Based on the adjustment of marketing and distribution massive data, this paper proposes a system for power grid line loss and stealing electricity prediction with the key technologies of electric power big data platform. The system provides the integration of calculation, analysis and visualization for line loss. According to the result of line loss calculation and the existing massive data of power grid enterprise, the system adopts the big data technology of Hadoop offline distributed computing and Spark memory computing algorithm to identify the potential acts of stealing electricity by big data mining for the abnormal line loss rate of line and supplying district. Therefore, it provides the stealing electricity prediction and data support services for power supply enterprise with anti-stealing inspection and improves the economic efficiency of enterprise. Through the construction of the system, the application of big data technologies in the electric power industry is verified and implemented. This work is supported by National Natural Science Foundation of China (No.61672276).