Implementasi GridSearch dalam Meningkatkan Kinerja Model Support Vector Regresion (SVR) utuk Prediksi Penjualan Produk (Studi kasus : Meubel Rohman Jaya)

Ahmad Baidowi Eko Fitra Firmanda (Universitas Nurul Jadid)
Ahmad Hudawi AS (Universitas Nurul Jadid)
Abu Tholib (Universitas Nurul Jadid)
Juvinal Ximenes Guterres (Universidade Oriental Timur Lorosa’e)

Explore IT : Jurnal Keilmuan dan Aplikasi Teknik Informatika

In the era of digitalization, product sales forecasting plays a crucial role for companies in estimating future demand. Meubel Rohman Jaya, a furniture business established since 2010, requires accurate predictions to optimize stock availability with the variety of products they produce. This research aims to forecast furniture product sales using the Support Vector Regression (SVR) algorithm with GridSearch optimization. Sales data of 11 furniture products over 30 months (January 2021 - June 2023) were processed through data collection and preprocessing. Modeling was performed using SVR without optimization and SVR with GridSearch optimization to obtain the best parameters. Predictions were generated and then evaluated using the Mean Absolute Percentage Error (MAPE) metric. The results showed that SVR without optimization achieved a MAPE of 40.39%, while SVR with GridSearch achieved a MAPE of 0.45%, indicating a significant increase in accuracy. GridSearch optimization has proven effective in improving prediction performance and is highly recommended for implementation in forecasting product sales at Meubel Rohman Jaya.