AutoML-TimeSeries-Forecasting

Use of ATSPy, a library to automate time series forecasting, using multiple models and ensembles.

Automated Models

1. ARIMA - Automated ARIMA Modelling
2. Prophet - Modeling Multiple Seasonality With Linear or Non-linear Growth
3. HWAAS - Exponential Smoothing With Additive Trend and Additive Seasonality
4. HWAMS - Exponential Smoothing with Additive Trend and Multiplicative Seasonality
5. PYAF - Feature Generating Model (slow and underforms)
6. NBEATS - Neural basis expansion analysis (now fixed at 20 Epochs)
7. Gluonts - RNN-based Model (now fixed at 20 Epochs)
8. TATS - Seasonal and Trend no Box Cox
9. TBAT - Trend and Box Cox
10. TBATS1 - Trend, Seasonal (one), and Box Cox
11. TBATP1 - TBATS1 but Seasonal Inference is Hardcoded by Periodicity
12. TBATS2 - TBATS1 With Two Seasonal Periods

Why AtsPy?

1. Implements all time series models in a unified manner by simply running AutomatedModel(df).
2. Automatically identify the seasonalities in your data using singular spectrum analysis, periodograms, and peak analysis.
3. Identifies and makes accessible the best model for your time series using in-sample validation methods.
4. Combines the predictions of all these models in a simple (average) and complex (GBM) ensembles for improved performance.
5. Where appropriate models have been developed to use GPU resources to speed up the automation process.

A blog post on Atspy : https://analyticsindiamag.com/hands-on-guide-to-atspy-for-automating-the-time-series-forecasting/