基于粗糙集-模糊C均值聚类的Elman神经网络农村需水量预测

农村需水量影响因素作用机理复杂导致农村需水量预测值与实际值差别较大,采用模糊C聚类分析与Elman神经网络模型结合的方法建立农村需水量预测模型。首先,将用水方差和年用水均量等用水数据作为特征向量对2010—2017年海南省16个村落进行模糊C聚类,将村落分为三类;其次,以数据分析为基础,结合文献分析和官方数据分析提取关键因素,借助SPSS软件对关键因素进行降维处理,得到三类村落的关键影响因素;最后,将所得关键因素和2010—2016年用水数据作为Elman神经网络算法的输入对模型进行校核并运用粗糙集理论对模型进行修正,经误差分析,平均绝对百分比误差(MAPE)从0. 27下降到0. 127,对称平均绝对百分误差(SMAP)从0. 082下降到0. 041,平均绝对偏差(MAE)从3 832. 32减少到1 325. 53,表明模型可以相对全面的模拟农村需水量变化规律,可以用于农村水资源精准预测,为城乡供水一体化提供理论依据。

Elman Neural Network Forecast of Rural Water Demand Based on Rough Set-Fuzzy C-Means Clustering

The complicated action mechanism of influencing factors of rural water demand leads to a large difference between the predicted value and the actual value of rural water demand. This paper uses the method of combining fuzzy C-cluster analysis and Elman neural network model to establish the prediction model of rural water demand. Firstly, 16 villages in Baoting County of Hainan Province from 2010 to 2017 are classified into three categories by fuzzy C clustering using water consumption data such as variance of water consumption and average annual water consumption as feature vectors. Secondly, on the basis of data analysis, combined with literature analysis and official data analysis, the key factors are extracted and dimension reduction is carried out on the key factors by SPSS software to obtain the key influencing factors of the three types of villages. Finally, the obtained key factors and water consumption data from 2010 to 2016 are used as inputs of Elman neural network algorithm to check the model and rough set theory is used to revise the model. After error analysis, MAPE decreases from 0.27 to 0.127, SMAP decreases from 0.082 to 0.041, MAE decreases from 3832.32 to 1325.53, which indicates that the model can relatively comprehensively simulate the change rule of rural water demand, can be used for accurate prediction of rural water resources, and provides theoretical basis for integration of urban and rural water supply.