Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal Journal of Artificial Intelligence and Technology Information
Muhammad Iszul Wilsa
Universitas Islam Negeri Siber Syekh Nurjati Cirebon
Computer Science & IT Engineering

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Comparative Analysis of Hybrid ARIMA-LSTM against Statistical and Machine Learning Benchmarks for Commodity Stock Muhammad Iszul Wilsa; Heru Purnomo Kurniawan; Rizki Dewantara; Dinda Febrihastatiwi; Indri Setiawati
Journal of Artificial Intelligence and Technology Information (JAITI) Vol. 4 No. 2 (2026): Volume 4 Number 2 June 2026
Publisher : PT. Tech Cart Press

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.58602/jaiti.v4i2.260

Abstract

Predicting stock prices in Indonesia’s commodities and energy sectors is a complex challenge due to high volatility influenced by global market dynamics and macroeconomic factors. This study aims to test the robustness of the ARIMA-LSTM hybrid model in predicting closing stock prices for six major issuers: ADRO, PTBA, MEDC, ANTM, MDKA, and AALI. The proposed approach employs a dual-input strategy that integrates 27 technical indicators with the linear residuals from the ARIMA model. The research methodology begins with data decomposition using the ARIMA model to capture linear components, followed by modeling the residuals using Long Short-Term Memory (LSTM) to capture complex non-linear patterns. The experimental results show that the hybrid model consistently delivers the best performance compared to single models such as ARIMA, Random Forest, and Single LSTM across all test datasets. In the 1-step-ahead scenario, the hybrid model achieved the lowest average MAPE of 2.20%, while in the 5-step-ahead scenario, the error rate remained at 3.98%. A key finding of this research is the hybrid architecture’s ability to mitigate the extreme overfitting experienced by the Single LSTM model, while providing better prediction stability against variations in issuer characteristics. This study concludes that the integration of statistical decomposition and deep learning provides a reliable framework for investors and analysts to make data-driven decisions amid the volatile fluctuations of the Indonesian capital market.
Co-Authors Dinda Febrihastatiwi Heru Purnomo Kurniawan Indri Setiawati Rizki Dewantara