基于RBF网络的城市垃圾产量预测及可视化

为了预测并控制未来几年城市垃圾产量，以我国城市为例，利用K-近邻互信息的多变量特征从18个拟影响因素中确定了8个影响垃圾排放量因子，分别为常住人口、地区生产总值、社会消费品零售值、金融业增加值、工业增加值、批发和零售业增加值、住宿和餐饮业增加值和第三产业增加值.以2006~2013年数据为训练样本，2014~2015年数据为检验样本，根据影响因素建立径向基函数（RBF）神经网络预测模型，并基于平均相对误差对模型反向修正.然后结合两段式径向基预测模型，对全国各省市2017~2018年的垃圾总产量预测并可视化.结果表明，本文建立的两段式径向基预测模型平均相对误差是6.43%，预测精度为93.57%.可见，该模型的预测精度较高，能较好的在现实生活中对城市垃圾的产生量进行预测.

Prediction and visualization of municipal solid waste production based on RBF network

In order to predict and control the solid waste production in city for the following few years, Chinese cities, for example, the paper was determined 8factors from 18factors that have influence in solid waste production by using multivariate feature of K-neighbor mutual information. The 8factors that have influence in solid waste production was permanent resident population, regional GDP, total retail sales of consumer goods, added value of the financial industry, added value of the industry, added value of wholesale and retail industry, added value of accommodation and catering industry, and added value of the tertiary industry. The data from 2006 to 2013 was used as the training sample. And the data from 2014 to 2015 was used as the test sample. To predict and visualize the total solid waste production in all provinces and cities from 2017 to 2018, firstly, paper was established a prediction model of radial basis function (RBF) neural network based on above-mentioned 8influence factors. Second, paper was corrected the prediction model based on the mean relative error (MRE). Third, this paper was proposed two-stage radial basis prediction model. The results showed that the MRE of the two-stage radial basis prediction model proposed in this paper was 6.43%, which was equivalent to 93.57% prediction accuracy. It was therefore obvious that the prediction accuracy of this model was high, and it was capable to predict the municipal solid waste production in real life.