Nowadays, a major cause of death is traffic accidents because of the increased traffic congestion in today's world. The present study attempts to identify high accident rate locations using ArcGIS and predict accident severity from accident data using machine learning models. Data collected from various government agencies on accidents that occurred in a district of India is used for this paper. Accident hotspot analysis is based on the number of grievous and fatal injuries per unit area and accident severity prediction using machine learning models like logistic regression, random forest, decision tree, and k-Nearest Neighbor (K-NN) models, etc. A comparative study of these machine learning models was conducted to examine which ones are more effective in accident prediction. Severity prediction concerning location is helpful in accident management on roads, thereby reducing the number of road accidents. Random Forest and K-NN gave the highest accuracy when the accuracies of the models were compared.