Reasonable use of household electricity consumption data obtained from smart meter can infer household income， which is conducive to the control of consumer groups，so that businesses can provide more targeted services and products for users． In order to improve the accuracy of inference，a data fusion method is proposed to estimate household income based on the information of total household electricity consumption and housing area． Several different machine learning classification algorithms are used to train and analyze the data． Finally，the classification accuracy of annual household income can reach 81% ． Compared to using the information of total household electricity consumption only，the classification accuracy is improved by 15% ． It can be seen that the method of increasing housing area information can achieve certain evaluation purposes， provide help for businesses and users，and enable people to enjoy a more intelligent and quality life．