Dividend Prediction with LSTM
Forecasting future dividends using a sequence model (LSTM) on historical market data. Built with a clean data pipeline, reliable modeling, and an interactive Tableau dashboard.
Data Pipeline
- Fetched price/dividend data via
yfinanceand unified it to a consistent datetime index. - Handled missing values and ensured sorted, aligned series per ticker.
- Prepared supervised sequences with a look-back window → next dividend format.
- Split data by time (train, validation, test) to prevent leakage.
LSTM Model
- Implemented an LSTM in Keras to predict future dividends.
- Tuned hyperparameters with Adam optimizer and early stopping.
- Evaluated predictions on held-out time segments with clear visualizations.
Deployment
The trained model was deployed for reuse within my data pipeline, enabling reproducible inference and integration into visualizations.
Tableau Dashboard
- Interactive ticker selection to explore multiple ETFs/stocks.
- KPIs: latest predicted dividend, year-over-year change, trailing 12-month totals.
- Timeline view overlaying historical and LSTM-predicted dividends.
- Dynamic filters and tooltips for exploring results.
- Export option for downloading filtered datasets.
Screenshots
