基于MIC-XGBoost算法的居民用水量数据预测

PREDICTION OF RESIDENTS WATER CONSUMPTION DATA BASED ON MIC-XGBOOST ALGORITHM

为了进一步提升水务企业工作效率,解决抄表数据审核工作量大、保证数据准确性等问题,在传统的XGBoost预测算法基础上提出一种基于MIC-XGBoost的混合预测模型。以某水务企业近两年的用户历史用水数据为基础,利用最大信息系数(MIC)得出不同影响因素与用户用水量之间变量的关联程度,构建不同单一影响因素下的用水量数据预测模型,采用实际用水数据调整算法参数,得到最终预测模型。实验结果表明,在预测精度上该模型比单一的XGBoost模型提高了约21%,能有效提升数据审核效率。

In order to further improve the working efficiency of water utilities and solve the problems of heavy workload of meter reading data review and to ensure data accuracy,a hybrid forecasting model based on MIC-XGBoost is proposed based on the traditional XGBoost(eXtreme Gradient Boosting)forecasting algorithm.Based on a user s historical water consumption data of a water company in the past two years,the maximum information coefficient(MIC)was used to obtain the degree of correlation between the different influencing factors and the user s water consumption,and the different single influencing factors were constructed.The actual water consumption data was used to adjust the algorithm parameters to obtain the final prediction model.The experimental results show that the MIC(temperature)-XGBoost model is about 21%higher than the single XGBoost model in prediction accuracy,which can effectively improve the efficiency of data review.