Back to Home

COURSE DETAILS

16:332:568
Software Engineering of Web Applications
Spring 2008
Electrical and Computer Engineering Department
Rutgers University, NJ

INSTRUCTOR

Professor Ivan Marsic

PROJECT STATEMENT

There are many tools available to investors but none of them removes entirely the element of chance from investment decisions. Large trading organizations can employ sophisticated computer systems and armies of analysts. Our goal is to help the individual investor make better investment decisions. Our system will use the Delphi method,8 which is a systematic interactive forecasting method for obtaining forecasts from a panel of independent experts.

The goal of this project is to have multiple student teams implement Web services for stock-prediction. Each Web service (WS) will track different stocks and, when queried, issue a forecast about the price movement for a given stock. The client module acts as a “facilitator” which gathers information from multiple Web services (“independent experts”) and combines their answers into a single recommendation. If different Web services offer conflicting answers, the client may repeat the process of querying and combining the answers until it converges towards the “correct” answer.

Here are some of our project goals:

Potential services that will be provided by the forecaster service include:

A key step in specifying the forecaster service is to determine its Web service interface: what will go in and what will come out of your planned Web service? Below I list all the possible parameters that I could think of, which the client and the service could exchange. The development team should use their judgment to decide what is reasonable and realistic for their own team to achieve within the course of an academic semester, and select only some of these parameters for their Web service interface.

Parameters sent by the facilitator to a forecaster (from the client to a Web service) in the inquiry include:

Some of the above may not be necessary, particularly in the first instantiation of the system. Also, there are privacy issues, particularly with the last two items above, that must be taken into account. The forecaster Web-services are run by third parties and the trader may not wish to disclose such information to third parties.

Results returned by a forecaster to the facilitator (for a single stock per inquiry):

The performance of each prediction service should be evaluated as follows. Once activated, each predicted price value should be stored in a local database. At a future time when the actual value becomes known, it should be recorded along with the previously predicted value. A large number of samples should be collected, say over the period of tens of days. We use absolute mean error and average relative error as indices for performance evaluation.