANFIS model's performance with other models in similar studies and discussing potential limitations.

Author Response

Comments: 

First of all, according to the main research content and objectives of the article, the title is not very accurate, and it is recommended to modify it. At the same time, the writing of the abstract does not conform to the general norms of scientific papers, so it is suggested to rewrite according to the general paradigm.

In the introduction part, the paper presents the research background to a certain extent, but it is not detailed enough and the problem statement is not clear enough. It is suggested that the research questions and objectives should be put forward directly. Of course, the author has made a detailed review of relevant research in this field, but the sorting and refining of rap research is not clear enough. It is suggested that the research gap should be put forward clearly.

In the part of materials and methods, this paper mainly introduces the data collection and cleaning process, as well as the corresponding descriptive statistics and correlation analysis. But the methods section is missing. It is recommended to separate data source and methods into two separate sections, which will be clearer and easier to read. The effects of Figures 1 and 2, in my opinion, are somewhat repetitive and reflect similar information.

For the results section, what is the relationship between the findings from the regression learner models, neural networks, and the ANFIS model? Meanwhile, the discussion part is too general and fails to correspond with the results part. It is suggested to improve this part. In addition, the authors could further enhance the discussion by comparing the ANFIS model's performance with other models in similar studies and discussing potential limitations.

Response: 

We made some corrections in abstract section:

This study addresses the challenge of accurately determining the fair market value of real estate in Kazakhstan, leveraging a multi-methodological approach that encompasses statistical models, regression analysis, data visualization, neural networks, and particularly, an Adaptive Neuro-Fuzzy Inference System (ANFIS). The integration of these diverse methodologies not only enhances the robustness of real estate valuation but also introduces new insights into effective asset management. The findings suggest that ANFIS provides superior precision in real estate pricing, demonstrating its potential as a valuable tool for strategic management and investment decision-making.

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

In this version, the author's modification is too simple and lacks sufficient responses to the relevant comments. It is suggested to further respond to the previous comments one by one. Although many valuation methods are integrated in this study, different from technical reports, scientific papers are not just a simple hybridisation of technology and methods. More important is the contribution to the research problem and theory, to which this research needs to be strengthened.

1. The research title does not fit well with the theme of this research. This research is not an analysis of the real estate sector, nor of organizations.

2. Research questions need to be clearly presented, not left to the reader to find out.

3. What is the gap that needs to be solved in this study? Where is the theoretical contribution?

4. The methods part is still missing, so I suggest you supplement it.

5. Compared with the presentation of results, the discussion of this study is too simple, and the suggestions correspond to the results section.

Author Response

Comments 1: The research title does not fit well with the theme of this research. This research is not an analysis of the real estate sector, nor of organizations.
Response 1: We changed the title of the article:
Enhancing Real Estate Valuation in Kazakhstan: Integrating Machine Learning and ANFIS for Improved Precision

Comments 2: Research questions need to be clearly presented, not left to the reader to find out.

Response 2: We add research questions to introduction section.

Comments 3: What is the gap that needs to be solved in this study? Where is the theoretical contribution?

Response 3: We add information to introduction section.

Comments 4: The methods part is still missing, so I suggest you supplement it.

Response 4: We rewrote methods section.

Comments 5: Compared with the presentation of results, the discussion of this study is too simple, and the suggestions correspond to the results section.

Response 5: We rewrote discussion section.

Author Response File: Author Response.pdf