Availability of real-world road traffic flow data can enable researchers to find traffic-related patterns, accident prevention, traffic optimization, and much more. Researchers can use machine learning models with data to predict traffic-related outcomes. Lack of real-world traffic flow data for complex and high-traffic countries led us to the use of simulation to replicate real-world traffic to develop solutions and improve existing methods. The current research is about simulating rear-end collisions, one of the most frequent road accidents caused by tailgating, distracted driving, or sudden braking, leading to injuries and property damage. It utilizes traffic flow data from these accidents and applies well-known machine learning algorithms like logistic regression, random forest, support vector machine, Naive Bayes, K-nearest neighbor, and long short-term memory to determine whether the data is from an accident or not. The proposed work is simulating traffic using a simulator SUMO used to simulate necessary traffic conditions and extract traffic data. The results show how the LSTM algorithm, which has an F1 score of 0.91 for predicting accidents for time shift 3, works with traffic data over time and how different time dependencies change the prediction.