Abstract: Computer-Aided Decision Making (CADM) has been widely used in various decision contexts. This study is focused on CADM for Choosing Education Programs (CASCEP) to rank the performance of a set of decision alternative with respect to multiple criteria. CASCEP has been developed in Visual Basic (VB) that will select and rank the programs that are more suitable to the Malaysian Certificate of Education (SPM) leavers based on 2 inter-related factors: the student’s SPM results and minimum programs requirements. In this prototype, 5 programs are offered by Universiti Teknologi Malaysia (UTM) have been selected as a sample programs. Rule-based reasoning technique is used as an engine in order to search the most suitable education program. In this engine, score or value of each program is generated to rank the programs.

INTRODUCTION

Computer-Aided Decision Making (CADM) is widely used in decision making contexts include managing all the sub-decision, educating decision maker and consumer purchase in an online sales environment (Grosser and Ghaed, 2004) and also, to assist decision makers in the building industry (Papamichel et al., 1999). Decision Support System (DSS) can be used in problem solving to recommend the best alternative solution to human (Barkhi et al., 2005). DSS can be classified into soft computing techniques, knowledge engineering techniques and agent-based techniques. The first technique uses the concepts related to rough set theory, fuzzy logic, neural networks and genetic algorithms. The second technique uses knowledge-based systems and expert systems (Sueyoshi and Tadiparthic, 2008). DSS can be also integrated with agent element that can be implemented by using rule-based reasoning, knowledge based reasoning and fuzzy agent’s techniques. Venkatramana built the decision support system and hence a rule-based approach was chosen as the implementation model for technical problem in inventory issue (Venkatraman and Venkatraman, 2000).

In this study, agent acts as a decision making tool to facilitate student with a list of suitable programs that Malaysian Certificate of Education (SPM) leavers could apply for. Open Certificate System has widen the choice of educational programs that student is qualified for, hence in order to be offered a place at Malaysia Public Higher Learning Institution (IPTA), it is very important for students to make the right choice in applying the right programs. The student must determine their subject strength in making decision to avoid selecting unsuitable programs. In short, it is hard to find, which program (s) is (are) most suitable based on their result. Therefore, CASCEP has been developed to assist them in decision making process effectively. In order to improve, the effectiveness of decision making, DSS can be integrated with agent elements using rule-based reasoning, knowledge-based reasoning and fuzzy agent’s technique (Wu et al., 1992). In this study, agent acts as a decision making tool to facilitate student with a list of suitable programs that SPM leavers could apply for. The agent will match the students SPM results with programs requirements to search the most suitable education programs. Therefore, any misunderstanding about the programs requirements can be avoided. In this study, 5 programs offered by Universiti Teknologi Malaysia (UTM), are chosen as the sample education programs of the prototype system (Anonymous, 2002).

Related research: This study discussed past research and issues on decision making in education, decision support system and rule-based.

There are many factors, which one should consider before making a decision in choosing education program. Ibrahim (2002) indicated in this study, SPM leavers should make the right choice in program selection. The strength of the subject must be determined to avoid bad selection program. The student needs a counseling session in order to choose the most suitable education program. Hence, it is important to have counseling teacher to give advice and monitor the student in making decision (Yahya, 1999). Therefore, many past studies developed Computer Aided Decision Making (CADM) to aid people in decision making include managing all the sub-decision, educating decision maker and consumer purchase in an online sales environment (Grosser and Ghaed, 2004), construction (Papamichel et al., 1999) and selection problems. Song (1992) conducted a study on using computer-aided decision-making system to assists educational practitioners in making curriculum decision about innovative programs features and implementation requirements recommended by the system.

The selection of an appropriate university or college is of vital importance to the student for acquisition of proper educational experience. There are thousands of universities and programs out there. The information available to students about each program is plentiful but is rather tedious to be obtained. The decision process of college selection is further complicated by many factors such as curriculums, location, rank, size of the universities and so forth. These factors play an important role in the final selection of a college. Wu et al. (1992) developed a computer-based decision support system to help users make better decisions in the selection of a college. It can be run on any IBM XT/AT or compatible machines with a DOS environment. It will allow users to make better decisions in their college selection process. Mohamad et al. (2005) integrated Fuzzy Multiple Attribute Decision Making MADM with an expert system for selecting a university program, which involves program assessment based on multiple criteria of the user’s SPM results. The prototype focused on UiTM’s Science and Technology Cluster (college) 6 diploma programs, which are selected from 6 different faculties of each sub-group grouping. Paul et al. (2004) explained the development and implementation of a Group Decision Support System (GDSS) in selection and prioritization of the attributes of Master of Business Administration (MBA) programs. They use a collective memory concept in an iterative decision. This may help decision makers adopt less conflicting decision.

Sieker et al. (2006) defined DSS as an integrated, interactive computer system, consisting of analytical tools and information management capabilities, designed to aid decision makers in solving relatively large, unstructured problems. DSS is applied to guides and supports an improved water resources management on the level of small watersheds. DSS in oil spill management (Pourvakhshouri et al., 2006) aids the decision maker to choose the most reasonable combating method for prevention, control and/or cleanup way against the oil spills pollution. DSS is also increasingly in tourism (Laniado et al., 2004). SFIDA can be used to generate information and stimulate participation, making the decision transparent, repeatable and participated.

Sueyoshi and Tadiparthic (2008) in their study combined soft computing, knowledge-based and agent-based techniques to build an intelligent decision making tool. The proposed software uses soft-computing techniques such as probabilistic reasoning and reinforcement learning. The software uses a knowledge-base to fully utilize knowledge on a wholesale market of electricity. Each player in the wholesale market is represented by an intelligent agent. Nammuni et al. (2004) used rule-based to build DSS that assists trial designers in designing and planning clinical trials. It knowledge base included medical, statistical, ethical and trial design information, to provide guidance during the trial design process and thus help produce more rigorous protocols more rapidly and easily. Lazarov and Shoval (2002) presents a system for automatic assignment of technicians to service faults. A rule-based selection was used to refine the decision that is, to choose the most appropriate technicians from the list of relevant technicians. Dupuit et al. (2007) integrates rule-based reasoning and non-parametric measurement for the optimal sampling points and to control the wastewater quality and its progression over time in industrial wastewater network management.

Rule-based approach has been in use for a couple of decades and their usefulness has been demonstrated in many domains such as agriculture (Debaeke et al., 2006), farming (DelaOssa et al., 2007), pattern recognition (Frauel et al., 2006) and construction industry (Furusaka et al., 2000). Venkatraman and Venkatraman (2000) mentioned in their study, the use of rule-based for a typical inventory problem of steel pipes in a construction industry where the scrap has to be minimized. They describe that every knowledge system consists of 2 core components. There are knowledge representation schemes and inference strategies.

For the inventory problem, the expert knowledge is coded into a rule-based system by transforming the knowledge and constraints into a set of if-then rules. They have used the forward chaining inference strategy for implementation. The study also, brings out the benefits that this rule-based system offers to the organization such as scrap reduction and lead time reduction. DelaOssa et al. (2007) stated the capability of Rule-Based System (RBS) as predictive systems that is, the system can be used to infer the output for a target variable given an input. RBS is also described as descriptive systems that is, the rules describe interesting relations between the problem variables. Based on that capability, they developed application to a farming problem.

MATERIALS AND METHODS

The aim of this study, is to develop a prototype of Computer-Aided Decision Making in Choosing Education Programs (CASCEP) at the higher educational institutions using rule-based search agent for the SPM leavers. This study focuses on SPM leavers, who wanted to apply for the programs that are offered by IPTA.

DSS is developed using agent element to search for suitable programs to be applied by SPM leavers as their 1st step to study in IPTA. In this prototype system, it is based only on SPM result and minimum program requirements. In this study, the following 5 programs offered by Universiti Teknologi Malaysia (UTM), form the education programs of the prototype system (Anonymous, 2002).

Architecture design of agent: In this prototype, the agent is used to suggest the most suitable education programs to the user. Figure 1 shows the basic agent architecture for rule-based search agent. The rule-based engine consists of initial rule base, content rule base and final rule base. The data that has been key-in by the user will be stored in the database. The agent will then use the data to fire the rules in the rule-based engine. The output of the rule-based engine is the most suitable education programs.

Fig. 1:	Basic agent architecture for rule-based search agent

Fig. 2:	Rule base sequence

Fig. 3:	Initial rule base

Rule-based engine: The rule-based engine is divided into initial rule-based, content rule base and final rule base (Fig. 2). The engine will compare the data in the database with rules, to produce the suitable education programs.

Fig. 4:	Content rule base flow

Initial rule base: The initial rule base contains initial rules. The flow of the initial rule base start by filtering subject requirement for each program extracted from the database to produce qualified result (Fig. 3).

Content rule base: Content rule base is used to generate the total scores of the programs. The calculation of the total scores is shown.

Total 1 represents the scores for group one, where else total 2 for group 2 and total 4 for group 4. Nilai is the grade’s value.

Fig. 5:	Suggested education programs flow

The score calculation process can be illustrated in the flowchart as shown in Fig. 4. The input to the process is SPM result. Subjects that are graded ≤6C are chosen. The subjects will be compared with programs requirements from the database to produce the qualified result. The qualified results must fulfill the condition of each group in order to calculate the total score. Otherwise this calculation process will be stopped.

Final rule bas: The final rule base is used to generate the result. In this part, 5 programs are selected based on the lowest score. In the case, where the total scores for the programs are found to have the same value, then Rule 3 (priority requirement) will be applied. The calculation to generate the scores for rule 3 is Total 3 = Total 3 + nilai. For the programs that met the rule 3 requirement, it will be ranked ahead among the programs that have the same scores. Figure 5 shows the process flow to display the result.

Table 1:	Sample of SPM result
Examination syndicate ministry of Malaysian education. Malaysian certificate of education (2001), Qualify to get certificate, Examination director

Table 2:	Education programs sample

The data from temporary file is the input to the sorting process. The programs are sort based on the lowest score. Then, the best 5 programs will be chosen. However, in the condition where programs generated the same value, then rule 3 is the decision factor. Whichever, programs that meet Rule 3 requirement is chosen.

Actual implement: In this prototype, system analysis phase seeks to systematically analyze data input, data clustering and process modeling. There are 2 categories of inputs. SPM result as a user input and programs requirements. SPM result as a user input (Table 1).

Programs requirements to generate rules: Information on minimum requirement for education programs are referred to university programs handbook (Anonymous, 2002). The prototype focused on UTM 5 programs, which are selected from difference faculties (Table 2). Diploma in civil engineering is used as a sample to represent the whole process.

Raw data: Table 3 shows the raw data, which is indicated the minimum program requirements for T001. These data will be used to generate rules.

Clustering: The raw data in Table 3 is then clustered into 4 groups (Table 4).

The clustering requirement in Table 4 then could be further simplified as presented in Table 5. Referring to the group column, G1, G2 and G4 are categorized as compulsory requirement and G3 is categorized as priority requirement.

Table 3:	Minimum program requirement for T001
Programs: Diploma in Civil Engineering (T001)

Table 4:	Clustering requirement for T001

Referring to the requirement column, No. 1 or 2 refers to the number of subjects and ≤6C refers to the minimum grade scored by the mentioned subject (s). The purpose of clustering is to simplify the process of calculating score for each group (the detail process will be explained in the content rule base phase). The clustered data is then used to build rule base (Table 6).

Table 6 shows the rule base for T001. In the rule base, the set of rules is coded in IF-THEN structure which where particular rule is fired. Goal is executed if all the rules are true. The calculation algorithm is developed to find score for each program.

Table 5:	Simplified clustered requirement of T001

Table 6:	Rule base for T001 programs

RESULTS AND DISCUSSION

Initial rule bas: The output to the initial process is to identify qualified result. The sample of SPM result from Table 1 will be summarized as indicated in Table 7. It shows the result that has been used to match with programs requirements. There are only nine subjects from SPM result having a grade not >6C.

Table 8 shows the qualified subjects for each program for sample data after filtering process. Only T018 have nine subjects to fulfill the program requirement. The rest of the programs need 8 subjects.

Content rule bas: In this rule base, the total score for the programs were generated. Qualified subjects from the initial rule base then will be used in score calculation process for each subject’s group. Grade is presented as a value for each subject and it will be calculated if the subject meets the group’s requirement. Otherwise, the score for that group will be zero and the total score for the programs will also be zero. This means the student is not qualified for the particular program. Table 9 shows the calculation score for T001 is 13. The SPM result meets all of the requirements for all of the groups.

Table 10 shows the calculation number of subject score for T002 is 13. T001 and T002 are similar due to the same requirement of G1, G2 and G4. T002 does not have G3 and only T001 has this group. In this phase, group 3 is not used to generate score for program but it will be useful when more than one group has same total score.

Table 7:	Summarized SPM result for ID: 840508-05-5151
Identity card number: 840508-05-5151

Fig. 6:	GUI of main menu

Fig. 7:	GUI of student menu

Table 11 shows a total score for T004 is zero. This is because, student must have 2 qualified subjects for G1 and if not, the score for G1 will be zero. If either one of group, has zero score then the overall score will be zero.

Table 12 and 13 show that the total score for T018 and T022 are also zero. In this case, this means there is no subject that is qualified for group 1. Table 14 is summarizing score for all of the programs. It is found that only T001 and T002 have score and the others are zero.

Final rule base: The results in the initial rule based and the content rule based shows the most suitable education programs for identity card number: 840508-05-5151 T001 is suggested as a first choice because the SPM result has met priority requirement for T001 whereas T002 does not have this requirement. Both of the suggested programs are offered by UTM (Table 15).

Table 8:	Simplified qualified subjects and subject groups for all programs
Identity card number: 840508-05-5151

Table 9:	Calculation score for T001
Identity card number: 840508-03-5151; Program code: T001

Table 10:	Calculation score for T002
Identity card number: 840508-03-5151; Program code: T002

Table 11:	Calculation score for T004
Identity card number: 840508-03-5151; Program code: T004

Table 12:	Calculation score for T018
Identity card number: 840508-03-5151; Program code: T018

Table 13:	Calculation score for T022
Identity card number: 840508-03-5151; Program code: T022

Table 14:	Score for each program

Table 15:	Final suggested education programs
Identity card number: 840508-05-5151

Fig. 8:	GUI of student menu with input

Fig. 9:	GUI of CASCEP result

Figure 6-9 demostrate the Graphical User Interface (GUI) of CASCEP. Users only need to key in SPM result on student menu (Fig. 7) and CASCEP will display the output on result menu (Fig. 9).

CONCLUSION

Generally, the main purpose of the system, which is to develop a prototype of computer-aided decision making system in choosing education programs for SPM holders have been achieved. The CASCEP uses rule-based search agent to search the most suitable education programs by matching the SPM result with the programs requirements. The agent is capable of assisting the student in choosing the right education programs that is suitable with their SPM result. By using the CASCEP, the possibility of choosing the wrong education programs can be decreased.

The sample of the education programs that are suggested by the agent in the prototype are only five education programs. By including more education programs from different higher education institutions, the agent could make comparison and would be able to suggest more education programs to the user. It is also hoped that this system can be enhance to Web based application programs (Noor et al., 2006; Wena et al., 2008; Zhang and Goddard, 2007). In such a case, every user will be able to access the system anywhere and use it as a mechanism in helping them to make the right decision. The prototype employed the rule-based technique and search agent. In future, the study could be carried out using other techniques such as fuzzy neural networks (Kuo and Chen, 2004), genetic algorithm (Lee, 2008) and additive value function (Noor et al., 2006).