A method was proposed to use big data to mine charging demand information of electric vehicles (EVs) and select location and construction scale of charging point. Firstly, the travel data of EV users were cleaned and analysed to obtain the available rules and corresponding digital characteristics. Secondly, according to Markov chain principle, the roulette wheel method was used to predict the spatial and temporal distribution of charging demand. Then particle swarm algorithm was used to optimize the selection of candidate sites. Finally, a typical region was taken as an example to prove the feasibility of this scheme.