Pattern recognition and forecasting from multiple financial time series data and news

Stock price prediction is becoming popular to many researchers and it is a challenging task. With the increased advantages of using machine learning models, the creation of an accurate prediction model becomes a hot topic in the market. With the application of recurrent neural networks, this projec...

Full description

Saved in:
Bibliographic Details
Main Author: Yee Aung, Su Wai
Other Authors: Loke Yuan Ren
Format: Final Year Project
Language:English
Published: Nanyang Technological University 2021
Subjects:
Online Access:https://hdl.handle.net/10356/149273
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Stock price prediction is becoming popular to many researchers and it is a challenging task. With the increased advantages of using machine learning models, the creation of an accurate prediction model becomes a hot topic in the market. With the application of recurrent neural networks, this project proposes a time series prediction model to capture the complex features such as non-linearity, non-stationary and sequence correlation of financial time series. This project presents a hybrid model of convolutional network (CNN) and long short-term memory neural network (LSTM) with Attention Mechanism for classifying finance data from Yahoo Inc. and the prediction of the 3-day ahead opening and closing prices. Historical price data for each stock and related tweets from Twitter will be used to train the proposed model. The empirical results show that the CNN-LSTM+Attention model provides a better prediction, and it shows excellent effects on the static prediction and dynamic trend prediction of the financial time series. Additionally, the transformation of output values to price change instead of the actual stock prices increases the accuracy in prediction results. The experimental results show that the proposed approaches give good performance in predicting the stock market prices. It also provides a lower mean squared error (MSE), lower mean absolute error (MAE), higher R-squared values and thus can be considered as superior to other models in stock price prediction.