Given the seriousness of road traffic accidents as a public health concern, it is critical to comprehend the variables linked to an increase in the severity of injuries in accidents. To improve road safety, make better decisions about road safety, and lessen the severity of crashes in the future, it is crucial to identify these elements. The study aimed to collect traffic data, analyse it to identify suitable variables for accident prediction, and to develop an accident predictive model suitable for regional conditions. Random Forest, Support Vector Regression, and Multi Linear Regression models were developed to predict traffic accidents. The dataset comprised of 67 blackspots, each containing 17 variables collected using different survey from the identified blackspot of Karnataka state. Mean Absolute Error, Root Mean Squared Error, and Coefficient of Determination metrics were used to assess the models’ performance after the data was divided into training and validation sets, backward stepwise multilinear regression fared best with R2 of 0. 87 for validation. The study emphasises the potential of machine learning to improve traffic safety.