Importance of Trading Volume-Based Features and Modelling Parameters in Daily Stock Trading Using Neural Networks

Abstract

This thesis analyses the role of combination of Close Price and Volume Trading value in terms of retrieved profit after short-term stock trading by using machine learning method. Thus, the Volume trading value and Close Price is selected as input of 2 methods Support Vector Machines (SVM)-based predictor and Artificial Neural Network (ANN) because both 2 methods prove itself excellent performance among machine learning methods. To decide the moment trading on simulation, strategy rule based on parametric model and close price is also produced. SVM conducts training and testing procedures to find trading signal. Profitability of these simulations is evaluated against a traditional Buy-and-hold strategy and compared with total average profit among the whole dataset. We tested the proposed model with daily trading data from 2001 to 2015. The empirical result showed satisfactory performance that the accuracy and ratio of trading profit of proposed model trading volume can be a good factor on forecasting profit.