The dataset used is AirPassengers.csv the public dataset which contains the number of passengers who traveled by airline per month from 1949 to 1960. The dataset has 144 observations and two columns: Month and the number of passengers..
-
Time series decomposition (STL & Classical Decomposition)
- Level
- Trend
- Seasonality
- Noise
-
Stationarity
- AC and PAC plots
- Rolling mean and std
- Dickey-Fuller test
-
Making our time series stationary
- Difference transform
- The Box-Cox transformation
- Smoothing
- Moving average
I used Rolling Forward for evalution next day
- Mean Absolute Error (MAE)
- Mean Absolute Percentage Error (MAPE)
- Root Mean Squared Error (RMSE)
| Model | mae | rmse | mape |
|---|---|---|---|
| Moving Average (MA) | 36.1 | 48.36 | 12.25 |
| Weighted Moving Average (WMA) | 30.61 | 40.90 | 10.45 |
| Linear Model | 34.4 | 45.7 | 12.3 |
| Linear Model With STL | 34.4 | 45.7 | 12.3 |
| Naïve | 25.8 | 33.7 | 9.01 |
| MLR | 5.259989974446297e-14 | 6.422239031165233e-14 | 1.8281416595513634e-14 |
| ARIMA | 26.0 | 32.98 | 32.5 |
| SARIMA | 17.8 | 22.63 | 5.79 |
| AR (Autoregressive) | 24.8 | 31.036 | 8.94 |
| Advanced Smoothing Single Exp | 28.43 | 36.26 | 8.9 |
| Advanced Smoothing Double exp | 31.5 | 39.9 | 3.34/td> |
| Advanced Smoothing Triple exp | 9.5 | 12.7 | 9.7 |
| Prophet | 6.8 | 6.87 | 3.14 |
| XGBOOST | 30.5 | 41.3 | 6.97 |
| LGBMRegressor | 27.0 | 38.0 | 6.22 |
| KNeighborsRegressor | 22.15 | 25.48 | 4.72 |
| SVR | 44.57 | 56.77 | 9.259 |