The competition is about building a machine learning model to predict what event will come next given an ordinal sequence of customers’ life events. The dimension of the original training dataset is (1122007, 3) and the test data set size is 100870.
The most important reason why I won this competition is my creative feature engineering. The original dataset has only three columns and the number of available features is very small. I figured out the underlying meaning of event codes and created many new features based on the classification of event codes.
You can download the dataset from https://inclass.kaggle.com/c/allstate-event-prediction-challenge/data. Just run the file run.sh after downloading the data set.