Intelligent Agents as Data Mining Techniques Used in Academic Environment Irina Tudor 1, Liviu Ionita 2 (1) Petroleum-Gas University of Ploiesti, Department of Informatics, Bd. Bucuresti, No. 39, 100680, ROMANIA E-mail: tirinelle@yahoo.com (2) Petroleum-Gas University of Ploiesti, Department of Informatics, Bd. Bucuresti, No. 39, 100680, ROMANIA E-mail: iliviu@upg-ploiesti.ro Abstract A knowledge-based society determines organizations to focus their activities on improving management quality by using knowledge. Huge data stores become important once the real significance of data is discovered. Data mining techniques are involved in different knowledge processes, as one can notice in various public applications of the researchers. Managers can use these techniques in order to extract patterns, relations, associations from data initially considered of little value. Nowadays, intelligent agents represent an important opportunity to optimize knowledge management. Agents and data mining can work together in various domains such as finance, assurance, medicine, engineering and education. In this paper the authors considered an example of "data mining agents", outlining their major involvement in the complex process of knowledge management in academic environment. Keywords: Intelligent Agents, Data Mining, Knowledge Discovery, Academic Environment 1 Introduction A society based on knowledge determines managers to develop better methods and techniques to organize their data, as these become increasingly significant. In a competitive world, modern organizations focus on locating, storing, transferring and efficiently using their own information in order to better manage their intellectual capital. Concepts of knowledge management, decision support, data mining are well-known in different areas such as business, engineering, communications, transport, medicine, education etc. Data can be transformed into usable knowledge as part of knowledge management initiatives using data mining techniques to increase organizations assets. To manage knowledge is not an easy task. Data from various activities fields are produced and stored daily, processed, transmitted in different locations without taking into account their meanings. Managers focus their activity mainly on finding methods and techniques to organize huge data provided by transactions or other activities and to extract useful patterns, relations, associations from data etc. Data mining task is to
The 4 th International Conference on Virtual Learning ICVL 2009 381 translate structured data into knowledge. In recent years, organizations have attempted to transform raw data into usable knowledge as part of their knowledge management initiatives. In different applications, it is necessary to know what to do, when and how to do it, in order to complete the pre-established tasks for the proposed objectives, by means of selfdecision systems. These systems are known in literature as agents. Intelligent agents act robustly in a flexible, open environment. The recognized domains of intelligent-agents applications are education, communication, engineering, business, e-commerce, assurance, telecommunication etc. Knowledge discovery process can be assisted by agents in order to increase the quality of knowledge and to simplify the main processes of identifying patterns from huge data volumes. Intelligent agents and data mining share the same objectives in order to assist decision making process. A data mining agent is a software program built for the primary purpose of efficiently finding information that operates in a data store. This type of agent is able to detect both major trend changes and new information. In the current paper we discuss data mining agents that make a significant contribution to a knowledge management effort in the education field. Authors illustrate how agents such as DM techniques can be used for building educational knowledge, which would lead to a better performance in the academic environment. 2 Intelligent Agents and Data Mining Agents, i.e. special types of software applications, have become increasingly popular in computing world in recent years. Some of the reasons for this popularity are their flexibility, modularity and general applicability to a wide range of problems (data filtering and analysis, information brokering, condition monitoring and alarm generation, workflow management, personal assistance, simulation and gaming). Because of the explosive development of information source available on the Internet and on the business, government, and scientific databases, it has become quite necessary for the users to utilize automated and intelligent tools to extract knowledge from them [Seydim, 1999]. Intelligent agents can help making computer systems easier to use, enable finding and filtering information, customizing views of information and automating work. An Intelligent agent is software that assists people and acts on their behalf. Intelligent agents work by allowing people to delegate work that they could have done to the agent software [Gilbert, 1997]. On the other hand, data mining is the process of posing queries and extracting useful information, patterns and trends previously unknown from large quantities of data [Thuraisingam, 2000]. Data mining is also a multidisciplinary field, working in areas that include artificial intelligence, machine learning, neural networks, pattern recognition, knowledge-based systems, information retrieval, high performance computing, and data visualization [Han and Kamber, 2001]. The concept of knowledge is very important in data mining. In order to get the correct knowledge from the data mining system, the user must define the objective and specify the algorithms and its parameters exactly with minimum effort. If the data mining system produces large number of meaningful information by using a specialized data mining algorithm (association, clustering, decision trees etc.), it will take more time for the end-
382 University of Bucharest and Gh. Asachi Tehnical University of Iasi users to choose the appropriate knowledge for the problem discussed. Even choosing the correct data mining algorithm involves more time for the system. A solution for this problem could be an intelligent system based on agents. Data mining and intelligent agents can make a common front to help people in the decision making process, to elaborate decisional models and take good decision in real time. Data mining is a difficult and laborious activity that requires a great deal of expertise for obtaining high quality results. New methods are necessary for intelligent data analysis to extract relevant information with minimum effort. With the use of the autonomous intelligent agents several data mining steps are possibly be automated [Rajan and Saravanan, 2008]. 3 Agents in Academic Environment Intelligent agents can successfully perform complex tasks within the educational process. An example of intelligent agents in academic environment is given by euniv project [Oprean et al, 2002], consisting of five categories of applications: educational (courses, seminars, practical activities, lectures, assessment sessions, graduation and admission, curricula, text-books, e-learning); research (projects: national grants, international grants, co-operation, internal; reports; scientific papers, books; events); administrative; secretarial; others. For this project the authors [Oprean et al, 2002] used the architecture presented in the figure below. A casual scenario is the following: when a department prepares the new academic year structure, a software agent presents a snapshot of the situation providing the information needed in the educational process for a given situation. Users can obtain information about all the professors and assistants specialized in a certain course, the software available in the department, the configuration of the networks in laboratories and the necessities for students practical works, how many workstations are needed taking into account the number of hours/student and the number of students in the last educational year. The system [Oprean et al, 2002] offers alternatives Figure 1. E-Univ architecture [Oprean et for the location of courses and practical works, if the timetable is implemented. An intelligent agent checks the pre-requisite for al, 2002] attending the course. All these operations being automatic provide a valuable support to the department staff.
The 4 th International Conference on Virtual Learning ICVL 2009 383 On the other hand, timetable planning can be a significant task for agents. The constraints are related to the availability, timetabling and preferences of each professor, to rooms availability, number of students, and curricula. In order to solve this problem for the particular case of university course timetable scheduling, an agent-based approach is a viable solution. The designed multi-agent system, MAS_UPUCT, has as main purpose the modelling of the university courses timetable scheduling [Oprea, 2006]. The authors propose a multi-agent system designed to offer advisory support for candidates in order to enrol them at postgraduate school courses from a certain faculty. The application works online and an online registration request form is necessary to help the advisor in accepting or declining candidate registration request. A registration form contains personal information and education history (e.g. graduated domain) for each candidate. When the candidate fills the registration form, there must be mentioned the postgraduate school(s) he/she wants to attend. An advisor agent assigned to the respective candidate may accept or decline his/her online registration request on the basis of candidate s past study domain. In case of acceptance, the candidate can be enrolled, following the university methodology for postgraduate school examination approved by the university senate. In case of the request being declined by the advisor, the candidate repeats the process of filling the form manually and submitting the online registration request after rectifying the previous errors. If the candidate request is rejected, he/she may contact the board of examiners to obtain a supplementary advisor support. The multi-agent system contains both mobile agents (Candidate Agent (CAgent), Advisor Agent (AAgent)) and stationary agents (University Agent (UAgent)). CAgent is a personalized mobile agent and it is created when a candidate initiates a registration request. It sends the request to the advisor and back to the candidate, after having been accepted or declined. To each candidate there is assigned a personalized Advisor Agent. AAgent is an intelligent mobile agent that performs two tasks: collects the academic and financial information and provides advice, once it has an intelligent analysis on the collected data and the request based on the specified registration rules. The purpose of University Agent is to provide information on the academic history of the candidate to AAgent (in case the candidate graduated from the same university as the one he/she wants to enrol on for postgraduate school), as well as to inform the candidate about the registration s confirmation through e-mail. The UAgent task consists in book keeping the taxes paid by each candidate. The multi-agent system works as follows (fig.2): first of all, the candidate initiates the request by selecting postgraduate schools out of the offered list in the registration form. To initiate the request, a personalized mobile CAgent is invoked that takes the request to the advisor and waits until acceptance/ rejection is provided. When receiving a request, a personalized AAgent is activated. After getting information about the candidate, AAgent returns and performs the critical task of an intelligent advisor Advisory process works as follows: On the basis of the collected information on the candidate (the domain of study attended before, his/her final examination marks etc.), he/she is being evaluated according to pre-established rules. Rules checking is performed by an inference engine of the AAgent. In simple cases, acceptance/rejection is provided on the basis of basic or inferred rules. Having received the response to the request from
384 University of Bucharest and Gh. Asachi Tehnical University of Iasi the AAgent, the proposed multi-agent informs the candidate about the acceptance or rejection of his/her registration request form. The candidate comes in front of the examiners board and follows the steps stipulated in the postgraduate school methodology examination. Finally, the UAgent generates an e-mail confirming tax payment by the candidate. For a robust behaviour of the system, there must be a good coordination of the agents. Communication between agents of the proposed system is realized by means of interaction strategies in which there are specified the conditions to which agents may pass when receiving messages that contain certain information. Among the advantages of using intelligent agents, one may mention higher work efficiency, meaning that the user saves time, as agents work autonomously and more effectively, as they can search and filter huge amount of information, which would be impossible for humans. This opens new approaches for researchers in combining data Figure 2. The multi-agent system components mining with intelligent agents. This paper proposed a multiagent workflow-based system for postgraduate school registration in order to automate this complex process. The proposed system is characterized by the advantages of autonomy, mobility and collaboration of different software agents in order to provide simple and fast registration workflow process for a candidate. Using agents as data mining techniques to reduce enrolling time in the described process is a new approach within artificial intelligence field. The proposed system is in the design phase and the presented theories will be tested by authors in their future research work. REFERENCES Gilbert, D. (1997): Intelligent Agents: The Right Information at the Right Time, IBM Corporation, Research Triangle Park, NC USA. Han, J., Kamber, M. (2001): Data Mining: Concepts and techniques, Morgan Kaufmann Publishers, 5-7. Oprea, M. (2006): Multi-Agent System for University Course Timetable Scheduling, The 1st International Conference on Virtual Learning, ICVL 2006, Bucuresti, 231-238. Oprean, C., Moisil, I., Candea, C. (2002): euniv: an e-business solution for a university academic environment. In Proceedings of 3rd Global Congress on Engineering Education, Glasgow, Scotland, United Kingdom, 363-366. Rajan, J., Saravanan, V. (2008): A Framework of an Automated Data Mining System Using Autonomous Intelligent Agents, International Conference on Computer Science and Information Technology, 700-704. Seydim, A.Y. (1999): 'Intelligent Agents: A Data Mining Perspective, Dept.of Computer Science and Engineering, Southern Methodist University, Dallas, TX 75275. Thuraisingam, B. (2000): Data Mining: Technologies, Techniques, Tools, and Trends, CRC Press, 4-6.