README.md

In AviBlinder/tsSelect: Execution of Time Series Models

"tsSelect Package - Pick the best time series model"

This package has as input a Time Series object.

The first step fits a series of models:

• Mean model
• Naive model (all forecasts = last observation)
• Seasonal Naive model (each forecast is equal to the last observed value from the same season)
• 'drift' model (adds a "trend" over time to the naive method)
• tslm_1 (linear-regression by 'trend')

• tslm_2 (linear-regression by 'trend' and 'season')

• Exponential smoothing models (with several variations)

• ARIMA (with and without stepwise selection)

The second step picks the model with lowest error measures, either by a specific measure or by majority vote, among all possible error measures.

The different error measures are: ME, RMSE, MAE, MPE ,MAPE, MASE, ACF1

Execution example:

library(tsSelect)

run_models(ts_object)

AviBlinder/tsSelect documentation built on May 5, 2019, 9:22 a.m.