Prediction of user attributes based on big data of smart grid is of great significance for constructing smart grid analysis system and intelligent building.Traditional machine learning methods based on single user attribute analysis can not improve the accuracy by using the relationship between the various attributes, Digging missing data, these two issues have constrained the design of smart grid systems and the sophistication of intelligent building systems.
Cong Yang, a researcher at Shenyang Institute of Automation, Chinese Academy of Sciences, based on the research of many years of machine learning algorithms, proposed a multi-task learning supervised / semi-supervised user attribute prediction model by using each attribute prediction as a single task, Multiple tasks at the same time learning and decision-making.At the same time, the relationship between multiple user attributes was excavated to improve the accuracy of multiple attribute predictions, and the full use of missing data sample information to further improve the model generalization ability.
Relevant research results are published in IEEE Transactions on Smart Grid and Pattern Recogniton, respectively, by the Joint Household Characteristic Prediction via Smart Meter Data and User attribute discovery with missing labels. The research was supported by the State Key Laboratory of Robotics, the National Natural Science Foundation of China support.
