Integrasi Model Sarima dengan Optimasi Algoritma Genetika dalam Peramalan Penjualan Sepeda Motor di Indonesia

Abstract

A variety of factors will influence the sales of a product, so it is necessary to analyze it to anticipate
changes in the business environment. This research explores how accuracy in forecasting can help
companies make policies and determine the right strategy. This research aims to test the
performance of the Seasonal Autoregressive Integrated Moving Average (SARIMA) and Genetic
Algorithm (GA) methods in forecasting motorcycle sales in Indonesia. The SARIMA model
forecasts seasonal patterns formed from sales data, while GA is implemented to optimize model
parameters to improve prediction accuracy. The data used in this research comes from motorcycle
sales reports in Indonesia from 2021 to 2024. Forecasting performance with the GA-SARIMA
model is tested and evaluated using the AIC, MAD, MSE, MAPE, and MAE parameters. The
performance of all parameters shows that the GA-SARIMA model has the lowest value, so it
performs better than the traditional SARIMA. The validation results using actual sales show that
the GA-SARIMA model shows an error in January 2025 of 2,698, while in February 2025, it is
71,924. Despite the error, the forecasting is still within the upper and lower limits, so the resulting
model can still forecast motorcycle sales in Indonesia.