A novel sequence space related to [Formula: see text] defined by Orlicz function with application in pattern...

Area of Science:

• Pattern Recognition
• Data Mining
• Machine Learning

Background:

• Clustering algorithms group data based on similarity, often using distance measures.
• The effectiveness of clustering relies on appropriate distance measures tailored to the dataset.
• Standard k-means clustering can be limited by its default distance metric.

Purpose of the Study:

• To propose a new sequence space and a novel distance measure.
• To modify the k-means clustering algorithm using the proposed distance measure.
• To evaluate the performance of the enhanced k-means algorithm against the standard version.

Main Methods:

• Development of a new sequence space using an Orlicz function.
• Definition of a new distance measure with component-wise weighting.
• Modification of the k-means algorithm incorporating the new distance measure.
• Implementation and comparison on two-moon and path-based datasets from the UCI repository.

Main Results:

• The proposed distance measure exhibits useful properties within the new sequence space.
• The modified k-means algorithm achieved higher clustering accuracy than the standard k-means.
• Empirical results on real-world datasets validate the proposed method's efficacy.

Conclusions:

• The novel distance measure and modified k-means algorithm offer improved clustering performance.
• This approach provides a more effective way to handle data similarity in pattern recognition.
• The study highlights the importance of tailored distance measures for clustering algorithms.