A time series forecasting project analyzing financial instrument prices using Vector Error Correction Model (VECM) and Vector Autoregression (VAR). It includes full pipeline from data import, stationarity testing, cointegration testing, model building, diagnostics, forecasting, and performance evaluation.
This repository contains a detailed Jupyter-based analysis and implementation of a time series forecasting workflow for financial instruments. The project performs:
- Exploratory data analysis and differencing
- Stationarity testing with the Augmented Dickey-Fuller (ADF) test
- Pairwise OLS regression for residual analysis
- Johansen cointegration test to identify long-run relationships
- Estimation of VECM with short- and long-run dynamics
- Model diagnostics (autocorrelation, normality)
- Reparameterization of VECM as VAR
- Forecasting and error evaluation (MAE, MAPE, etc.)
- Impulse Response Functions (IRF) and Forecast Error Variance Decomposition (FEVD)
TSA_2025_project_data_1.csv: Input time series dataset with 10 financial instrument seriesmain_notebook.ipynb: Full pipeline with explanations, tests, and visualizationREADME.md: Project documentation
| Technique | Description |
|---|---|
| ADF Test | For checking unit roots and stationarity of series |
| Johansen Test | For detecting cointegrated relationships among variables |
| VECM | To capture both long-run equilibrium and short-run dynamics |
| VAR | For vector autoregressive modeling and forecasting |
| IRF/FEVD | For interpreting shock dynamics and source of variances |
- Forecasting horizon: 25 periods
- X MAPE: ~3.92%, Y MAPE: ~2.43%
- Residuals: No autocorrelation, approximately normal
- X drives Y in long-run and short-run dynamics
- System is stable and well specified