Source-linked AI summary
Data Mining : A prediction of performer or underperformer using classification
Umesh Kumar Pandey, Saurabh Pal
TL;DR
Educational institutions hold large datasets whose hidden relationships may support understanding and decision making. This paper uses Bayesian classification on previous-year student data to predict performance categories and examination divisions. It reports that the resulting predictions can help identify students needing special attention and support timely action.
Problem
Educational data mining addresses the need to extract useful information from growing educational databases for understanding students and improving educational decision making.
Method
The study applies Bayesian classification to previous-year student data, estimating class probabilities from conditional probabilities of attribute values.
Results
The model predicts student examination divisions from attributes such as medium, caste category, and academic class, selecting the class with the highest probability.
Takeaways & Limitations
The paper states that these predictions can identify students needing special attention and support timely action to reduce failing ratios.
Abstract
from arXiv · showhide
Now a day's students have a large set of data having precious information hidden. Data mining technique can help to find this hidden information. In this paper, data mining techniques name Byes classification method is used on these data to help an institution. Institutions can find those students who are consistently perform well. This study will help to institution reduce the drop put ratio to a significant level and improve the performance level of the institution.
I. INTRODUCTION
Educational databases are growing, creating a need to uncover useful relationships and information. This study uses Bayesian classification to predict whether new students will be performers or underperformers from previous-year data.
- Educational data storage is increasing as processing power and storage become cheaper, creating larger databases for analysis.
- Data mining techniques can uncover hidden relationships and information in large educational databases.
- Classification assigns attribute values to predefined discrete classes, making it suitable for categorical prediction.
- The study analyzes previous-year data to predict whether new students will be performers or underperformers using Bayesian classification.
II. DATA MINING
Data mining uses encoded procedures to discover models or patterns from databases. These outputs support analysis beyond traditional database querying.
- Data mining is also described as exploratory analysis, data-driven discovery, and deductive learning.
- Data mining algorithms take data as input and produce models or patterns through precisely encoded finite rules.
- Discovered structures describing most of a dataset are called models, while structures with local data properties are called patterns.
- Data mining database access differs from traditional access in its query, data, and output.
III. BACKGROUND AND RELATED WORK
Educational data mining applies specialized methods to student data to understand learning and support educational decision making. Prior work has used prediction and grouping for student-focused analyses.
- Educational data mining identifies, extracts, and evaluates variables related to students’ learning processes.
- Its applications include summarization, classification rules, association discovery, change analysis, and anomaly detection.
- Educational data mining is used to enhance educational processes and improve decision making.
- Prior studies used prediction to identify factors associated with test scores and grouping to distinguish students’ course-completion patterns.
IV. CLASSIFICATION
Classification is a predictive modeling approach that maps known input values to an unknown variable or predefined class. It uses previously known results to make predictions about new data.
- Predictive modeling creates or selects a model to estimate the probability of an outcome from input data.
- Classification predicts data values using known results from other data.
- A predictive model learns a mapping from an input vector x to a scalar output y.
- Classification maps data into predefined classes and is often called supervised learning because the classes are established before examining the data.
V. BAYESIAN CLASSIFICATION
Bayesian classification estimates class membership from conditional and prior probabilities. The naïve approach assumes attribute independence and uses these probabilities to predict target classes.
- Bayes classification applies Bayes rule of conditional probability to estimate a property from data evidence or input.
- The naïve Bayes approach assumes independence between the various attribute values.
- Conditional and prior probabilities describe the data and are then used to predict class membership for a target tuple.
- Naïve Bayes requires only one scan of the training data and can handle missing values by omitting the corresponding probability.
VI. EDUCATIONAL DATA MINING IN HIGHER EDUCATION
Higher education faces social and structural inequalities that affect students’ ability to compete, while quality assurance is a major policy concern. Educational data mining develops methods to understand students and learning settings and address knowledge gaps in higher education.
- Higher education confronts social inequalities rooted in historical, cultural, and economic structures that influence individual competitiveness.
- Quality assurance in higher education has become a top policy priority in many nations.
- Educational Data Mining explores educational data to better understand students and the settings in which they learn.
- Data mining methodology can help address knowledge gaps that may prevent higher-education management from achieving quality objectives.
VII. APPLICATION
The application uses student data from affiliated colleges to build a naïve Bayes classification model. The resulting probabilities predict student divisions from selected demographic and academic attributes.
- Application: The study analyzes data from degree colleges affiliated with Dr. R. M. L. Awadh University to predict student performance.
- Application: The initial dataset contains 600 PGDCA students from the 2009–10 computer science department course.
- Application: Data preparation joins tables into a single first-normal-form table and removes errors.
- Application: Selected fields include sex, language medium, bachelor’s stream, and obtained division.
- Application: The naïve Bayes model estimates class priors and conditional probabilities by counting occurrences in the training data.
- Application: The class with the highest product of conditional probabilities is selected for the target tuple.
- Results and discussion: For a medium, OBC, BA non-calculative student, the model predicts a second-division final-exam result.
VIII. CONCLUSION
The paper applies Bayesian classification to previous-year student data to predict divisions. It presents the method as a way to support improvement efforts and identify students needing special attention.
- Bayesian classification predicts students’ divisions using their previous-year database records.
- The study is intended to help students and teachers improve student division outcomes.
- The approach is also intended to identify students needing special attention so action can be taken at the appropriate time.