Abstract: Aiming at the problem of low accuracy of crime distribution prediction and serious lack of historical crime data, a crime distribution prediction algorithm, TWcS, was proposed based on a transition probability matrix model, the historical crime data and integrating social environmental factors in the studied area. In this paper, the social environment factors including distance information, area information and population information were introduced as weights into the gradient descent strategy, and the transition probability matrix self-learning of TWcS algorithm was realized by gradient descent. The experimental results show that the performance of TWcS algorithm is superior to other prediction algorithms including TPML-WMA, LR, AR, Lasso regression algorithm, Bayesian algorithm, decision tree algorithm, etc.The MAE value of TWcS algorithm is only 33% of the average MAE value of the other algorithms.