Oil well pump change is an important means to adjust formation pressure and improve plane contradiction. Aiming at the problem that it is difficult to accurately determine the displacement of changing pump in oil well, the method of big data analysis is used to optimize the displacement of changing pump quantitatively. The sample database of oil well history pump change was established, and the main control factors of pump change effect were determined by using big data correlation analysis and expert experience. Machine learning algorithm is optimized to establish the prediction model of daily oil gain in the early stage of pump change. In combination with exhaustive search, the optimal displacement is determined with the aim that the water content does not rise and the daily oil increase is the maximum. This method is applied in M oilfield, and 7 main controlling factors of pump change effect, such as daily oil production, submergence, production thickness and pump efficiency, are determined. The average mean square error of the daily oil increase prediction model in the training set and the test set is 0.73, and the oil increase effect is good in the typical well implementation, and the water cut is basically not rising. This method provides a new way to optimize the displacement of pump replacement well, and greatly improves the efficiency and accuracy of displacement determination.