Data Validity Analysis and Hybrid Soft Computing Prediction Method of Water Resources in Mining Area

Abstract

In order to solve the problem of detecting and correcting the outliers of mining area water resources monitoring data, a hybrid soft computing method based on pSO and Bp network was proposed to detect the outliers of time series data. First, through the classification and screening of the initial outlier data, the mining area water resource big data “full data” associated model was constructed to fit the regression curve between each of the two different data parameters, then use the “full data” associated model of mining area water resources big data to define another parameter set whose linear correlation coefficient is the largest, which is the correlation monitoring data parameter set of the monitoring data parameter. Through the correlation analysis between data, the overall change trend of most data can be fitted, without the influence of abnormal values, and the abnormal values can be detected effectively according to the deviation ratio. Finally, taking an area's annual mining area water consumption data as an example, we make an empirical analysis. The results show that the hybrid soft calculation method proposed in this paper can effectively detect abnormal values and the corrected data can truly reflect the situation of mining area water consumption in the area, and can provide more real and reliable data for subsequent analysis.