The flood depth-damage function is commonly used to estimate the damage in structures caused by floods. Although depth-damage functions are convenient to use for flood damage assessment due to its simplicity, they ignore the other flood damage influencing parameters such as the duration of the flood, age of the structure, floor area, type of wall and roof materials, etc. A machine learning-based approach is proposed in this study for the flood damage modelling incorporating all the relevant flood damage parameters. Data from the devastating 2018 flood in Kerala, India’s southernmost state, is gathered, and the collected data is utilised to investigate the applicability of the recently developed Machine learning (ML) classification algorithms, namely Naive Bayes, K-Nearest Neighbors, Decision Tree, Random Forest, Ada Boost, XG Boost, Light GBM, Cat Boost, and Support Vector Machine to model the flood damage incorporating all the influencing parameters. The Random Forest model is obtained as the best performing algorithm for flood damage prediction, with an accuracy of 84% for test data set. Further, a SHAP (SHapely Additive exPlanations) analysis is used to estimate the order of significance of the input variables and also to explain the reason for the prediction of flood damage state by the best performing machine learning model.