Earthquake-Damage-Modelling

Predicting damage grade to buildings after an earthquake based on their construction and location. The Government of Nepal carried out a survey after the 2015 earthquake to assess the damage caused and provide appropriate assistance for re-building. The purpose of this project is to use this data, that has been made available as a part of a competition and predict the level of damage of buildings. It is a classification problem. The project focussed on feature engineering techniques that addressed categorical features with high cardinality. Two techniques, namely target encoding and entity encoding were adopted. Altogether, the results of three models namely, Random Forests, Neural Networks and Logistic Regression; were compared. It is a multi-class classification problem wherein the model predicts the damage severity that could be low, medium or high; based on the location and construction of the buildings.