Analysis of Min-max Algorithm and PSO Algorithm for Data Clustering
Shashwat Mehta, Manmohan Singh, Rajendra Pamula · 2020
This paper propounded the approach of PSO (Particle Swarm Optimization) to cluster data. It has helped us to advent the use of PSO in finding the centroids of a cluster specification by the users. It then looked into two sections of PSO where the first part is using K-means clustering for seeding initial swarm and the second chunk uses PSO to sift the cluster formation by K-means. Unlike previous papers, our evaluation has been extended to some new datasets. Also the use and application of PSO is very pertinent in many sectors (Banking, Telecommunication, E-commerce, IT Industry, Hospital) as this algorithm provides comprehensive approach to a particular problem with its properties such as Fast convergence rate, minute amount of parameters and easy implementation without any computational complexities. But it faces certain fallacy and catastrophic results when the algorithm tries to learn on its parameters such as initial weight, acceleration coefficient (r1 and r2) and to avoid that it needs to be incorporated with some other unsupervised and efficient Machine Learning algorithm that is Min-max K-mean. This paper proposed an idea to get the initial cluster inputs from Min-Max K mean algorithm and to use that factor on the improved PSO algorithm where the stress to find the Gbest does not rely only on the initial parameters and the Particle's Best position (XJ)) can iteratively learned and enhanced their performance.