In order to solve the described insufficient problem of load feature selection and weight calculation in the past clustering analysis of residential electricity behavior, enhance the accuracy of clustering analysis in residential electricity behavior and reduce the time of clustering analysis operation, a data model based on ReliefF algorithm is proposed. The data model is characterized by electricity consumption rate during peak hour, the peak load time, the valley of the power, daily load cycles, the minimum load rate feature, and so on. The massive data of residential electricity behavior can be processed by the model, and clustering analysis of the model is made through k-means algorithm. Experimental data is obtained from a built-up smart community, and the result accuracy reaches to 94.61%, showing the proposed model based on ReliefF algorithm in clustering analysis of residential electricity behavior is effective.