This repository contains a collection of time series analysis and forecasting projects, implemented in Python using libraries such as pandas, NumPy, statsmodels, and Matplotlib.
- Autoregressive Models (AR, MA, ARMA) – capturing short-term dependencies in time series.
- ARIMA, SARIMA, SARIMAX – modeling and forecasting non-stationary series with trend and seasonality.
- Vector Autoregression (VAR) – analyzing and forecasting multiple interdependent time series.
- ARCH & GARCH – modeling time-varying volatility, commonly applied in finance and econometrics.
- Stationarity Testing – Augmented Dickey-Fuller (ADF) test for detecting unit roots and verifying stationarity.
- Heteroscedasticity Testing – Breusch-Pagan and White tests to detect changing variance in residuals.
- Normality & Autocorrelation Checks – Shapiro-Wilk, Jarque-Bera, and Ljung-Box test to validate residual behavior.
- Autocorrelation Analysis – ACF (Autocorrelation Function) and PACF (Partial Autocorrelation Function) plots for model identification and residual diagnostics.
- Parameter Significance – Z-statistics and p-values for evaluating model coefficients.
- Model Selection & Diagnostics – comparing models using AIC, BIC, and residual analysis to ensure robustness.
- Rolling Forecast & Backtesting – evaluating forecast stability over time with out-of-sample predictions.
- Forecast Evaluation Metrics – RMSE, MAE, MAPE, and Theil’s U-statistic to assess accuracy.
- Data Preprocessing – differencing, scaling, and seasonal decomposition to prepare data for modeling.