Improving the Forecasting of Car Sales using a Hybrid of Time Series Regression and Support Vector Regression with Google Trend Indices as Additional Predictors
Reny Dyah Puspitasari (a*), Dedy Dwi Prastyo (b)

a) Interdisiplinary School of Management and Technology,
Institut Teknologi Sepuluh Nopember,
Surabaya, 60111, Indonesia
*Correspondence, email: reny.dyah[at]gmail.com

b) Department of Statistics, Faculty of Science and Data Analytics,
Institut Teknologi Sepuluh Nopember,
Surabaya, 60111, Indonesia
e-mail: dedy-dp[at]statistika.its.ac.id

Abstract

The automotive industry is a leading sector that significantly contributes to the national economy. During the COVID-19 pandemic, there was a decrease in annual car sales by up to 40%, even a decrease in monthly car sales in May 2020 by up to 81,8%. On the other hand, car sales may relate with the searching behavior of internet users before they buy the car. The high growth of internet users in Indonesia recently, which has reached 210 million users, plays a significant role in people^s behavior for seeking information online before buying products. Google has summarized this search traffic for particular terms into the so-called Google Trend index. This research utilizes the Google Trend indices of chosen terms as additional predictors to improve the forecasting performance of car sales.
This research performs car sales forecasting model by combining linear and non-linear models through two modeling stages. The first step employs Time Series Regression (TSR) with dummy variables to accommodate the trends patterns, seasonality, and calendar variations pattern. In the second step, the non-linear model Support Vector Regression (SVR) is employed to model the residual of TSR. In addition, the Google Trend indices are used as additional predictors during the modeling TSR in the first step. The involvement of Google Trend indices in the model, along with the hybrid approach, is expected to improve forecasting accuracy.

Keywords: Car Sales Forecasting, Google Trends, Hybrid Model, Time Series Regression, Support Vector Regression

Topic: MATHEMATICS AND STATISTICS