K-means

Purpose

The objective of the program is to automatically cluster similar elements.

Rules

• The user can choose one of the files - "noraml.txt" or "unbalance.txt"
• The user can choose the count of the clusters.
• The program should use random restart, keeping the best solution in the new generations.
• The best solution is chosen based on internal cluster distance.

About the program

The program is using the algorithm K-means to automatically cluster similar elements with two attributes, provided as points in the Euclidean space in the files "normal.txt" and "unbalance.txt".

• "normal.txt" - 4 Gaussian clusters
• "unbalance.txt" - 8 Gaussian clusters

To run the program

• Run python3 main.py
• Input file name ("normal" or "unbalance")
• Input number of clusters

Output

• The internal cluster distance of each generation is displayed.
• Image that shows the different clusters in different colors.

Output - "normal.txt"

Output - "unbalance.txt"

Bonus optimizations that can be applied

• Compare results with K-means++
• Compare results with Soft k-means
• Use intercluster distance.
• Use combination of both internal cluster distance and intercluster distance.