基于多因素权重分析的农村居民清洁取暖接受度分类预测方法研究

RESEARCH ON CLASSIFICATION AND PREDICTION METHOD OF RURAL RESIDENTS’ACCEPTANCE FOR CLEAN HEATING BASED ON MULTI-FACTOR WEIGHT ANALYSIS

提出一种基于多因素权重分析的分类模型(K-means-EWM-BP)来预测农村居民的清洁取暖接受度。首先,基于实地调研数据,选取农村居民家庭年总收入、性别、年龄、受教育程度作为聚类特征对农村居民分类;其次,在分类的基础上,对各类别农村居民的清洁取暖接受度影响因素进行多因素权重分析;最后,构建K-means-EWM-BP模型,实现对农村居民清洁取暖接受度的预测及验证。结果表明:1)受访农村居民可分为3类,其中清洁取暖接受度主要受教育程度影响的农村居民(类别1)占比31%,清洁取暖接受度主要受家庭年总收入影响的农村居民(类别2)占比43%,清洁取暖接受度主要受性别影响的农村居民(类别3)占比26%。2)类别1农村居民清洁取暖接受率预测值为95%,类别2农村居民清洁取暖接受率预测值为100%,类别3农村居民清洁取暖接受率预测值为72%。3)与EWM-BP模型和BP模型相比,K-means-EWM-BP模型预测准确性达到91.43%,高于准确性为87.14%的EWM-BP模型和准确性为80%的BP模型,同时标准误差(RMSE)与EWM-BP模型和BP模型相比分别降低0.01和0.06。

In this study,a classification model based on multi-factor weight analysis(K-means-EWM-BP)is proposed to forecast the acceptability of clean heating for rural residents.Firstly,rural residents are classified based on data from a field survey by taking gender,age,education level,and total annual household income as clustering characteristics.Secondly,on the basis of classification,multi-factor weight analysis is carried out on the influence factors of the acceptability of clean heating of various rural residents.Finally,the K-mean-EWM-BP model is constructed to forecast and verify the acceptability of clean heating for rural residents.The results show that:1)the rural residents can be divided into three categories,with 31%influenced by education level(category 1),43%by annual household income(category 2),and 26%by gender(category 3).2)The forecasted acceptance rate of clean heating for rural residents in category 1 is 95%,the forecasted acceptance rate of clean heating for rural residents in category 2 is 100%,and the forecasted acceptance rate of clean heating for rural residents in category 3 is 72%.3)The K-means-EWM-BP model achieves an accuracy of 91.43%in forecasting the acceptance rate of clean heating by farmers,surpassing both the EWM-BP model(with an accuracy of 87.14%)and the BP model(with an accuracy of 80%).Meanwhile,the root mean square error of the K-means-EWM-BP model declines by 0.01 and 0.06 relative to the EWM-BP model and the BP model,respectively.