MACHINE LEARNING FOR STOCK MARKET SECURITY MEASUREMENT: A COMPARATIVE ANALYSIS OF SUPERVISED, UNSUPERVISED, AND DEEP LEARNING MODELS

• DOI: 10.55640/ijns-04-01-06
• Publication Date: 2024-11-22T15:38:04Z
• Authors: Md Shakhaowat Hos...; S M Shadul Islam ...; ; Md Mohibur Rahman; ; Sanjida Akter Tisha; ; Farhan Shakil; ; Ashim Chandra Das; ; Radha Das; ; Sadia Sultana
• Keywords: Interpretability; Feature Engineering; Autoencoder; Supervised Learning; Data pre-processing

This study presents a comprehensive analysis of machine learning techniques for measuring and predicting security in stock markets, comparing the performance supervised, unsupervised, deep models. Using diverse dataset from Kaggle that includes historical prices, financial news sentiment, company fundamentals, macroeconomic indicators, we applied feature engineering rigorous preprocessing methods to optimize model accuracy. The evaluated Random Forest, Support Vector Machines (SVM), K-Means clustering, Long Short-Term Memory (LSTM) networks across key metrics. Results indicate Forest outperformed other models classification tasks with an accuracy 92%, making it highly effective real-time assessment. SVM also demonstrated strong capabilities, particularly high-dimensional spaces, 88%. DBSCAN clustering algorithms excelled anomaly detection, identifying unusual patterns could signal market irregularities. LSTM models, designed time-series forecasting, achieved root mean square error (RMSE) 1.78, proving their utility future trends but requiring more computational resources.Our findings suggest hybrid approach, combining strengths supervised can provide robust solution measurement. By leveraging explainable AI such as SHAP LIME, improved interpretability, these predictions actionable stakeholders. research highlights potential monitoring supports growing integration finance industry.