Spatial Differentiation Effect of Rural Logistics in Urban Agglomerations in China Based on the Fuzzy Neural Network
Abstract
:1. Introduction
2. Literature Review
2.1. The Importance of Logistics
2.2. Evaluation of the Logistics Development
2.3. Rural Logistics Related Research
3. The Rural Logistics Index System under the Context of Urban Agglomerations
4. Methodology
4.1. The T-S Fuzzy Neural Network
4.1.1. The Construction of Fuzzy Neural Network
4.1.2. The Algorithm of T-S Fuzzy Neural Network
4.2. Moran’s I
4.3. Kernel Density
5. Results
5.1. The Study Area and Data Sources
5.2. Evaluation Criteria of Rural Logistics in Urban Agglomeration
5.3. The Comprehensive Index of Rural Logistics in Chengdu-Chongqing Agglomeration
5.4. Spatial Distribution of Rural Logistics in Chengdu-Chongqing Agglomeration
6. Discussion
6.1. Differentiation Characteristics of Rural Logistics in Urban Agglomeration
6.2. Dynamic Evolution of Rural Logistics Distribution in Urban Agglomeration
6.3. Comparison between the Fuzzy Neural Network and the Traditional BP Neural Network
7. Conclusions
Author Contributions
Funding
Data Availability Statement
Conflicts of Interest
Appendix A
Fuzzy Neural Network | BP Neural Network |
---|---|
0.1071 | −0.5548 |
0.0496 | −0.4788 |
−0.037 | −0.0296 |
−0.092 | −0.5013 |
−0.0534 | −0.5449 |
−0.0088 | −0.4233 |
0.0722 | −0.5412 |
0.0102 | −0.5409 |
−0.1002 | −0.4441 |
−0.0581 | −0.5626 |
0.0681 | −0.2662 |
−0.0818 | −0.4952 |
−0.0854 | −0.7700 |
0.0997 | −0.4294 |
−0.0737 | −0.5409 |
−0.0072 | −0.5441 |
0.1171 | −0.5681 |
−0.0458 | −0.1517 |
0.0872 | −0.4917 |
−0.1166 | 0.0002 |
0.1282 | −0.4068 |
−0.1385 | −0.5474 |
−0.0902 | 0.1285 |
−0.0392 | −0.4415 |
0.1423 | 0.1501 |
−0.0283 | −0.5636 |
−0.0952 | −0.4182 |
0.076 | −0.8079 |
−0.112 | −0.6168 |
−0.076 | −0.7269 |
0.0947 | −0.3827 |
0.1146 | −0.8120 |
0.0311 | −0.4626 |
0.0597 | −0.5409 |
0.0414 | −0.7154 |
0.0366 | −0.5428 |
−0.0441 | −0.1885 |
−0.0264 | −0.5097 |
0.0972 | −0.5462 |
0.0634 | −0.4844 |
−0.0664 | −0.0779 |
0.1096 | −0.4576 |
−0.0865 | −0.6770 |
0.1183 | −0.5409 |
−0.0349 | 0.1628 |
−0.0516 | −0.4653 |
0.0797 | 0.0226 |
−0.0749 | −0.5415 |
−0.0629 | −0.3301 |
−0.1401 | 0.1511 |
References
- Li, L.; Ma, S.G.; Zheng, Y.L.; Xiao, X.Y. Integrated regional development: Comparison of urban agglomeration policies in China. Land Use Policy 2022, 114, 105939. [Google Scholar] [CrossRef]
- Oxford University Press. Our Common Future. In WCED-World Commission on Environment and Development; Oxford University Press: Oxford, UK, 1987. [Google Scholar]
- Shen, Z.C.; Xu, X.L.; Xu, S.W.; Sun, D.W. A comparative study of land development patterns and regional thermal environments (RTEs) in typical urban agglomerations of China and America: A case study of Beijing-Tianjin-Hebei (BTH) and Boswash. Sci. Total Environ. 2022, 803, 149735. [Google Scholar] [CrossRef]
- Muñoz-Villamizar, A.; Solano-Charris, E.L.; AzadDisfany, M.; Reyes-Rubiano, L. Study of urban-traffic congestion based on Google Maps API: The case of Boston. IFAC-PapersOnLine 2021, 54, 211–216. [Google Scholar] [CrossRef]
- Peters, A.; Hernández, D.; Kioumourtzoglou, M.-A.; Johnson, M.A.; Chillrud, S.N.; Hilpert, M. Assessing neighborhood-scale traffic from crowd-sensed traffic data: Findings from an environmental justice community in New York City. Environ. Sci. Policy 2022, 133, 155–163. [Google Scholar] [CrossRef]
- Dubie, M.; Kuo, K.C.; Giron-Valderrama, G.; Goodchild, A. An evaluation of logistics sprawl in Chicago and Phoenix. J. Transp. Geogr. 2020, 88, 102298. [Google Scholar] [CrossRef]
- Rice, M.; Sorenson, M.; Aversa, J. The geography of lifestyle center growth: The emergence of a retail cluster format in the United States. J. Retail. Consum. Serv. 2022, 65, 102835. [Google Scholar] [CrossRef]
- Sakai, T.; Kawamura, K.; Hyodo, T. Spatial reorganization of urban logistics system and its impacts: Case of Tokyo. J. Transp. Geogr. 2017, 60, 110–118. [Google Scholar] [CrossRef]
- Bartle, C.; Chatterjee, K. Employer perceptions of the business benefits of sustainable transport: A case study of peri-urban employment areas in South West England. Transp. Res. Part A Policy Pract. 2019, 126, 297–313. [Google Scholar] [CrossRef]
- Raimbault, N. From regional planning to port regionalization and urban logistics. The inland port and the governance of logistics development in the Paris region. J. Transp. Geogr. 2019, 78, 205–213. [Google Scholar] [CrossRef]
- Song, Y.; Yeung, G.; Zhu, D.L.; Zhang, L.X.; Xu, Y.; Zhang, L.Y. Efficiency of logistics land use: The case of Yangtze River Economic Belt in China, 2000–2017. J. Transp. Geogr. 2020, 88, 102851. [Google Scholar] [CrossRef]
- Yang, X.C.; Wang, M.J. Diversification and Spatial Differentiation of Villages’ Functional Types in the New Period of China: Results from Hierarchical Urban-Rural Spatial Relations and Townships Size. Land 2022, 11, 171. [Google Scholar] [CrossRef]
- Hauge, J.B.; Birkie, S.E.; Jeong, Y. Developing a holistic decision support framework: From production logistics to sustainable freight transport in an urban environment. Transp. Res. Interdiscip. Perspect. 2021, 12, 100496. [Google Scholar] [CrossRef]
- Büyüközkan, G.; Ilıcak, Ö. Smart urban logistics: Literature review and future directions. Socio-Econ. Plan. Sci. 2021, 81, 101197. [Google Scholar] [CrossRef]
- Liu, R. Incomplete Urbanization and the Trans-Local Rural-Urban Gradient in China: From a Perspective of New Economics of Labor Migration. Land 2022, 11, 282. [Google Scholar] [CrossRef]
- Lan, S.L.; Tseng, M.-L.; Yang, C.; Huisingh, D. Trends in sustainable logistics in major cities in China. Sci. Total Environ. 2020, 712, 136381. [Google Scholar] [CrossRef]
- Aydın, G.T. Kent içi lojistik: İstanbul için bir uygulama; Istanbul Technical University: Maslak, Turkey, 2017. [Google Scholar]
- Lagorio, A.; Pinto, R.; Golini, R. Urban Logistics Ecosystem: A system of system framework for stakeholders in urban freight transport projects. IFAC-PapersOnLine 2017, 50, 7284–7289. [Google Scholar] [CrossRef]
- Strulak-Wójcikiewicz, R.; Wagner, N. Exploring opportunities of using the sharing economy in sustainable urban freight transport. Sustain. Cities Soc. 2021, 68, 102778. [Google Scholar] [CrossRef]
- Gong, X.W. Coupling coordinated development model of urban-rural logistics and empirical study. Math. Probl. Eng. 2019, 2019, 1–12. [Google Scholar] [CrossRef] [Green Version]
- Filippi, A.C.G.; Guarnieri, P.; Carvalho, J.M.; Reis, S.A.; da Cunha, C.A. New configurations in Brazilian agribusiness: Rural warehouse condominiums. J. Agribus. Dev. Emerg. Econ. 2020, 10, 41–63. [Google Scholar] [CrossRef]
- Liu, W. Route optimization for last-mile distribution of rural E-commerce logistics based on ant colony optimization. IEEE Access 2020, 8, 12179–12187. [Google Scholar] [CrossRef]
- Morel, M.; Balm, S.; Berden, M.; van Amstel, W.P. Governance models for sustainable urban construction logistics: Barriers for collaboration. Transp. Res. Procedia 2020, 46, 173–180. [Google Scholar] [CrossRef]
- Taniguchi, E.; Noritake, M.; Yamada, T.; Izumitani, C. Optimal size and location planning of public logistics terminals. Transp. Res. Part E Logist. Transp. Rev. 1999, 35, 207–222. [Google Scholar] [CrossRef]
- Iwan, S.; Kijewska, K. The integrated approach to adaptation of good practices in urban logistics based on the Szczecin example. Procedia-Soc. Behav. Sci. 2014, 125, 212–225. [Google Scholar] [CrossRef] [Green Version]
- Schliwa, G.; Armitage, R.; Aziz, S.; Evans, J.; Rhoades, J. Sustainable city logistics—Making cargo cycles viable for urban freight transport. Res. Transp. Bus. Manag. 2015, 15, 50–57. [Google Scholar] [CrossRef] [Green Version]
- Rześny-Cieplińska, J.; Szmelter-Jarosz, A.; Moslem, S. Priority-based stakeholders analysis in the view of sustainable city logistics: Evidence for Tricity, Poland. Sustain. Cities Soc. 2021, 67, 102751. [Google Scholar] [CrossRef]
- Russo, S.M.; Voegl, J.; Hirsch, P. A multi-method approach to design urban logistics hubs for cooperative use. Sustain. Cities Soc. 2021, 69, 102847. [Google Scholar] [CrossRef]
- Kang, X.Y.; Pang, H.Y. City selection for fresh produce e-commerce’s market entry strategy: Based on the perspective of urban logistics competitiveness. Transp. Res. Interdiscip. Perspect. 2022, 13, 100537. [Google Scholar] [CrossRef]
- Morfoulaki, M.; Kotoula, K.; Stathacopoulos, A.; Mikiki, F.; Aifadopoulou, G. Evaluation of specific policy measures to promote sustainable urban logistics in small-medium sized cities: The case of Serres, Greece. Transp. Res. Procedia 2016, 12, 667–678. [Google Scholar] [CrossRef] [Green Version]
- Viana, M.S.; Delgado, J.P.M. City logistics in historic centers: Multi-criteria evaluation in GIS for city of Salvador (Bahia–Brazil). Case Stud. Transp. Policy 2019, 7, 772–780. [Google Scholar] [CrossRef]
- Juvvala, R.; Sarmah, S.P. Evaluation of policy options supporting electric vehicles in city logistics: A case study. Sustain. Cities Soc. 2021, 74, 103209. [Google Scholar] [CrossRef]
- Zhao, L.J.; Zhou, J.P.; Li, H.Y.; Yang, P.L.; Zhou, L.X. Optimizing the design of an intra-city metro logistics system based on a hub-and-spoke network model. Tunn. Undergr. Space Technol. 2021, 116, 104086. [Google Scholar] [CrossRef]
- de Oliveira, L.K.; Lessa, D.A.; Oliveira, E.; Calazans, B.F.G. Multi-agent modelling approach for evaluating the city logistics dynamic in a vulnerability situation: An exploratory study in Belo Horizonte (Brazil). Transp. Res. Procedia 2017, 25, 1046–1060. [Google Scholar] [CrossRef]
- Liu, W.H.; Wei, W.Y.; Yan, X.Y.; Dong, D.; Chen, Z.X. Sustainability risk management in a smart logistics ecological chain: An evaluation framework based on social network analysis. J. Clean. Prod. 2020, 276, 124189. [Google Scholar] [CrossRef]
- Qazi, A. Adoption of a probabilistic network model investigating country risk drivers that influence logistics performance indicators. Environ. Impact Assess. Rev. 2022, 94, 106760. [Google Scholar] [CrossRef]
- Zheng, W.J.; Xu, X.H.; Wang, H.W. Regional logistics efficiency and performance in China along the Belt and Road Initiative: The analysis of integrated DEA and hierarchical regression with carbon constraint. J. Clean. Prod. 2020, 276, 123649. [Google Scholar] [CrossRef]
- Liang, Z.J.; Chiu, Y.-H.; Guo, Q.; Liang, Z. Low-carbon logistics efficiency: Analysis on the statistical data of the logistics industry of 13 cities in Jiangsu Province, China. Res. Transp. Bus. Manag. 2021, 43, 100740. [Google Scholar] [CrossRef]
- Zhuang, Z.L.; Fu, S.Q.; Lan, S.L.; Yu, H.; Yang, C.; Huang, G.Q. Research on economic benefits of multi-city logistics development based on data-driven analysis. Adv. Eng. Inform. 2021, 49, 101322. [Google Scholar] [CrossRef]
- Lan, X.J. On development strategy of rural logistics market. In Proceedings of the 2014 2nd International Conference on Education Technology and Information System (ICETIS 2014), Jinan, China, 26–27 April 2014; pp. 155–158. [Google Scholar]
- Yan, Z.; Zhang, Y.-Y.; Li, Y.F.; Liu, S.; Yang, J.A. A Study of Rural Logistics Center Location Based on Intuitionistic Fuzzy TOPSIS. Math. Probl. Eng. 2017, 2017, 2323057. [Google Scholar] [CrossRef] [Green Version]
- Jiang, X.H.; Wang, H.Y.; Guo, X.C.; Gong, X.L. Using the FAHP, ISM, and MICMAC approaches to study the sustainability influencing factors of the last-mile delivery of rural E-commerce logistics. Sustainability 2019, 11, 3937. [Google Scholar] [CrossRef] [Green Version]
- Yuan, S.; Chen, J. Research on Rural Logistics Development Strategy and Construction of Common Logistics System in Hunan. In Proceedings of the 5th International Conference on Economics, Management, Law and Education (EMLE 2019), Krasnodar, Russia, 11–12 October 2020; pp. 767–769. [Google Scholar]
- Jiang, H.; Yang, J. Applying rough set theory to analyzing the evaluation index system of modern rural logistics construction. In Proceedings of the 2009 International Conference on Test and Measurement, Hong Kong, China, 5–6 December 2009; Volume 2, pp. 418–421. [Google Scholar] [CrossRef]
- Luo, Y.; Bai, Y. Evaluation Model Construction and Empirical Analysis of Rural E-Commerce Logistics Service Quality. In Proceedings of the International Conference on Applications and Techniques in Cyber Security and Intelligence, Fuyang, China, 19–21 June 2020; pp. 28–34. [Google Scholar] [CrossRef]
- Lin, P.-C.; Cheng, T.C.E. The diffusion and the international context of logistics performance. Int. J. Logist. Res. Appl. 2019, 22, 188–203. [Google Scholar] [CrossRef]
- Lu, H.; Lu, X.; Jiao, L.D.; Zhang, Y. Evaluating urban agglomeration resilience to disaster in the Yangtze Delta city group in China. Sustain. Cities Soc. 2022, 76, 103464. [Google Scholar] [CrossRef]
- Yang, Z.D. Comparison and empirical analysis of the urban economic development level in the Yangtze River urban agglomeration based on an analogical ecosystem perspective. Ecol. Inform. 2021, 64, 101321. [Google Scholar] [CrossRef]
- Ekici, Ş.Ö.; Kabak, Ö.; Ülengin, F. Improving logistics performance by reforming the pillars of Global Competitiveness Index. Transp. Policy 2019, 81, 197–207. [Google Scholar] [CrossRef]
- Wang, X. The research on the evaluation index system of livable rural areas in China—by the case of rural areas in Henan Province. Agric. Agric. Sci. Procedia 2010, 1, 456–461. [Google Scholar] [CrossRef] [Green Version]
- Qi, G.Q.; Shi, W.M.; Lin, K.-C.; Yuen, K.F.; Xiao, Y. Spatial spillover effects of logistics infrastructure on regional development: Evidence from China. Transp. Res. Part A Policy Pract. 2020, 135, 96–114. [Google Scholar] [CrossRef]
- Zhang, W.D.; Lu, J.; Zhang, Y. Comprehensive evaluation index system of low carbon road transport based on fuzzy evaluation method. Procedia Eng. 2016, 137, 659–668. [Google Scholar] [CrossRef] [Green Version]
- Xie, B.K.; Zhang, X.X.; Lu, J.L.; Liu, F.; Fan, Y.Q. Research on ecological evaluation of Shanghai port logistics based on emergy ecological footprint models. Ecol. Indic. 2022, 139, 108916. [Google Scholar] [CrossRef]
- Wang, T.; Kang, J.W.; Valentine, V.F. A holistic analysis of national e-commerce and logistics development. Marit. Econ. Logist. 2020, 22, 500–513. [Google Scholar] [CrossRef]
- Helo, P.; Rouzafzoon, J. Logistics Informatization Systems. Int. Encycl. Transp. Sci. 2021, 105, 76–84. [Google Scholar] [CrossRef]
- Lu, H.; Shang, T.; Wei, Y.; Wu, P. Safety assessment of exit advance guide signs in mountainous highway tunnel based on eye-tracking Technology. IEEE Access 2021, 9, 111315. [Google Scholar] [CrossRef]
- Moayedi, H.; Aghel, B.; Vaferi, B.; Foong, L.K.; Bui, D.T. The feasibility of Levenberg–Marquardt algorithm combined with imperialist competitive computational method predicting drag reduction in crude oil pipelines. J. Pet. Sci. Eng. 2020, 185, 106634. [Google Scholar] [CrossRef]
- Nasr, A.K.; Tavana, M.; Alavi, B.; Mina, H. A novel fuzzy multi-objective circular supplier selection and order allocation model for sustainable closed-loop supply chains. J. Clean. Prod. 2021, 287, 124994. [Google Scholar] [CrossRef]
- Hosseini, S.; Vaferi, B. Determination of methanol loss due to vaporization in gas hydrate inhibition process using intelligent connectionist paradigms. Arab. J. Sci. Eng. 2022, 47, 5811–5819. [Google Scholar] [CrossRef]
- Mamdani, E.H.; and Assilian, S. An experiment in linguistic synthesis with a fuzzy logic controller. Int. J. Man-Mach. Stud. 1975, 7, 1–13. [Google Scholar] [CrossRef]
- Larsen, P.M. Industrial applications of fuzzy logic control. Int. J. Man-Mach. Stud. 1980, 12, 3–10. [Google Scholar] [CrossRef]
- Liu, Y.J.; Gao, Y.; Tong, S.; and Li, Y. Fuzzy approximation-based adaptive backstepping optimal control for a class of nonlinear discrete-time systems with dead-zone. IEEE Trans. Fuzzy Syst. 2015, 24, 16–28. [Google Scholar] [CrossRef]
- Chen, Y.C.; Teng, C.-C. A model reference control structure using a fuzzy neural network. Fuzzy Sets Syst. 1995, 73, 291–312. [Google Scholar] [CrossRef]
- Tang, J.-J.; Liu, F.; Zou, Y.J. An improved fuzzy neural network for traffic speed prediction considering periodic characteristic. IEEE Trans. Intell. Transp. Syst. 2017, 18, 2340–2350. [Google Scholar] [CrossRef]
- Barbounis, T.; Theocharis, J.B. A locally recurrent fuzzy neural network with application to the wind speed prediction using spatial correlation. Neurocomputing 2007, 70, 1525–1542. [Google Scholar] [CrossRef]
- Pan, Y.; Er, M.J.; Li, X.; Yu, H.; Gouriveau, R. Machine health condition prediction via online dynamic fuzzy neural networks. Eng. Appl. Artif. Intell. 2014, 35, 105–113. [Google Scholar] [CrossRef]
- Wang, X.C. Matlab Neural Network Analysis of 43 Cases; Beijing University of Aeronautics and Astronautics Press: Beijing, China, 2013; pp. 288–298. [Google Scholar]
- Tepanosyan, G.; Sahakyan, L.; Zhang, C.S.; Saghatelyan, A. The application of Local Moran’s I to identify spatial clusters and hot spots of Pb, Mo and Ti in urban soils of Yerevan. Appl. Geochem. 2019, 104, 116–123. [Google Scholar] [CrossRef]
- Huang, J.B.; Ding, A.; Li, Y.; Lu, D. Increasing the risk management effectiveness from higher accuracy: A novel non-parametric method. Pac. Basin Financ. J. 2020, 62, 101373. [Google Scholar] [CrossRef]
- Yang, T.; Jin, Y.; Yan, L.; Pei, P. Aspirations and realities of polycentric development: Insights from multi-source data into the emerging urban form of Shanghai. Environ. Plan. B Urban Anal. City Sci. 2019, 46, 1264–1280. [Google Scholar] [CrossRef]
- Sá Marques, T.; Saraiva, M.; Ribeiro, D.; Amante, A.; Silva, D.; Melo, P. Accessibility to services of general interest in polycentric urban system planning: The case of Portugal. Eur. Plan. Stud. 2020, 28, 1068–1094. [Google Scholar] [CrossRef]
- Heitz, A.; Dablanc, L.; Tavasszy, L.A. Logistics sprawl in monocentric and polycentric metropolitan areas: The cases of Paris, France, and the Randstad, the Netherlands, REGION. Eur. Reg. Sci. Assoc. (ERSA) Louvain-La-Neuve 2017, 4, 93–107. [Google Scholar] [CrossRef] [Green Version]
Perspective | Indicators | Methods | Source |
---|---|---|---|
Efficiency improvement | Cost; Implementation time; Specialized technical requirements; Social reaction, a requirement for synergy between all stakeholders; Exploitation of existing infrastructure, policies, and actions. | Multi-Criteria Analysis | Morfoulaki et al., 2016 [30] |
Frequency of merchandise delivery; Time of loading and unloading of merchandise; Typology/size of the establishments; Type of vehicle used; Location and offer of dedicated spaces for loading/unloading; Street design and geometric characteristics associated with the quality of urban logistics operations; Degree of saturation of the street (service levels) associated with the quality of urban logistics operations. | Multi-Criteria Evaaluation in GIS | Viana and Delgado, 2019 [31] | |
Vehicle capacity; Customer demand; Distance between two node points; Fixed cost of the vehicle; Subsidy on the purchasing price of the vehicle; Annual circulation tax of the vehicle; Entrance fee paid by the vehicle; Energy price for the vehicle; Energy tax for the vehicle; Energy efficiency for the vehicle; The amount of carbon emits; Carbon tax on the vehicle; A sufficiently large number; The speed limit of the vehicle. | Ant Colony Optimisation | Juvvala and Sarmah, 2021 [32] | |
Degree centrality; Betweenness centrality; Closeness centrality; Network efficiency centrality. | Entropy-weighted TOPSIS | Zhao et al., 2021 [33] | |
Risk management | Demand for daily delivery; Daily route; Opening hours of the establishment; NOx count; Complaint count; Occurrence and duration time; Travel time information; Speed information; Pollutant emission parameters. | Multi-Agent Modeling | Oliveira et al., 2017 [34] |
Internal stakeholder risks; External stakeholder risks; Indirect stakeholder risks; The risk of disruptive technology upgrades; The risk of technological route changes; Government management risks; Industry management risks; Intermediary service risks; Industrial policy risks; Legal and regulatory risks. | Social Network Analysis; TOPSIS | Liu et al., 2020 [35] | |
Business Environment Risk; Corruption Risk; Economic Risk; Environmental Risk; Financial Risk; Health and safety Risk; Political Risk; Customs; Infrastructure; International Shipments; Logistics Competence; Timeliness; Tracking and Tracing. | Probabilistic network model | Qazi, 2022 [36] | |
Urban agglomeration | Investment in fixed assets; Length of logistics network; Postal outlets; Terminal energy consumption; Cargo volume; Gross product of logistics industry; Carbon emissions. | Data envelopment analysis | Zheng et al., 2020 [37] |
Input and output indexes; Total mileage; Capital stock; Number of employees; Freight volume; Cargo turnover; GDP of the logistics industry; CO2 emission; Environmental variable index; Logistics industry density; Urbanization level; Logistics specialization level. | The three-stage Super-SBM model | Liang et al., 2021 [38] | |
Number of people employed in logistics; Ownership of civil trucks; Total logistics investment in fixed assets; Total of Post business; Road/highway freight turnover; Added-value of the tertiary industry; GDP per capita; Total national investment in fixed assets; Total retail sales of consumer goods; Resident population; Household consumption; Public income. | Data envelopment analysis | Zhuang et al., 2021 [39] | |
Rural logistics from the perspective of urban agglomeration | Per capita GDP; Per capita disposable income of urban residents; Total retail sales of social consumer goods; Total sown area of crops; GDP of agriculture, forestry, animal husbandry, and fishery; Total power of agricultural machinery; Total grain output; Highway mileage; Highway freight turnover; Highway density; Internet broadband access users. | Fuzzy neural network | This paper |
Level | Index | Variable | Attribute | Unit | Description | Source |
---|---|---|---|---|---|---|
Per capita GDP | + | CNY | Regional economic scale | Yang, 2021 [48] | ||
Rural logistics demand | Per capita disposable income of urban residents | + | CNY | Consumption demand of regional urban residents | Yang, 2021 [48] | |
Total retail sales of social consumer goods | + | 100 mn CNY | Realization degree of social commodity purchasing power | Lu et al., 2022 [47] | ||
Total sown area of crops | + | Hectares | Regional agriculture production scale | Wang, 2010 [50] | ||
Agricultural production capacity | GDP of agriculture, forestry, animal husbandry, and fishery | + | 100 mn CNY | Regional comprehensive agricultural production capacity | Wang, 2010 [50] | |
Total power of agricultural machinery | + | 10,000 kw | Regional agriculture development scale | Wang, 2010 [50] | ||
Total grain output | + | 10,000 tons | Regional agricultural production capacity | Wang, 2010 [50] | ||
Highway mileage | + | Kilometer | Regional transportation infrastructure | Xie et al., 2022 [53] | ||
Rural infrastructure | Highway freight turnover | + | 10,000-ton kilometers | Freight transport efficiency | Xie et al., 2022 [53] | |
Highway density | + | Km/100 square kilometers | Regional logistics circulation capacity | Zhang et al., 2016 [52] | ||
Rural informatization | Internet broadband access users | + | Thousand households | Development degree of regional Informatization | Wang and Kang, 2020 [54] |
Evaluation index | I | II | III | IV | V |
---|---|---|---|---|---|
Per capita GDP | 12,638~23,769 | 23,768~30,806 | 30,905~38,712 | 38,846~49,376 | 49,830~85,679 |
Per capita disposable income of urban residents | 12,708~19,386 | 19,451~24,565 | 24,619~28,851 | 28,920~34,371 | 34,549~48,592 |
Total retail sales of social consumer goods | 98~283 | 285~420 | 422~580 | 586~1013 | 2473~117,872 |
Total sown area of crops | 118~340 | 341~448 | 449~525 | 528~778 | 810~3372 |
GDP of agriculture, forestry, animal husbandry, and fishery | 93~224 | 225~286 | 287~354 | 356~500 | 528~2749 |
Total power of agricultural machinery | 84~152 | 153~193 | 194~230 | 231~281 | 10,591~825,062 |
Total grain output | 35~127 | 127~165 | 166~216 | 217~258 | 268~1081 |
Highway mileage | 5656~8160 | 8201~11,482 | 11,524~14,846 | 14,914~20,292 | 22,298~180,796 |
Highway freight turnover | 240,303~423,410 | 424,931~545,112 | 546,691~671,405 | 677,515~1,160,247 | 1,597,072~36,106,277 |
Highway density | 38~99 | 100~136 | 136~156 | 157~184 | 185~258 |
Internet broadband access users | 88~268 | 270~410 | 413~646 | 653~1266 | 1421~13,724 |
City | 2010 | 2011 | 2012 | 2013 | 2014 | 2015 | 2016 | 2017 | 2018 | 2019 | 2020 |
---|---|---|---|---|---|---|---|---|---|---|---|
Chongqing | 2.77 | 3.01 | 3.39 | 2.63 | 3.85 | 4.13 | 3.56 | 5.26 | 4.59 | 5.38 | 5.74 |
Chengdu | 1.95 | 2.63 | 3.23 | 3.84 | 4.32 | 4.35 | 4.47 | 5.42 | 6.04 | 5.15 | 6.50 |
Zigong | 0.51 | 0.95 | 1.38 | 1.82 | 2.09 | 2.54 | 2.80 | 3.31 | 3.83 | 4.25 | 4.70 |
Luzhou | 0.56 | 0.96 | 1.40 | 1.74 | 2.13 | 2.20 | 2.75 | 3.22 | 3.67 | 4.16 | 4.52 |
Deyang | 0.71 | 1.46 | 1.95 | 2.45 | 2.79 | 3.33 | 3.29 | 3.84 | 4.35 | 4.47 | 5.24 |
Mianyang | 0.71 | 1.18 | 1.61 | 1.98 | 2.37 | 2.97 | 3.06 | 3.64 | 4.12 | 4.44 | 5.00 |
Suining | 0.39 | 0.58 | 0.98 | 1.26 | 1.62 | 1.51 | 2.29 | 2.73 | 3.29 | 3.88 | 4.18 |
Neijiang | 0.48 | 0.77 | 1.19 | 1.45 | 1.76 | 1.80 | 2.48 | 2.88 | 3.34 | 3.92 | 4.20 |
Leshan | 0.57 | 1.05 | 1.50 | 1.95 | 2.31 | 2.98 | 2.92 | 3.39 | 3.95 | 4.35 | 4.93 |
Nanchong | 0.62 | 0.63 | 0.98 | 1.17 | 1.52 | 1.86 | 2.16 | 2.60 | 3.04 | 3.71 | 3.87 |
Meishan | 0.47 | 0.79 | 1.22 | 1.59 | 1.93 | 1.97 | 2.61 | 3.06 | 3.51 | 4.04 | 4.31 |
Yibin | 0.60 | 0.98 | 1.43 | 1.80 | 2.18 | 2.46 | 2.72 | 3.22 | 3.90 | 4.34 | 4.91 |
Guang’an | 0.46 | 0.78 | 1.21 | 1.53 | 1.83 | 1.47 | 2.37 | 2.77 | 3.17 | 3.69 | 3.91 |
Dazhou | 0.54 | 0.61 | 0.95 | 1.12 | 1.46 | 1.74 | 2.16 | 2.57 | 2.94 | 3.54 | 3.73 |
Ya’an | 0.38 | 0.75 | 1.18 | 1.60 | 1.94 | 1.88 | 2.57 | 2.95 | 3.38 | 3.94 | 4.25 |
Ziyang | 0.45 | 0.78 | 1.20 | 1.43 | 1.73 | 1.05 | 2.32 | 2.65 | 3.00 | 3.38 | 3.65 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Lu, H.; Bao, J. Spatial Differentiation Effect of Rural Logistics in Urban Agglomerations in China Based on the Fuzzy Neural Network. Sustainability 2022, 14, 9268. https://doi.org/10.3390/su14159268
Lu H, Bao J. Spatial Differentiation Effect of Rural Logistics in Urban Agglomerations in China Based on the Fuzzy Neural Network. Sustainability. 2022; 14(15):9268. https://doi.org/10.3390/su14159268
Chicago/Turabian StyleLu, Hao, and Jie Bao. 2022. "Spatial Differentiation Effect of Rural Logistics in Urban Agglomerations in China Based on the Fuzzy Neural Network" Sustainability 14, no. 15: 9268. https://doi.org/10.3390/su14159268
APA StyleLu, H., & Bao, J. (2022). Spatial Differentiation Effect of Rural Logistics in Urban Agglomerations in China Based on the Fuzzy Neural Network. Sustainability, 14(15), 9268. https://doi.org/10.3390/su14159268