ABSTRACT

ln our count r) mn nua l work is be ing con verted lo Computer based applicat ions for efficiency as \\ el l as for better servi ng the int ended users. 'l hrough web tech no logy, the.c applicat ions arc being remotely accessed by their users across geographica l boundaries. At the Karachi Stock Exchange large business transact ions are carr ied-out da il) and the brokers arc manually involved to provide facility of pu rchasing and selling of 1he u nits of shares of certain companies and the) arc the key persons to keep their eyes on t he market and provide consu ltanc} to customers for the best time to purchase and sell the shares. This business is growing a ll over the world so there 1s a need of an intelligent system which might be used d irect ly by a common customer without involving the broker. The system pred icts future trend of stock market shares on da ily basis so that a customer may decide whether to purchase or sell the shares. This thesis presents the evaluation of Karachi stock exchange pred ict ion model \\ here Dyna mic Bayesian Network (DBN) technique is used to implement the system to provide the plat form for pred ict ing shares. I n this system the data are provided b) Karach i stock exchange are used and the results shows that the proposed system gets almost 93% of hit rate for pred icted results when compared with actual results u sing sample of 10 days of network, whi le usi ng data of month JUN E 201 1 hit rate for
pred icted results 74%, JULY 20 1 1 hit rate of 68.3%, AUGUST 20 1 1 hit rate of 62% and SEPTEMBER 201 1 hit rate of 66. 15% and final lly t he mean of four consecut ive months hit rate as accuracy of the system 68%. The research has augmented more interest to work on yearly data for making t he system more intelligent.