Horizontal curves provide a smooth transition between two tangent sections of roadways and these curves account for more than 25% of fatal crashes. There have been numerous studies to model crash frequency in terms of various influencing roadway and traffic factors. Statistical methods met with limited success due to complex interactions and effects of influencing variables. In the past few years, there has been a growing interest in the use of Machine Learning (ML) techniques in road safety. This study attempts to model the frequency of fatal crashes using three ML techniques; Artificial Neural Networks (ANN), Classification and Regression Tree (CART), and Light Gradient Boosting Machine (Light GBM). This study is limited to the modelling of fatal crashes at simple horizontal curves on two-lane two-way non-urban roads in Kerala.