Effective Short-Term Forecasting for Daily Time Series with Complex Seasonal Patterns

Abstract

Time series forecasting is a process of estimating future value based on historical data and it plays a crucial role in business decision making in various domains. The selection of a suitable time series forecasting technique depends upon the presence of the following four components: trend, seasonal, cyclical and irregular. Traditional time series techniques like ARIMA, SARIMA, ETS are designed to handle single seasonality in a time series, but with theexistence of multiple seasonality, these techniques fail to perform satisfactorily. Thus, there is a need to use advanced techniques like BATS and TBATS for multiple seasonal data. The main objective of this paper is to develop a successful prediction model after comparing BATS and TBATS models for short-term forecasting of complex time series. The daily time series of natural gas consumption for a manufacturing unit of BHEL, India that exhibits multiple seasonality is used for evaluation purpose. The results of the analysis conclude that TBATS outperforms BATS model for a span of different short-term prediction horizon. The main reason behind is be attributed to thefact that less number of parameters is required to be estimated in TBATS model in comparison to the BATS model.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 41 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.