A Multi-agent System for Data Mining

Abstract

ln this section, we will focus on the implementation steps of our case study. For example, we will describe the details of how to use DBMiner,·Agent-tool and others to build a data mining system. We will base our work on the models, which we built in section 5.8.

Our case study will focus on evaluation process for post graduation studies and training scholarships. Students apply in IT! (Information Technology Institute) to attend 9-month program. Then they can complete their studies by applying in master program. The evaluation process consists of three phases. In the first phase, the student must pass three writing exams, one interview and a toefl exam. If the student passes all exams then he/she will attend the training program. During the training program, he/she attends courses and exams for each course. He/she can complete master studies depend on their total grade in all courses. The manager wants a data mining system to support him in evaluation process in each phase. He wants to answer the following questions: (i) can he/she join the training program. (ii) Will he/she success in the program. In other word, will he/she achieve the required total grade to join the master? We will design data mining system to help the decision maker.

2- Data preparation

We got the data of some students whom applied on 9-month program and their evaluation and grades. In addition, whether they accepted in the diploma or not? These records are stored in excel files. The first step to analyze these records, we must convert these files to records in database. We used MS SQL SERVER 2000 to import these records. The excel file contains 501 rows. In addition, it contains 12 columns. The first three columns represent the university, faculty and total grade of the applier. Each university and each faculty has a rank to calculate the level of the appliers. Then the second three columns represent the grades of the written exam. There are three exams IQ, Aptitude, General Knowledge. The interview records the following three columns: language,