Autoregressor: simple and robust time series model selection

auto_regressor.py automates the process of fitting and using a robust time series regression (OLS) model with lags. Plug in your data, run the script and the code will fit the best model + provide predictions.

Autoregressor: Simple and Robust Time Series Model Selection

This is what the code does:

1. Loading Your Dataset:

• The load_df function is employed to import your dataset. It's crucial that your dataset contains either a datetime column or a column explicitly named 'date' or 'Date'.

2. Preprocessing and Analysis:

• Call remove_colinear_features to weed out collinear variables, ensuring model integrity.
• Conduct an exploratory analysis to scrutinize relationships within your data and inspect for any non-linear dynamics between variables.

3. Creating Splits and Fitting Models:

• Use create_splits to segment your data into distinct training and testing sets, integrating lagged features for thorough analysis.
• Apply regression_OLS to fit models, selectively honing in on significant features through an automated process.

4. Evaluation and Prediction:

• Gauge model performance using oos_summary_stats, allowing for comparison of models across different data splits.
• Facilitate both in-sample and out-of-sample predictions with fit_and_predict, providing visual representation for in-depth evaluation.

Optional Features:

• Non-linearity Detection: Automatically detects and adjusts for non-linear relationships between variables, enhancing model accuracy.
• Customization Options: Offers flexibility in adjusting the number of lags, splits, training share, VIF threshold, and p-value cutoff for tailored analysis.