Mexican Automotive Industry Sales Behavior During the COVID-19 Pandemic
Journal
Advances in Soft Computing
Lecture Notes in Computer Science
ISSN
0302-9743
1611-3349
Date Issued
2021
Author(s)
Ramírez, Jorge
Alarcón, Joaquín
Calzada, Gustavo
Type
text::book::book part
Abstract
This work shows the results of forecasting the behavior of sales in the Mexican automotive industry in a simulated scenario without COVID-19 and comparing it with the actual sales numbers. As this pandemic has caused traditional forecasting techniques to show poor performance and low prediction quality, this work aims to estimate the number of sales lost during the pandemic, using a machine learning model based on several explanatory variables and predicting those variables without the influence of the COVID-19 pandemic. Three different regression models were tested (Linear regression, Random Forest and Neural Network) creating scenarios and incorporating different variables into the models. Random Forest with 3 variables shows the highest predictive power. This model applied on forecast variables without pandemic’s impact predicts 1,342,028 units sold between February 2020 and January 2021, representing a 29.76% drop in sales and a total impact of 416,324 sales lost due to the pandemic. © 2021, Springer Nature Switzerland AG.