Flower pollination Algorithm based on triple differential evolution and its application in Research on the classification and leveling of power cloud data
Jiamao Han, Yaxing Wei, Yingchao Zhen, Ji Shi, Lei Wang, Yu Zhang, Wenbo Sun, Rong Ling Wu · 2024
For the clustering problem of data mining, this paper designs a flower pollination algorithm based on triple differential evolution (FPA-TDE). FPA-TDE uses DE/best/1, DE/best/2 and DE/rand/2 to enhance the search orientation and population richness of global search process. In local search process, DE/rand/1 and DE/current-to-rand/1 are used to improve the optimization accuracy, and the MDE strategy is designed to improve the convergence speed while maintaining the searching continuity. In addition, the external archive is applied to dynamically update the scaling factors of all differential evolution strategies. It forces the search process and guides individuals to search for potential high-quality solution sub-regions, so as to avoid trapping into local optimum. The extensible benchmark functions are used to verify the optimization performance of FPA-TDE. The results show that FPA-TDE can maintain good optimization performance and operation efficiency under different dimensional benchmark functions. Moreover, FPA-TDE is applied to solve clustering problem of data mining, and the experimental results show that the algorithm can satisfy the clustering requirements of high accuracy and stability.