Abstract

The purpose of this research is to develop sophisticated machine-learning models that can predict stock price movements, taking into account the multifaceted influences that create volatility in financial markets. To navigate the complexities of market dynamics effectively, investors, financial institutions, and academia rely on precision stock price predictions. This research generates actionable buy and sell signals by using historical stock data, technical market indicators, Linear Support Vector Machines (LSVM), Neural Networks, and Logistic Regression. A distinctive feature of this model is it uses market indicators such as PPO, MACD, RSI Signal, Bollinger Bands, ROC Signal, and DX Signal instead of conventional methods that might include sentiment analysis from external data sources. By creating additional indicators, this thesis broadens the analytical scope by using a carefully curated dataset from Yahoo Finance, with a focus on Johnson & Johnson. In addition to its improved prediction accuracy, precision, and F1 scores, the proposed model also allows traders to make informed decisions about when to buy or sell, potentially enhancing portfolio performance by as much as 70%. Among all models, Logistic Regression emerged as the best performer, yielding an impressive 96.467% accuracy for predicting stock price movements. As part of this research, we examine how stock price prediction mechanisms work in detail and introduce a model that combines technical analysis with machine learning insights, paving the way for more accurate and reliable investment decisions.

Publication Date

5-23-2024

Document Type

Thesis

Student Type

Graduate

Degree Name

Professional Studies (MS)

Department, Program, or Center

Graduate Programs & Research

Advisor

Sanjay Modak

Advisor/Committee Member

Hammou Messatfa

Campus

RIT Dubai

Download

Share

COinS