TRAVELLING SALESMAN FOR FINANCIAL PREDICTION

Department of Computer Engineering Faculty of Engineering, Cairo University , Giza , EGYPT

This work deals with the problem of forecasting a complex environment behavior. The word

“complex behavior” is here used in the sense that the environment may be modeled as an

almost random walk environment. Stock market prediction is known to be such an

environment and is the case study of this work. The ultimate goal for human investors is to

build an intelligent tool for stock market prediction. The heart of any such tool is the

intelligent technique used for earning forecast. This paper presents a swarm based tool

tailored for stock market prediction. Stock market prediction is first formulated as a traveling

salesman problem that is solved using swarm optimization. The solution is used as the basic

indicators for prediction. Although empirical results are comparable to conventional statistical

and evolutionary based techniques, the technique is superior in terms of execution efficiency

and speed. In addition, the technique may be used in case of missing data and may be based

on simple raw data rather than more sophisticated financial indicators that may not be

available to the average novel investor.