Landslide susceptibility mapping and debris flow modelling are critical components of disaster risk management, particularly in regions such as the Western Ghats in Kerala, which are susceptible to rainfall-induced shallow landslides. Traditional physics-based models for landslide susceptibility and debris flow often encounter challenges in acquiring geotechnical and hydrological data, making large-scale implementation laborious. Similarly, statistical models face difficulties due to the complexity and variability of environmental factors influencing landslide occurrences. The study seeks to enhance the accuracy and predictive capabilities of landslide runout area models by integrating advanced machine learning (ML) techniques with physics-based landslide susceptibility models. The study explores four classification models, including Random Forest, Support Vector Machine, Gradient Boosting Machine, and Sequential Neural Networks, to predict debris flow paths. These models incorporate a range of input features from digital elevation models, such as slope, flow direction, and various curvatures, in addition to hydrological data, soil depth, as well as slope failure indicators such as factor of safety and probability of slope failure derived from grid-based infinite slope stability models. The study employed three analytical methods, namely, correlation tests, Pearson tests, and Fisher scores, to assess and quantify the influence of 15 features on the likelihood of landslide runout areas. The effectiveness of these models is cross-validated using data from the 2019 Kavalappara landslide event, demonstrating that the Sequential Neural Network model outperforms other classifications. The paper not only presents a novel integration of ML with traditional landslide analysis techniques but also lays the groundwork for developing more sophisticated models for the Western Ghats. By refining these models and expanding the datasets, the approach could significantly improve landslide predictions and aid in devising more effective disaster management strategies. The findings offer a fresh perspective on the dynamics of land movement and the development of early warning systems.