This repo aims to serve as a beginner friendly guide to various time series analysis, decomposition and forecasting techniques.
- Analysis And Visualization Of The Airline Passengers Data
- Apply Data Decmposition Techniques
- Apply Different Forecasting Techniques And Comment On each Technique Whether It Should Be Used Or Not And Under What Conditions
- Evaluate Different Forecasting Techinques Using Fixed Partitioning, Roll-Forward Partitioning, Cross-Validation
The Data Used Can Be Found Here. The Data Contains Monthly Data Of Airline Passengers From 1949-01 To 1960-12. The Below Is A Visualization of The Data:
1 - Simple Moving Average
2 - Naïve Forecasting
3 - Weighted Moving Average
4 - Simple Linear Regression
5 - Classical Decomposition
6 - STL
7-1 Arima
7-2 S-ARIMA
8-1 Single Exponential Smoothing
8-2 Double Exponential Smoothing
8-3 Triple Exponential Smoothing
9 - Facebook Prophet Algorithm
10 - Supervised Machine Learning (XGBoost)
The evaluation metric used is RMSE and each techniques is evaluated using fixed partitioning, roll-forward partitioning, cross-validation. Below are the results of all the forecasting techniques used in the project:
| Technique | Fixed | Roll Forward | Cross Validation |
|---|---|---|---|
| Simple Moving Average | 113.8 | 55.38 | 139.60 |
| Naïve Forecasting | 137.3 | 44.20 | 142.60 |
| Weighted Moving Average | 84.6 | 52.40 | 183.00 |
| Linear Regression | 74.7 | 55.40 | 80.00 |
| Classical Decomp. | 60.0 | 54.70 | 84.10 |
| STL Decomp. | 68.7 | 66.20 | 80.70 |
| ARIMA | 126.1 | 39.72 | 141.83 |
| S-ARIMA | 39.9 | 11.50 | 55.10 |
| Single Exp. Smoothing | 139.7 | 47.10 | 140.30 |
| Double Additive Exp. Smoothing | 118.3 | 44.00 | 93.70 |
| Double Multiplicative Exp. Smoothing | 107.60 | 43.70 | 104.90 |
| Triple Additive Exp. Smoothing | 35.70 | 12.30 | 57.40 |
| Triple Multiplicative Exp. Smoothing | 13.80 | 10.30 | 40.10 |
| Facebook Prophet | 40.3 | 31.50 | 45.40 |
| XGBoost Regressor | 53.1 | NaN | 104.20 |