In response to the poor effectiveness of water plugging and profile control measures in production and injection wells in a high water cut oil reservoir block of a certain oilfield in the east, how to efficiently select wells and adopt appropriate profile control or water blocking measures has become an urgent problem that needs to be solved. For the conventional work-over decision method, the weight values of indexes was defined manually and relied on experience, and not considering the production performance and laws of overall block. Machine learning random forest algorithm is used to establish a new decision model for work-over well selection of overall block. The algorithm model is established by normalization and discretization of production data. The weight values of four decision index and comprehensive evaluation decision index is calculated based on the model. The work-over well selection is determined by comprehensive evaluation decision index eventually. The difference between conventional model and new algorithm model is compared. The result shows the algorithm decision effect is better than the conventional decision system, which improving the accuracy of well selection decisions in the overall block water plugging and profile control measures.