The purpose of the present invention is to provide a method for predicting the amount of electricity lost by electricity customers so as to solve the problem of predicting the amount of electricity lost through manual experience in the prior art. The method includes the following steps: step 1: classifying customers who use electricity by cluster analysis; step 2: calculating predicted loss of customers for each type of electricity customers during the forecast month; step 3: obtaining historical electricity consumption of each electricity customer, calculating the average monthly electricity consumption of each type of electricity customers based on the historical electricity consumption in the forecast month; and step 4: according to the average monthly electricity consumption of each type of electricity customers in the forecast month and the predicted loss of customers for each type of electricity customers during the forecast month, calculating the total lost electricity consumption of all electricity customers in the forecast month. The beneficial effect is that the total lost electricity consumption can be calculated, and the defects the predicted lost electricity consumption can only be obtained empirically in the prior art are overcome.