UCSC Digital Library >
Computer Science Masters >
Master of Computer Science - 2017 >
Please use this identifier to cite or link to this item:
|Title: ||CSE Stock Selection According to the Future Profit Gain using Pattern Recognition|
|Authors: ||Altaf, M.A.M.|
|Issue Date: ||2017|
|Abstract: ||The prediction of Colombo Stock Exchange stock price direction may serve as an early recommendation system for short-term investors and as an early financial distress warning system for long-term shareholders. Many stock prediction studies focus on using statistical methods, Technical Analysis, fundamental market analysis, and while others use hybrid model based approaches. These methods are employed with different success rate in practice. None of these methods is proved as a consistently acceptable stock price prediction tool.
Artificial Neural Network, a field of Artificial Intelligence, is a popular way to identify unknown and hidden patterns in data which is suitable for share market prediction. For predicting of share price using two modules, one is training session and the other is predicting price based on previously trained data. Used the Bayesian regularization backpropagation algorithm for the training sessions and a specialized multilayer feedforward network known as Nonlinear Autoregressive with External (Exogenus) Input network architecture model for stock price prediction. This study has found better methodologies which can predict share market price using systematic feature selection process, regularized backpropagation algorithm, specialized multilayer feedforward network and selected few technical indicators as input features.
This prototype stock picking model for Colombo Stock Exchange initially exploits only raw stock market data as input. The input features are initially further filtered using feature selection and later on increased to gain a better price prediction. The prototype model is based on function approximation and prediction. Having tested the model on the Colombo Stock Exchange daily stock dataset, where the prediction of the values/profitable stocks on the basis of values from the past days is made. This study has achieved the best case accuracy of over 80% at least on the dataset.
This study hopes to make further progress by introducing more input parameters (Technical Indicators and other information sources). The parameter tweaking, altered architectural models or introducing separate models to predict profitable stocks trained for different sectors in the Colombo Stock Exchange and that will provide more realistic results.|
|Appears in Collections:||Master of Computer Science - 2017|
|CSE stock Selection according to the future profit gain using pattern recognition- 13440021.pdf
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.