Artificial Intelligence in the Stock Market: Quantitative Technical Analysis, Model Weight Optimization, and Financial Sentiment Evaluation to Predict Stock Prices

Abstract

This research paper demonstrates the modern implementation of artificial intelligence in predicting the stock market. In doing so, it focuses on stock price prediction through A.I. models and machine learning algorithms to maximize profit potential, improve investments, and eliminate risk. Potentially lucrative company stocks and shares have attracted investors as well as general interest in the stock market for decades (Malkiel, 1973, p. 269), leading more people to try to forecast the rise or fall of market prices. However, industry volatility and the seemingly unpredictable nature of the stock market have led many buyers to invest impulsively or make poor purchasing decisions like selling or buying shares at the wrong times. The paper outlines the training and testing of A.I. models, such as linear regressions and neural networks, on collected, classified data to generate accurate predictions. The program also utilizes natural language processing (NLP) through a deep learning model with transformer-produced sentence embeddings, allowing the algorithm to consider relevant socioeconomic and sociopolitical news to produce predicted prices at an even higher level of accuracy. These models achieved average prediction errors of 0.12% for the stock prices of Amazon, 0.13% for the stock prices of Google, and 0.07% for Microsoft’s stock prices on the testing data sets, over a two-week testing period. This paper ultimately evaluates existing prediction methods and builds on robust machine learning systems to offer more efficient estimation models.