Related Experiment Video
Updated: Apr 25, 2026
12:18
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
7.0K
A procedure for extending input selection algorithms to low quality data in modelling problems with application to
José Otero1, Ana Palacios2, Rosario Suárez1
1Computer Science Department, Universidad de Oviedo, Sedes Departamentales, Edificio 1, Campus de Viesques, 33203 Gijón, Spain.
Thescientificworldjournal
|August 13, 2014
Summary
This study introduces a novel method to enhance feature selection for vague data, improving input relevance ranking in low-quality datasets. The technique effectively utilizes available information for better predictive modeling in computer science education.
Area of Science:
- Computer Science
- Data Science
- Machine Learning
Background:
- Selecting informative inputs for modeling is challenging with low-quality data, leading to uncertain feature rankings.
- Existing feature selection algorithms often struggle with vague or imprecise data, limiting their applicability.
- Automatic grading in massive online computer science courses requires identifying relevant code metrics for student qualification.
Purpose of the Study:
- To develop a new procedure for extending crisp feature selection algorithms to handle vague data.
- To model partial knowledge of feature order using possibility distributions and rank them effectively.
- To demonstrate the technique's utility in a real-world application for automatic grading in computer science education.
Main Methods:
- Developed a procedure to adapt crisp feature selection algorithms for vague datasets.
- Utilized possibility distributions to model partial knowledge of feature ordinals.
- Applied ranking to sort possibility distributions for improved feature selection.
- Analyzed 800 source code files from computer science lectures (2013-2014) using the proposed technique.
Main Results:
- The proposed technique effectively makes the most of available information in certain vague datasets.
- Identified the most relevant code metrics for the automatic grading task in computer science courses.
- Demonstrated successful application in a real-world educational context.
Conclusions:
- The new procedure offers a robust method for feature selection with vague data.
- The approach enhances the reliability of identifying informative inputs, even with data uncertainty.
- The technique shows significant potential for improving automated assessment in computer science education.
