Agent-based extraction algorithm for computational problem solving

Modeling Computational Problem Solving (CPS) is the main issue in teaching programming where the given text is needed to be transformed into a model then later into a programming language. Various CPS techniques have been proposed such as PAC, IPO, Flowchart and Algorithm. Also, various models have been proposed for these techniques either using agent or non-agent based solutions, but the literature shows the existing models and techniques still have limitation for novice programmers as they need prior knowledge on programming language before being able to do CPS. On the other hand, agent-based model offers a good method to solve complex computational problems. Therefore, utilizing the agent-based featured agents for this purpose is a benefit with the aim of helping novice programmers to understand the CPS without knowing the programming language. Therefore, this research aims at developing a tool for this purpose which can be classified into two objectives. The first is to propose an agent based model for CPS and the second is to evaluate the effectiveness of the prototype based on three factors consisting of understanding,efficiency and usability. Four agents have been proposed as an agent based model for CPS, which are User_Agent, PAC_Agent, IPO_Agent and Algorithm_Agent. User_Agent is responsible for receiving the problem from user and sending it to PAC_ Agent. Moreover, the extraction algorithm is located in User_Agent to perform extraction of appropriate information needed from the text and later send this information to PAC_Agent for analysis modeling and displaying input, process and output. Additionally, IPO_Agent not only produces the same PAC’s output results, but also generates module number and represents these results in another window. Finally,Algorithm_Agent employs the extracted information provided by IPO_Agent to produce the pseudo-code of the given problem and shows it in separate window. The proposed agent-based model for CPS has been designed using Prometheus Design Tools (PDT), which can be plugged-in with Eclipse Environment. The agent-based model for CPS has been developed in JADE environment which applied the basic Believe, Desire and Intention (BDI) architecture to support multi-agent system environment. The performance of the tool has been tested using 20 data sets of scenario CPS, created by the researcher. The tool is used to evaluate the accuracy of the proposed extraction algorithm to extract the appropriate information which are needed. The results show that the extraction algorithm has been able to extract 100 % of the information correctly. How far the proposed agent-based model for CPS is able to help novice programmer is evaluated by conducting a group based experiment with 35 students from Faculty of Computer Science and Information Technology (FSKTM). Several statistical tests such as normality test and reliability analysis were conducted. The basic statistical frequency method was used to analyse the performance result. The results indicate that the students’ understanding of the CPS techniques are improved by using the proposed tool. From the result, the proposed tool has obtained rank 4.17 for “understanding”, 4.01 for “efficiency” and 4.07 for “usability”. With such results, it can be concluded that our proposed agent-based tool is able to help novice programmer in CPS