Demand Prediction in the Automobile Industry Independent of Big Data

Research output: Contribution to journalArticlepeer-review

2 Citations (Scopus)

Abstract

In recent years, various kinds of big data have been handled, and many variables are used in prediction model research. However, a gap between research and practice is thought to exist. As a result of adding variables that cannot be obtained at present as data representing the future to the explanatory variable, predicting the explanatory variable to apply it is necessary. There are cases wherein customers’ purchase intentions and attractiveness of products are used as explanatory variables; however, this is also not realistic because it is impossible to obtain product information from other companies before the products are launched. Therefore, to be useful for the production/sales plan, it is important that predictions are done using only currently available data, without additional surveys. In this study, gross domestic product and population are used as future data, models are built to predict the demand by body type in Japan on a monthly basis, up to 36 months ahead. Furthermore, in addition to earthquake and subsidy events, model change features were designed and incorporated into the models. The results showed that the prediction accuracy with an error of approximately 5%. It is believed that this study could suggest the possibility of feature quantity design and modeling instead of relying on large amounts of data.

Original languageEnglish
Pages (from-to)249-270
Number of pages22
JournalAnnals of Data Science
Volume9
Issue number2
DOIs
Publication statusPublished - Apr 2022

Keywords

  • Earthquake
  • GDP
  • Model change
  • State space model
  • Subsidy

Fingerprint

Dive into the research topics of 'Demand Prediction in the Automobile Industry Independent of Big Data'. Together they form a unique fingerprint.

Cite this