A modified logisticMF Method based on experts and data characteristics for electric power applications

Based on the electric power data of Pudong Electric Power station in Shanghai, the model training and experiment are carried out with one million magnitude negative control data. Based on the logisticMF algorithm as the theoretical support, the abnormal data recommendation model is established on the basis of experts and data characteristics. The existing traditional recommendation method is automatic distribution methods, which has problems of chaotic distribution and low recommendation accuracy. Based on the recommendation model of logisticMF, this paper integrates the characteristics of expert attributes, user preferences and data labels, and puts forward the ABDR-LogisticMF model (Abnormal Data Recommend logisticMF) suitable for power data recommendation service. Experimental results show that the accuracy of the proposed ABDR-LogisticMF model is higher than that of automatic distribution, which can fully meet the recommended business requirements of the current number of abnormal data, and also provides support for subsequent distribution mechanisms of other abnormal energy data.