Finhacks Data Challenge 2018 is the first Indonesia Kaggle-like data science competition that consists of three challenges such as Fraud Detection, Credit Scoring and ATM Cash Optimization.
In ATM Cash Optimization category, the goal is to predict the withdrawal amounts of more than 10,000 ATMs, that can be used for optimization to minimize the operational cost.
Dilan's prediction can get the 1st rank and outperforms other 123 models submitted by the participants. The winning solution is based on model combination of best model selection that consists of machine learning algorithm using XGBoost (with 4 different payday effects scenario) and traditional statistical forecasting method using Naïve (applied to 150 ATMs that contains many missing values).