Far East Journal of Theoretical Statistics
Volume 46, Issue 1, Pages 1 - 39
(January 2014)
|
|
UNSUPERVISED DATA MINING TO PROPOSE STOCKS PORTFOLIOS IN EGYPTIAN EXCHANGE
Medhat Mohamed Ahmed Abdelaal, Moustafa Galal Moustafa and Mona Mahmoud Mohamed Ebada
|
Abstract: Trading through stock exchanges is considered one of the widespread channels of investment due to its highly predictable return nevertheless it similarly includes high risks. Therefore, many studies suggest various means to help investors to take decision. At this point, the portfolio composition is one of the most important means to reduce risk of securities investment where investor’s assets are distributed to a group of stocks for diversification purposes. Any loss resulted from a drop of a certain stock can be minimized by the historical behavior of other stocks. There are many tools that can be relied on to determine performance of stocks and its relationships to each other. One of these tools is called “Data Mining” which proves efficiency and immense advantage in application to the financial sector. This study aims at proposing some alternatives for securities portfolio. The techniques of “Data Mining” used herein are “Unsupervised” method, “Hierarchical Cluster Analysis”, and “Kohonen” networks which will be helpful to classify the Egyptian Exchange market and define relationships between different market sectors through “Association Rules Mining”. |
Keywords and phrases: unsupervised data mining, hierarchical clustering, Kohonen maps, association rules mining. |
|
Number of Downloads: 287 | Number of Views: 692 |
|