基于数据挖掘的各城市综合竞争力等级分类的研究

Research on Classification of Urban Comprehensive Competitiveness Based on Data Mining

近几年随着中国经济实力的不断发展,各个城市越来越注重自身的综合城市竞争力。城市竞争力评价体系、等级分类体系的建立能使各个城市有针对性的把握自身未来发展方向。为了建立评价体系与进行等级分类,本文通过因子分析提取出三个主因子(Fa1:综合经济和信息化程度、Fa2:城市环境与医疗服务水平因子、Fa3:经济增长效益),并建立了城市指标体系。在此基础上,学习并采用K-中心聚类、决策树、神经网络、KNN与加权KNN等方法,从三个主因子得分入手,对各城市进行等级分类,与2016年官方城市综合竞争力排名进行比对,判断各方法的分类准确率,比较得出城市等级分类的最优方法以及影响城市综合竞争力的主要因素。基于R语言软件分析,我们得到以下研究结论:在对城市等级进行分类的研究中,发现决策树、神经网络算法分类准确率最优,其次分别是加权KNN、KNN算法和K-中心聚类。并且得到影响城市综合竞争力的主要因素分别是财政预算内收入、社会消费品零售总额、电话普及率、互联网用户数、金融机构年末存款余额与人均公园绿地面积指标。

In recent years, with the continuous development of China’s economic strength, cities pay more and more attention to their comprehensive urban competitiveness. The establishment of the evaluation system and classification system of urban competitiveness can enable each city to grasp its own fu-ture development direction. In order to establish an evaluation system and grade classification, this paper extracts three main factors (Fa1: comprehensive economic and informatization degree, Fa2: urban environment and medical service level factor, and Fa3: economic growth benefit) through factor analysis, and set up city index system. On this basis, we learn and adopt K-center clustering, decision tree, neural network, KNN and the weighted KNN, and start from the three main factor scores to classify each city, and the comprehensive competitiveness of the official city in 2016. The rankings are compared, the classification accuracy of each method is judged, and the optimal me-thod of city classification and the main factors affecting the comprehensive competitiveness of the city are compared. Based on the analysis of R language software, we get the following research con-clusions: in the research of city classification, it is found that the classification accuracy of decision tree and neural network algorithm is the best, followed by weighted KNN, KNN algorithm and k-center clustering. Besides, the main factors affecting the comprehensive competitiveness of the city are the revenue within the budget, total retail sales of consumer goods, telephone penetration rate, Internet users, year-end deposit balance of financial institutions and per capita green space index.