← Back to Publications List

Comparative Analysis of Classification Techniques in Data Mining: K-Nearest Neighbors, Naive Bayes, and Decision Trees based Approach

Students & Supervisors

Student Authors
Md. Farseen Islam Sattique
Bachelor of Science in Computer Science & Engineering, FST
Md. Towfiq Bin Hasan
Bachelor of Science in Computer Science & Engineering, FST
Md. Fardin Hossain Neloy
Bachelor of Science in Computer Science & Engineering, FST
Md. Rakib Hasan
Bachelor of Science in Computer Science & Engineering, FST
Supervisors
Md. Mortuza Ahmmed
Associate Professor, Faculty, FST

Abstract

The world is running on rapid digital transformation right now; the explosion of data brings immense opportunities and some serious challenges for increasing the value of the information. Data mining is not just about getting success on picking an algorithm. It is about finding the right method with the most convenient computing environment and the details of a specific field. There are three classification techniques which are mostly used techniques in the world right now. K-Nearest Neighbors is often praised and used for the straightforward design and the ability to adapt localized data in the most useful way. But there are also some drawbacks of this technique, it struggles with scaling issues and can become unreliable in the large set of data. Naïve Bayes remains a go-to method for probability calculation, it is mainly valued for the speed it has and the surprisingly strong performance in areas such as text classification, biomedical data, and genetic studies like the cases where its assumption for featuring independence holds up well. Decision Trees such as J48 using the Gini Index bring something incompatible to the table like a simple, rule-based structure that is easy to illuminate. This transparency makes them reliable in domains like medical bioinformatics, where trust and clarity in decision-making are most crucial part. This is not only about algorithms. The platforms used for applying them to play a major role in their own effectiveness. The platforms are praised for their accessible and workflow-oriented design. These ecosystems make it easier to connect the algorithms with complex datasets, such as leukemia gene-expression profiles. The difference between the approaches are accuracy, efficiency and interpretability. They give different accuracy with different training times. The most noticeable thing is no single approach comes with everything in them.

Keywords

Supervised Learning k-Nearest Neighbors Naive Bayes Decision Trees (J48/Gini) Data Mining Tools Comparative Analysis.

Publication Details

  • Type of Publication:
  • Conference Name: 24th International Mathematics Conference
  • Date of Conference: 18/12/2025 - 18/12/2025
  • Venue: University of Chittagong
  • Organizer: Department of Mathematics, University of Chittagong