Discover Frequent Patterns from Academic Data Of Student Information System

Numerous researchers have explored the realm of data mining in education. The primary goal is knowledge discovery, aiming to support staff in efficiently managing educational units, refining student activities


Introduction
Many researchers have conducted studies on data mining in education, focusing on the data stored in databases of educational software and Learning Management Systems.The primary objective is knowledge discovery to assist staff in effectively managing educational units [1], redesigning student activities, and ultimately enhancing learning outcomes.A fundamental data mining technique involves uncovering hidden associations within the data stored in educational software databases.In this paper, we employ association rules mining, implementing the Apriori algorithm [2] to mine academic datasets from the student information system of Sebha University, Libya.Real data is collected from the cloud server.The algorithm is applied to extract relationships among 11 attributes of students' academic records over four years.The resulting patterns are experimentally evaluated based on support and confidence values.Specific rules are then mapped to four classes and analyzed for further verification.The proposed study yields valuable patterns related to students' academic progress and preserves essential knowledge for predicting JOPAS Vol.23 No. 1 2024 56 course add and drop decisions [3].It is one of the essential tasks in data mining to discover frequent patterns from previous datasets [4].Patterns represent a frequent occurrence of serial events in a database, constituting a common form of data that may contain crucial knowledge to be uncovered.Finding frequent patterns can provide interesting insights to experts in various domains, especially in areas such as alarm log analysis, financial events, and stock trend relationship analysis.Education is a fundamental element of a country's progress, and data mining proves to be highly useful in this context.Data mining, in this sense, refers to the method of uncovering hidden, previously unknown, and potentially significant tasks within a vast amount of data [5].Educational institutions can utilize data mining to extract valuable information from their databases, a field known as Educational Data Mining (EDM) [6].EDM requires the application of innovative approaches derived from statistics, automated learning, psychological measurement, scientific computing, etc., to better understand students.Educational Data Mining offers a range of technologies to support the education system.The methodology followed in Educational Data Mining mirrors the traditional methods of data mining, involving understanding the environment to be handled, compiling, cleaning, and arranging data, selecting applicable techniques, interpreting results, and verifying the validity of applied technologies.EDM is employed to extract knowledge patterns from the academic data in the student database of Sebha University (SUL) and to determine optimal solutions.This information may assist in predicting the courses students should add or drop each semester and recommend suitable courses for new students.The Student Information System [7] is one of the recently implemented electronic systems at SUL, assisting students and teaching staff in managing the educational process.It encompasses tasks such as student registration, course addition, examinations, attendance, absences, grade monitoring, and graduation.To address this, some colleges have academic assistants with extensive experience guiding students on the right path and suitable course selection based on their college records.However, the electronic system presents challenges as students require immediate assistance without seeking an academic advisor.This paper explores the possibility of studying and analyzing previous student records (specifically, records of adding and dropping courses over several years) to identify problems faced by former students and patterns that can help prevent these problems in the future.The study focuses on mining frequent patterns in academic student data (ASD) collected from Sebha University's cloud server (SUCS).The primary goal is to identify frequent patterns in ASD, categorized into four main events: Fail (F), Pass (P), Good (G), and Very Good (V).We apply a frequent algorithm to find the most common patterns during different time periods.This paper is organized as follows: Section 2 presents related work on association rule mining, Section 3 explains the proposed algorithm, and Sections 4 and 5 cover the experimental results and conclusions, respectively.

Related work
Researchers have conducted extensive studies on observing the performance of university students since the end of the last century.Many studies have focused on applying machine learning algorithms to discover knowledge from university databases, as outlined below: In [8], a data mining technique in higher education was implemented to examine various faculty factors affecting the quality of productive students in the global marketplace.The findings highlighted the crucial role of quality staff in promoting higher education, emphasizing the need for efforts to improve student performance in lower classrooms.[9] introduces a model for modeling and predicting academic performance based on academic records and forum participation.Three data mining techniques (Naïve Bayes, Neural Network, and Decision Tree) were employed, with Naïve Bayes outperforming others, achieving an overall prediction accuracy of 86%.This study aims to assist teachers in enhancing academic performance.Similarly, [10] introduces a new student performance prediction model with additional features, incorporating various aspects of student records and demographic background.Rule-Based proved to be the most effective among other techniques, achieving an accuracy of up to 71.3%.The model successfully predicted the success level of first-year students.In addition, [7] presents another data mining technique aiming to discover patterns through the use of algorithms on educational data.The study focuses on enhancing decision-making by identifying atrisk students, reducing student repetition rates, increasing success rates, and improving learning outcomes.The proposed method successfully detected 25 different types of survey data, revealing highfrequency patterns that could improve overall student performance.This paper [8] introduces a framework for predicting learner performance in higher education institutions by employing suitable data mining techniques.The data utilized is primarily collected from the Learning Management System (LMS) and other systems within the organization.The aim of this framework is to assist institutions in making informed decisions regarding learners' performance.Various clustering algorithms are proposed for application to the data collected from diverse activities within the LMS.This research [11] investigates the application of association rules after collecting student data over the years from the Learning Management System (LMS).Interestingness metrics and other relevant student preferences are captured, considering several association rule algorithms.Visual presentations of rules and their corresponding results will be demonstrated in terms of performance metrics, highlighting their suitability in eLearning environments.Further, [10] utilizes the Apriori and K-means algorithms in educational databases to extract and define students' academic performance based on various criteria.This study aids in predicting academic trends and patterns, allowing for the classification of students into good, medium, or weak groups and helping lecturers identify hidden patterns in student learning.[2] implements the Apriori algorithm to discover meaningful patterns in the relationship between mathematics scores and problem-solving patterns of students.The study analyzes the correlation of solution patterns to specific problems through association rule search.Another association algorithm is introduced in [12], proposing the FP-Growth algorithm to find association rules between placement test questions and grades in a programming fundamental course.The study analyzes common knowledge or skills of students who achieve good grades, suggesting essential skills for computer science students.This study [13] aims to explore various attributes related to academic students' achievements.The subjects of the study were 3,509 graduated students from the Applied Science Private University in Jordan, encompassing all department programs of the Faculty of Information Technology.The objective is to analyze students' profiles and identify their relation to and association with their university GPA (Grade Point Average) by selecting four main attributes: department, secondary certificate grade, secondary certificate city, and gender.The study then seeks to determine the highest impacting value for each JOPAS Vol.23 No. 1 2024 57 attribute and generate main association rules using data mining classifiers such as Decision Tree, Naïve Bayes, and association rules.Ultimately, the results of this study may assist current and incoming faculty students in predicting their potential GPA, enabling them to monitor and enhance their performance.This paper [14] focuses on association rule mining technique to elicit pragmatic information from the primary data collected.The work carried here explores the relationship between student's intellectual performance and their psychological factors with the use of Apriori algorithm.It aims at predicting and improving academic performance of students by reshaping their psychological parameters.Given the success of association rules mining in education data, particularly the Apriori algorithm, this study gives attention to applying it to find patterns that could help students at Sebha University overcome the problem of choosing optimal courses to start their academic journey.

Preprocessing Data
The data were collected from the ASD of Sebha University, College of Science, over a period of up to four academic years, focusing solely on result data.The following table lists the attributes extracted from the student data.

Proposed Algorithm
This algorithm is utilized to examine the relationship between data from the Academic Student Data (ASD) of Sebha University Cloud Server (SUCS).The algorithm presented in [2] is employed to discover association rules between attributes.The proposed algorithm takes as input the attributes representing the ASD.The output of the algorithm consists of the most frequent patterns that meet the minimum support and minimum confidence thresholds predefined by the user.The Apriori algorithm is designated as Algorithm 1 and is outlined below.The process for discovering frequent patterns is summarized in the following steps: This section describes the experimental results from the Apriori algorithm and processed ASD.The Apriori algorithm is a machinelearning algorithm.Table 3 describe the number of attributes that involved in the experiment.3 shows an example of generated rules (fixed at 300) for 37135 students records .The confidence is also fixed (at 0.7) and the support is changed nine times from 0.1 as min-sup to 0.9 as max.sup.The above table shows that for all the attributes (student attributes)

Fig. 1 :
Fig. 1: Education data mining process In the education system Fig 1 despite the existence of this electronic system, students, especially newcomers, still face challenges related to adding and dropping courses and choosing appropriate courses at the start of university life.With the College Hierarchy Law preventing students from studying a course based on a previous one, students need guidance to navigate these decisions without issues affecting their morale and academic performance.To address this, some colleges have academic assistants with extensive experience guiding students on the right path and suitable course selection based on their college records.However, the electronic system presents challenges as students require immediate assistance without seeking an academic advisor.This paper explores the possibility of studying and analyzing previous student records (specifically, records of adding and dropping courses over several years) to identify problems faced by former students and patterns that can help prevent these problems in the future.The study focuses on mining frequent patterns in academic student data (ASD) collected from Sebha University's cloud server (SUCS).The primary goal is to identify frequent patterns in ASD, categorized into four main events: Fail (F), Pass (P), Good (G), and Very Good (V).We apply a frequent algorithm to find the most common patterns during different time periods.This paper is organized as follows:

processed ASD data with class (students records) Min.con Min.sup Output: Association rules
Input :