1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Hybridization of simulated annealing with principal axis tree and k-nearest neighbor for attribute reduction on financial time series forecasting.
Authors:	Hani A.K. Ihlayyel (P59475)
Supervisor:	Salwani Abduallah, Prof. Dr.
Keywords:	Hybridization Financial Forecasting Expert systems (Computer science)
Issue Date:	2-May-2013
Description:	Forecasting of the financial time series is a challenging task, due to the huge price fluctuations in the stock market. Due to the complexity, evolutionary and non-linear dynamical system nature of the financial stock market, which are characterized by the intensity of data, noises, non-stationary and unstructured nature, high degree of uncertainty and hidden relationships, the risks involved with stocks trading are immense. A method considered to solve the complex issues surrounding data is the attribute reduction, which is known to be a NP-hard optimization problem. Finding a minimal attribute from a large set of attributes to identify the right indicators is always a challenging problem. This study aims to investigate the effectiveness of the proposed algorithm for attribute reduction in order to identify the most relevant technical indicators for each stock market, and also aims to enhance the forecasting accuracy for the direction of daily closed price. Over the past years, researchers have proposed many meta-heuristic algorithms to find the optimal solution for attribute reduction such as genetic algorithm (GA). In this research, Simulated Annealing (SA) with Principal Axis Search Tree and K-Nearest Neighbor (PAT-KNN) is proposed to solve the issue. This study is divided into two phases; in the first phase SA is used for finding the minimal number of attributes and in the second phase PAT-KNN is used to evaluate the selected attributes. The proposed method has been tested on four stock markets, which are VALE5, KRAFT, PETR4 and KOSPI, obtained from Yahoo financial website (http://finance.yahoo.com). The experimental result shows that the proposed method is able to obtain competitive result compared to the other available results in the literature. The proposed method has successfully reduced the number of technical indicators and enhanced the forecasting accuracy in two experiments out of four namely PETR4 and KOSPI datasets were the number of technical indicators reduced to 4 in both datasets from 5 and 6 respectively and the forecasting accuracy enhanced from 57.50% and 60.05% to 58.21% and 61.05% respectively, and has produced same accuracy for the VALE5 and KRAFT.,Master/Sarjana
Pages:	97
Call Number:	QA76.76.E95.I376 2013 3
Publisher:	UKM, Bangi
URI:	https://ptsldigital.ukm.my/jspui/handle/123456789/476162
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.