A Novel Model on Reinforce K-Means Using Location Division Model and Outlier of Initial Value for Lowering Data Cost

Area of Science:

• Computer Science
• Data Science
• Machine Learning

Background:

• Semi-structured and unstructured data analysis often involves challenges with outliers and noise.
• The K-means algorithm, a popular clustering method, suffers from issues with initial centroid selection, leading to biased results and higher costs.
• Outliers can significantly impact cluster accuracy and lead to misclassification.

Purpose of the Study:

• To address the limitations of the K-means algorithm concerning initial centroid selection.
• To propose a novel K-means algorithm that reduces computational costs and improves accuracy in handling outliers.
• To minimize error rates associated with outliers in data clustering.

Main Methods:

• Developed a new K-means algorithm focusing on improved initial centroid selection strategies.
• Implemented an initial center point approach based on space division and outlier identification.
• Utilized an improved algorithm for space division and distance measurement to minimize outlier impact.

Main Results:

• Reduced clustering calculation costs by approximately 13-14% compared to existing methods.
• Demonstrated a significant reduction in the frequency and impact of outliers.
• Achieved a reduction of outliers by about 60% and a lower effectiveness index.

Conclusions:

• The proposed K-means algorithm offers a more efficient and accurate approach to clustering data with outliers.
• The method effectively lowers execution costs and enhances data classification by optimizing initial centroid selection.
• This research contributes to more robust data analysis techniques for complex datasets.