Predicting the Number of Clusters (K) Without Distance and Statistical Analysis

Clustering is a fundamental task in unsupervised learning, essential for data analysis and pattern recognition. A significant limitation of traditional clustering algorithms is the need to predefine the number of clusters (k -value), a hyperparameter that can heavily influence results. This paper pr...

Full description

Saved in:
Bibliographic Details
Published in:International Conference on Knowledge and Smart Technology pp. 76 - 81
Main Authors: Rabari, Rohhan, Lursinsap, Chidchanok
Format: Conference Proceeding
Language:English
Published: IEEE 26.02.2025
Subjects:
Adaptive Clustering Algorithms
Classification algorithms
Cluster Number Prediction
Clustering algorithms
Clustering without Predefined Parameters
Data Clustering Optimization
Data Distribution Analysis
Data models
Dimensionality in Clustering
Latent Vector Transformation
Machine Learning Frameworks
Neural Network Architectures
Neural Network Pre-training
Neural networks
Pattern Detection in Data
Performance Evaluation in Clustering
Prediction algorithms
Predictive Clustering
Synthetic data
Synthetic Data Generation
Three-dimensional displays
Training
Unsupervised learning
Unsupervised Learning Techniques
Vectors
ISSN:2473-764X
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Clustering is a fundamental task in unsupervised learning, essential for data analysis and pattern recognition. A significant limitation of traditional clustering algorithms is the need to predefine the number of clusters (k -value), a hyperparameter that can heavily influence results. This paper presents a novel approach that leverages latent vector transformation to predict the optimal k -value based on the distribution patterns of data points, eliminating the need for prior parameter specification. By utilizing synthetic data distributions that range from single to multiple clusters, we demonstrate how a neural network can be trained to effectively identify the underlying structure of the data, thus enhancing the clustering process.
ISSN:2473-764X
DOI:10.1109/KST65016.2025.11003305