Stock Market Decision with Machine-Learning-Based Price Prediction Models and AHP

Ye Lanxin

Abstract

As stock market investment and stock market price prediction has always been a popular topic to explore on, this final year project paper is aimed to deliver a stock market investment decision analysis methodology for individual investors to decide if they should invest in the determined stock market or not. The stock market that this paper takes the dataset from is Shanghai Stock Price Composite Index (SSE). This paper firstly researches on the literature review of the existing stock price prediction machine learning models. This paper would adopt some of the existing machine learning prediction methodologies and also develop and optimize the current methodologies based on the existing ones. As the current prediction methodologies contain either regression models or classification models, then this paper makes the comparison the two regression models (Long Short-Term Memory and Linear Regression) and two classification models (Logistic Regression and K-Nearest Neighbors) to figure out the prediction model which could provide the best prediction results with lower error or higher accuracy. Then taking the machine learning results as one of the main criteria, this paper also constructs an Analytic Hierarchy Process (AHP) model to finally make the decision between two alternatives: invest in SSE or do not invest in SSE. The other decision criteria are policy intervention, inflation, COVID-19 affection, stock market outlook, and other potential factors. The main criteria matrix is designed wisely with previous research support. The last chapter also provides a user guideline on the entire decision process that individual investors could use as the instruction. As the models covered in this paper are not a comprehensive representative of the real-word situation because the real-world situation is almost impossible to simulate perfectly, and also the complexity of the machine learning prediction models would affect the result, users or the individual investors should not totally rely on this paper to make their choices and decisions.