ECONOPHYSICS APPLICATION IN STOCK PRICE TREND PREDICTION ANALYSIS IN THE INDONESIAN COAL SECTOR USING A RESTRICTED BOLTZMANN MACHINE AS A FEATURE EXTRACTOR

• Main Author: Graham, Ariel
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/81662
• Institution: Institut Teknologi Bandung

Econophysics is an interdisciplinary field that incorporates physics to address pheno- mena within economic systems, with one key focus being the analysis of financial mar- kets. Financial market prediction is crucial for investors’ financial decision-making. Approaches to financial market prediction include statistical models, machine learning, and deep learning methods in artificial intelligence. However, the complexity and high dimensionality of financial market data often hinder model performance. Therefore, a machine learning model known as the Restricted Boltzmann Machine (RBM) is pro- posed for feature extracting. This study aims to analyze the impact of RBM hyper- parameters on stock price trend prediction and to compare Gaussian-Bernoulli RBM with Bernoulli RBM. The study involves using RBMs to predict stock price trends for four coal sector stocks, namely PTBA, ITMG, KKGI, and BUMI, utilizing Support Vector Machine (SVM), Random Forest, Multi-Layer Perceptron Classifier, and Long Short-Term Memory (LSTM). The prediction results are evaluated based on accuracy, precision, and negative predictive value using evaluation matrices. The classification model performance is also reviewed using the receiver operating characteristic (ROC) curve and its area under the curve (AUC). The study concludes that optimizing RBM hyperparameters affects the resulting accuracy, and Bernoulli RBM yields better resul- ts than Gaussian-Bernoulli RBM. However, some stock data and classification models show no significant impact from optimized hyperparameters, due to factors such as da- ta characteristics, hyperparameter options in fine-tuning searches, and the selection of parameters to optimize.