The invention belongs to the field of power marketing services, and discloses a fee-paying channel preference analysis and prediction method for an electricity customer based on an RFM model. The method comprises the steps: extracting the fee-paying record of a customer from a marketing business system; setting an analysis date, extracting the time interval from the latest fee-paying date to the analysis date in the fee-paying record, the fee-paying frequency F of an analysis time period and the mean fee-paying amount M of the analysis time period of each fee-paying channel of each user, and converting the R, F and M values of each fee-paying channel of each user into Rscore, Rscore and Mscore; calculating the weight coefficients of Rscore, Rscore and Mscore of each fee-paying channel of each user through an entropy method, and determining an RFM model calculation model of each fee-paying channel; finally calculating the preference score of each channel of each user according to the RFM model, wherein the channel with the highest score is the fee-paying channel that is employed by the user next time at the largest probability. The method can achieve the accurate recognition of the channel preference of each customer, and also can directly reflect the actual demands of one fee-paying channel in one region.