Deep learning algorithms for classification of financial time series data

• Main Author: Saputra, Kevin
• Other Authors: Ponnuthurai Nagaratnam Suganthan
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/139396
• Institution: Nanyang Technological University

Nowadays in this modern world, deep learning methods are used and applied more often in our daily life. The main objective of this project is to evaluate and investigate the application of various deep learning methods in forecasting and classifying financial time series data. It is remarkably difficult in forecasting and classifying time series data due to the natural characteristic of financial time series data and financial world in general, which is non-linear, non-stationary, and unpredictable. FOREX, gold, and oil prices are considered as crucial indicators of the world’s economic stability and growth. This is the exact reason for the capability to predict the precise changes in these prices is highly considered. However, as it is preposterous to predict the price changes precisely, a robust forecasting and classifying models are greatly desired. One of the widely used deep learning methods is Recurrent Neural Network (RNN). RNN is a type of Artificial Neural Network (ANN) where a directed graph formed based on connections between nodes in which a sequence of information may flow. This architecture allows the information to be processed to be different for each time step while maintaining some important initial information. Various types of RNN models used in this project to forecast and classify the financial data. For the forecast to be applicable in real life, the classification of whether the price would increase or decrease would be the main point of interest. Thus, utilizing the forecasting results, we may calculate the increase and decrease the accuracy of the models. Moreover, to further boost the performance, some preprocessing data decomposition would be used, such as Discrete Wavelet Transform (DWT), Empirical Mode Decomposition (EMD), as well as Ensemble Empirical Mode Decomposition (EEMD) to be precise.