Nowadays, Road traffic crashes (RTC) are a worldwide fifth cause of deceased, especially in the transport and logistics sector. Consequently, the fatalities in accidents are fast increasing every single day. Thus, it is important to have some early prediction methods which may be help drivers, and riders to know the statistics of accidents in the region itself, knowing that the information about the speed limit, respect for traffic signs, traffic lights, pedestrian aisles, right of precedence, weather conditions, omissions, and sleep, as well as the excessive speed that causes RTC. As a result, they can be careful at all times to curb these traffic accidents. In the context of artificial intelligence, some ML prediction approaches have been investigated for decision support to avoid this dilemma of RTC. These methods investigate to claim good reports on these issues. However, this sticking method has required good results for 99% of accuracy vis a vis the recent related work such as Logistic regression for 81%, K-Nearest Neighbors for 82%, Random Forest for 87%, SVM for 77% respectively of accuracy, also for mean square error , we obtained 0.26% for sticking method, K-Nearest Neighbors for 0.23%, Random Forest for 13%, SVM for 29% respectively. In this proposition sticking method algorithm has been giving the good performance cost of various learning models vis a vis the classification approach over TRAFFIC ACCIDENTS_2019_LEEDS (TAL19) Datasets. Finally, the good causes for the advantage of the predictions have been developed.