Date of Award

12-1-2023

Language

English

Document Type

Master's Thesis

Degree Name

Master of Arts (MA)

College/School/Department

Department of Mathematics and Statistics

Dissertation/Thesis Chair

Karin Reinhold

Committee Members

Joshua Isralowitz, Boris Goldfarb, Yunlong Feng

Keywords

Forecasting, LSTM, Machine Learning, Neural Network, Stock Market, Time Series

Subject Categories

Statistics and Probability

Abstract

This thesis investigates applying a Long Short-Term Memory (LSTM) Neural Network for forecasting optimal times to enter and exit the stock market. Given the inherent imbalance in the dataset, where most days are not opportune for market actions, and the challenge of predicting such infrequent occurrences, we will employ a five-day window before and after the identified optimal market entry or exit time. A prediction is considered correct if it is within this five-day interval. The model was trained on historical S&P 500 data, using features such as exponential and simple moving averages of the closing price, the ten-day percentage change, the application of a smoothing function, and the calculation of the slope between each day and the preceding day, derived from the smoothing function. This approach aims to enhance the model's ability to capture patterns and improve the prediction accuracy for optimal market timing.

Download

Share

COinS