Forecasting the Used Car Price Index Using Time Series Forecasting Methods

Abstract

This research aims to: 1) study the forecasting of the used car price index using the Classical Decomposition Method, Double Exponential Smoothing, and Holt-Winters’ Exponential Smoothing Method, and 2) compare the forecasting methods for the used car price index, including Classical Decomposition, Double Exponential Smoothing, and Holt-Winters’ Exponential Smoothing Method. The most suitable forecasting method was determined by the lowest Mean Absolute Deviation (MAD) and Mean Absolute Percentage Error (MAPE) values.

The results revealed that: 1) the Classical Decomposition of time series using a Multiplicative Model provided the lowest error, with MAD = 6.1813 and MAPE = 6.5451. Forecasting using Double Exponential Smoothing (α = 0.8, y = 0.8) resulted in MAD = 3.9621 and MAPE = 4.1532, while Holt-Winters’ Exponential Smoothing Method (α = 0.8, y = 0.6, δ = 0.8) resulted in MAD = 3.9535 and MAPE = 4.0722. 2) The Holt-Winters’ Exponential Smoothing Method was found to be the most appropriate for this time series data, as it produced the lowest MAD and MAPE values.