Jogja Information Technology Session (JOINTS) was a nationwide Data Science competition conducted by Universitas Gadjah Mada, one of the most prestigious universities in Indonesia. It was conducted through Kaggle for many Indonesian data science enthusiasts. There was 44 teams with approximately 120 people that entered the competition, and my team successfully acquired first place in the public leaderboard & second place in the private leaderboard.
The challenge was to predict if a certain customer would like to prolong his/her insurance period. There were various features in the dataset, including Gender, Age, Area Code, Vehicle Age, and more. In this task, we were required to use machine learning of some sort. The dataset was heavily imbalanced and had a lot of missing values, requiring participants to preprocess and create a pipeline that can account for these phenomenons.
With a team of three whose linkedins can be found below, I successfully created a report and predictions to surpass many teams and acquired a high placement in the leaderboard. We utilized smote to handle imbalance, used CatBoost, XGBoost, LGBM, and RandomForests and ensembled it using a Voting Classifier in order to acquire high predictions on the dataset.
Link to the Kaggle Competition.
Johannes Joseph Billie Christian (https://www.linkedin.com/in/johannesjbilliec/)
I Made Atmavidya Virananda (https://www.linkedin.com/in/imavv/)