Monelytics
The problem
Comparing forecasting approaches requires a consistent data pipeline and evaluation rather than a single prediction.
How I solved it
I led exploratory analysis, preprocessing, feature engineering and LSTM implementation. The team compared classical ML, time-series and deep-learning approaches and connected the analysis to an interactive Streamlit interface.
- AI-powered stock prediction for Indonesia's big four banks
- Implemented multiple ML models (Linear Regression, SVR, Random Forest, ARIMA)
- Integrated DL models (CNN, ANN, LSTM, Prophet)
- Used RMSE and MAE for performance metrics
- Led data exploration and feature engineering
The results
- analysis coverage
- 4 banks
- team modeling scope
- 9 approaches
Compared nine approaches for four Indonesian banks. The team delivered five modeling notebooks, four bank-specific LSTM files and a Streamlit application with RMSE/MAE comparisons.
Evaluation context. Team project. Scaler fitting before a shuffled split, date alignment and the LSTM UI model-file selection limit forecasting claims. Model artifacts exist; reliable future-price performance is not established.