The invention relates to an intelligent traffic signal light transformation method, which is applied to the field of traffic signal light control at intersections and solves a problem that the green light is still on when the front phase is congested and a problem of how to control a signal light algorithm according to video vehicle data. The intelligent traffic signal light transformation methodcomprises a congestion prediction algorithm, a signal light combination selection algorithm, a preferential green light algorithm and a delayed green light algorithm. The congestion prediction algorithm is adopted, the naive Bayesian, decision-making tree, logistic regression, K-nearest neighbor, random forest, AdaBoost and gradient lifting methods in a statistical machine learning method are applied, historical data collected from video serves as a training data set, four characteristic attributes such as the intersection leaving distance, the average speed, the speed variation and the average density and the congestion condition are enabled to serve as training data, the average speed, the speed variation and the average vehicle density corresponding to the intersection leaving distancerange in the current real-time video of the road are enabled to serve as characteristic attributes, the congestion probability at a certain position after a certain time is predicted by applying the statistical machine learning method, the transformation of the signal lights is controlled according to the congestion probability, and judging the best phase combination, the preferential green lightphase and the green light delay time according to the number of vehicles inspected and calculated by the real-time video of the road.