Effect of Technical Indicator Combinations on the Performance of a Hybrid CNN-LSTM Model for Bitcoin Price Direction Classification

DOI

10.2991/978-94-6239-828-3_30How to use a DOI?

Keywords

Bitcoin; CNN-LSTM; Deep Learning; Technical Indicators; Ablation Study; Price Direction Classification

Abstract

Bitcoin price direction classification is challenging due to high volatility and complex temporal dynamics. This study investigates the contribution of four technical indicator categories-Momentum (MOM30), Relative Strength Index (RSI14, RSI30), Stochastic Oscillator (%K30, %D30, %K200, %D200), and Trend-Following (MACD)-within a hybrid CNN-LSTM model. Eight selected technical indicators are combined with OHLCV variables and evaluated through a systematic ablation study comprising six configurations. Minute-level BTC-USD data from January 2023 to January 2025 are resampled into 15-min intervals, with MA10/MA60 used to generate the binary target. To reduce look-ahead bias, each input sequence contains only the preceding 32 observations, while the target corresponds to the subsequent timestep. Model hyperparameters are selected using Random Search and fixed across experiments. The Full Model achieves 97.74% accuracy and 99.78% AUC-ROC, while the OHLCV-only Baseline achieves 55.88% accuracy. Trend-Following shows the largest observed marginal contribution (+0.65 percentage points), followed by Momentum (+0.38) and RSI (+0.22), whereas removing the Stochastic Oscillator yields a slight accuracy increase (+0.17). These results provide empirical evidence on the relative contribution of the evaluated indicator categories within the adopted CNN-LSTM setting.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Frederikus Felix Bintang Setyawan
AU  - Lailil Muflikhah
AU  - Rizal Setya Perdana
PY  - 2026
DA  - 2026/11/04
TI  - Effect of Technical Indicator Combinations on the Performance of a Hybrid CNN-LSTM Model for Bitcoin Price Direction Classification
BT  - Proceedings of the 2026 11th International Conference on Sustainable Information Engineering and Technology (SIET 2026)
PB  - Atlantis Press
SP  - 338
EP  - 348
SN  - 2589-4900
UR  - https://doi.org/10.2991/978-94-6239-828-3_30
DO  - 10.2991/978-94-6239-828-3_30
ID  - Setyawan2026
ER  -