Date of Award

12-4-2024

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

Mariofanna Milanova

Abstract

The inherent challenges of financial time series forecasting demand advanced modeling techniques for reliable predictions. Financial data, belonging to the category of multimedia data, contains an extensive quantity of information that is widely used for data analysis and decision-making tasks. Effective financial time series forecasting is crucial for financial risk management and the formulation of investment decisions. Accurate prediction of stock prices is a subject of study in the domains of investing and national policy. This problem appears to be challenging due to the presence of multi-noise, nonlinearity, volatility, and chaotic natures in stocks. This paper proposes a novel financial time series forecasting model based on the deep learning ensemble model, LSTM-mTrans-MLP, which integrates the Long Short-term Memory (LSTM) network, a modified Transformer network, and Multilayered Perception (MLP). Long Short-Term Memory (LSTM) models are designed to remember information for extended durations, rendering them very efficient for analyzing time series data when the context from earlier in the sequence is significant. Furthermore, LSTMs are relatively robust to noise and missing data compared to CNN and conventional RNN models because of their ability to selectively retain and discard information. On the other hand, Transformer models utilize a self-attention mechanism that enables the model to simultaneously consider all segments of the input sequence while processing a particular position. Transformers can be scaled to handle large datasets and complex tasks by increasing the number of layers and attention heads. The self-attention mechanism enables Transformers to selectively concentrate on the most relevant parts of the input sequence, hence enhancing the model's capacity to comprehend the context and relationships hidden in the data. Finally, MLPs can learn complex nonlinear relationships between input and output variables. MLPs can generalize well to unknown data, making them useful for real-world applications by approximating complex functions with arbitrary precision. By integrating LSTM, modified Transformer, and MLP, the suggested model demonstrates exceptional performance in terms of forecasting capability, robustness, and enhanced sensitivity. Extensive experiments are conducted on the multiple financial datasets such as Bitcoin, Shanghai Composite Index, China Unicom, CSI 300, Google, and Amazon Stock Market. The experimental results verify the effectiveness and robustness of the proposed LSTM-mTrans-MLP network model compared with the benchmark and SOTA models, which can give an important inference for investors and decision-makers.

Download

Share

COinS