Hydraulic fracturing is of great significance to increase the fluid production capacity of oil wells and improve the degree of oil formation utilization in the middle and late stages of oilfield development, and accurate prediction of fracturing results of production wells is a prerequisite for determining the wells with potential for the measures when implementing the decision-making of fracturing measures for production wells. The traditional mathematical regression prediction model has low prediction accuracy, while the reservoir numerical simulation prediction model has the problems of difficult parameter acquisition, cumbersome calculation process, and strong professional barriers. Therefore, the analysis of influencing factors and data-driven prediction method of fracturing results of production wells are put forward. Feature correlation analysis and feature importance evaluation are used to optimize features.Three machine learning algorithms, K-nearest Neighbor (KNN), Support Vector Regression (SVR) and Random Forest (RF), are used to establish the prediction model of fracturing results of production wells, and Particle Swarm Optimization (PSO) is combined to optimize the hyperparameters of the three models. Finally, taking the actual block of an oilfield as an example, the model is applied, the prediction results of the three models are compared and analyzed, and the SVR model is selected for the prediction of the daily oil gain after fracturing of production wells. The application results show that the SVR model has a coefficient of determination of 0.72, a root-mean-square error of 0.099, and an average absolute error of 0.081 on the training set, and a coefficient of determination of 0.68, a root-mean-square error of 0.115, and an average absolute error of 0.099 on the test set. The SVR model shows good generalization ability, and has the value of popularization and application in the same type of block. It can provide technical support for the decision of fracturing measures for production wells.