Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal STATISTIKA
Kamisan, Nur Arina Bazilah
Unknown Affiliation
Decision Sciences, Operations Research & Management Mathematics

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

Found 1 Documents
Search

 1 
Optimizing Train-Test Splits for LSTM and MLP Models in Bitcoin Price Forecasting Accuracy Kamisan, Nur Arina Bazilah; Lee, Muhammad Hisyam; Sulandari, Winita
Statistika Vol. 25 No. 2 (2025): Statistika
Publisher : Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Islam Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29313/statistika.v25i2.6989

Abstract

Abstract. This study investigates the application and efficiency of two machine learning models, Long-Short Term Memory (LSTM) and Multilayer Perceptron (MLP), for cryptocurrency price forecasting, using Bitcoin as a case study. MLP is a feedforward neural network that learns patterns from independent data, while LSTM is a recurrent network that remembers past information to handle sequential or time-series data. The rapid growth and volatility of cryptocurrencies underscore the need for accurate price predictions to support investor’s and trader’s decision-making. The study aims to identify the optimal train-test splitting ratio for each machine learning model and to forecast Bitcoin prices over a 120 days. The daily Bitcoin price data is obtained from the Bitcoin website recorded from January 2018 until March 2021. Model performance was evaluated using Akaike Information Criterion (AIC), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). Experimental results demonstrate that both models exhibit strong predictive capabilities; the LSTM model consistently outperforms MLP in accuracy and reliability, achieving lower MAE, MAPE, and AIC values. These findings highlight LSTM’s effectiveness for forecasting volatile financial data and provide insights into selecting appropriate data-splitting ratios to improved model performance.
Co-Authors Lee, Muhammad Hisyam Winita Sulandari