What is it about?
The main objective of this study is to analyse whether the combination of regional predictions generated with machine learning (ML) models leads to improved forecast accuracy. With this aim we construct one set of forecasts by estimating models on the aggregate series, another set by using the same models to forecast the individual series prior to aggregation, and then we compare the accuracy of both approaches. We use three ML techniques: Support Vector Regression (SVR), Gaussian Process Regression (GPR) and Neural Network (NN) models. We use an ARMA model as a benchmark.
Photo by Sergi Kabrera on Unsplash
Why is it important?
The main aim of this study is to analyse the relative improvement on forecast accuracy of ML methods over a linear stochastic process using two alternative approaches. First, we apply the direct approach, which consists in forecasting the aggregate series. Then, we use the same models to forecast the individual series for each region prior to aggregation at a national level. Finally, we compare the forecasting performance of both approaches. Several authors have found evidence that combining forecasts tends to yield more accurate predictions than direct approaches (Bates and Granger, 1969; Stock and Watson, 2004; Ruth, 2008). We extend previous research by assessing this approach with ML techniques at a regional level. We find that ML methods improve their forecasting performance with respect to the benchmark as forecast horizons increase, suggesting the suitability of these techniques for mid- and long-term forecasting. In spite of the fact that the disaggregated approach yields more accurate predictions, the improvement over the benchmark occurs for shorter forecast horizons with the direct approach.
Read the Original
This page is a summary of: Combination forecasts of tourism demand with machine learning models, Applied Economics Letters, September 2015, Taylor & Francis,
You can read the full text:
The following have contributed to this page