Source-linked AI summary
Data Mining: A prediction for performance improvement using classification
Brijesh Kumar Bhardwaj, Saurabh Pal
TL;DR
The paper addresses the need to predict higher-education performance from multiple influencing factors so low achievers can be identified for support. It constructs a classification-based prediction model using prepared student data and reports several high-potential performance variables, including prior grades, living location, and teaching medium.
Problem
Predicting higher-education performance matters because student outcomes are influenced by multiple personal, social, psychological, and environmental factors, while low achievers need timely identification and assistance.
Method
The study constructs a Bayesian classification prediction model using selected influential variables from a student database.
Results
The analysis identifies senior-secondary examination grade, living location, teaching medium, mother’s qualification, other habits, family income, and family status as high-potential variables affecting performance.
Takeaways & Limitations
The model is intended to help teachers identify students needing special attention and take timely action to improve performance and reduce failing.
Abstract
from arXiv · showhide
Now-a-days the amount of data stored in educational database increasing rapidly. These databases contain hidden information for improvement of students' performance. The performance in higher education in India is a turning point in the academics for all students. This academic performance is influenced by many factors, therefore it is essential to develop predictive data mining model for students' performance so as to identify the difference between high learners and slow learners student. In the present investigation, an experimental methodology was adopted to generate a database. The raw data was preprocessed in terms of filling up missing values, transforming values in one form into another and relevant attribute/ variable selection. As a result, we had 300 student records, which were used for by Byes classification prediction model construction. Keywords- Data Mining, Educational Data Mining, Predictive Model, Classification.
I. INTRODUCTION
The introduction motivates predicting higher-education students’ performance from diverse personal, social, psychological, and environmental factors. It proposes a classification-based predictive model to identify low achievers and support timely assistance.
- Student performance prediction is important because academic outcomes depend on diverse personal, social, psychological, and environmental factors.
- Data mining can uncover hidden patterns and relationships in large educational datasets for decision making.
- High-accuracy prediction can help identify low-achieving students early so teachers can provide additional assistance.
- The investigation aims to generate predictive variables, identify factors affecting performance, construct a classification prediction model, and validate it for Indian higher-education students.
- The proposed work focuses on students’ learning behavior and academic performance during their academic careers.
II. BACKGROUND AND RELATED WORK
The background presents Educational Data Mining as a way to study student learning and performance, alongside prior studies using tutoring, demographic, regression, association-rule, decision-tree, and Bayesian approaches.
- Educational Data Mining applies data mining to understand learning processes by identifying, extracting, and evaluating student-related variables.
- Prior Indian research used Bayes classification to examine whether incoming students would perform, using category, language, and background qualification.
- A Pakistan study found mother’s education and family income highly correlated with academic performance among 300 college students.
- Research has examined learning behavior, at-risk warnings, cognition, personality, demographic variables, and private tutoring in relation to student outcomes.
- A Jordanian comparison of ID3, C4.5, and NaïveBayes reported better prediction from a decision-tree model.
- Association rules have also been used to study students’ interest in choosing classroom teaching language.
III. DATA MINING PROCESS
The study analyzes data from degree colleges and affiliated institutions using classification to predict student performance through a sequence of data-mining steps.
- Data were gathered from degree colleges and institutions affiliated with Dr. R. M. L. Awadh University in Faizabad, India.
- The collected data were analyzed with a classification method to predict students’ performance.
A. Data Preparations
The data-preparation process assembled a 300-record dataset from BCA students and consolidated information from multiple tables after removing errors.
- The dataset came from BCA students in different colleges during the 2009–10 academic session.
- Data stored in different tables were joined into one table, and errors were removed after the joining process.
B. Data selection and transformation
The study selected relevant predictor and response variables and transformed several domain-specific attributes into defined values or classes. The resulting variables are documented in Table 1.
- Relevant predictor and response variables were selected for the data-mining analysis, with some variables derived from database information.Table 1 provides the resulting student-related variables.
- Student category values were defined as General, OBC, SC, and ST.
- Instruction medium was represented as Hindi, English, or a mixture of both in Uttar Pradesh degree institutions.
- Students’ other habits were categorized as Drinking, Smoking, Both, or Not-applicable.
- Family size was constrained to integer values from one to ten.The paper fixed ten as the maximum family size.
- Senior-secondary grades were mapped from O for 90–100% through F for below 40%.
- BCA marks or grades were used as the response variable and divided into performance classes.The listed classes are First, Second, Third, and Fail, with percentage thresholds specified in the paper.
C. Implementation of Mining Model
The study implemented Bayesian classification on student data, using naïve Bayes probability estimates and filtered feature selection to construct a performance-prediction model. The sample comprised 300 BCA students from five affiliated colleges.
- The study used Bayesian classification to predict a categorical student-performance outcome from other attributes.
- 300 BCA students from five affiliated colleges formed the study sample.The sample included 226 males and 74 females who appeared in the 2010 examination.
- Naïve Bayes estimates class priors and attribute probabilities from the training data by counting occurrences.
- For a target tuple, the algorithm uses prior and conditional probabilities generated from training data to make a class prediction.
- The class with the highest product of conditional attribute probabilities is selected for the tuple.
- Filtered feature selection selected the best variable subset using probability values.
D. Result and Discussion
The analysis identifies several high-potential variables associated with students’ performance and applies Bayesian classification to support division prediction and timely attention to students needing help.
- Students’ Senior Secondary Examination grade is identified as the strongest potential variable for performance.Its relationship with BCA examination grade is shown in Figure 1.
- Students’ living location is identified as the second high-potential variable for performance.Its relationship with BCA examination grade is shown in Figure 2.
- Students’ medium of teaching is identified as the third high-potential variable for performance.The paper notes greater comfort among mixed- and Hindi-medium students than English-medium students in Uttar Pradesh.
- Mother’s qualification, students’ other habits, family annual income, and family status are also reported as high-potential variables affecting higher final-examination grades.
- Bayesian classification is used to predict students’ divisions from previous-year data and identify students needing special attention.The stated purpose is to support timely action and reduce the failing ratio.
- The study reports that factors beyond students’ own effort significantly influence academic performance.The authors present this as improving insight into existing methods.
Author Profile
The paper profiles Brijesh Kumar Bhardwaj and Saurabh Pal as computer-application academics and researchers with backgrounds in data mining, mathematics, and research supervision.
- Brijesh Kumar Bhardwaj is an Assistant Professor researching Data Mining and Knowledge Discovery.He is affiliated with Dr. R. M. L. Avadh University, Faizabad, India.
- Saurabh Pal is a senior computer-applications lecturer who has authored research papers and guides research scholars.His academic background includes computer science and mathematics.