A multi-strategy improved marine predator algorithm for large-scale optimization

View at publisher Summarise with VIRAT
Authors: Bao Liu, Wenyao Ren, Haoran Li, Yuqing Liang, Jiaxin Li, Lei Gao
Cited by: 0
References: 21

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Large-scale complex optimization is critical for many industrial fields. While swarm intelligence algorithms are commonly used to tackle such problems, they often get trapped in local optima, as seen in the Marine Predator Algorithm (MPA). To overcome this limitation and enhance optimization accuracy, this paper proposes a multi-strategy improved MPA (IMPA). Tent mapping is introduced in the population generation phase to improve diversity. During position updates, Cauchy mutation is applied in high-speed and partial equal-speed stages to strengthen global search, while Particle Swarm Optimization (PSO) mechanisms are incorporated in other equal-speed and low-speed stages to enhance local convergence. An adaptive mechanism is also embedded to balance exploration and exploitation. The performance of IMPA is evaluated on eight benchmark test functions, where it demonstrates superior accuracy and stability compared to conventional algorithms such as GA, PSO, DE, and standard MPA. Furthermore, IMPA is integrated with a Bidirectional LSTM (BiLSTM) to improve oil well production prediction. The combined IMPA_BiLSTM model achieves accuracy improvements of 69.47%, 60.23%, and 64.91% over LSTM, BiLSTM, and MPA-BiLSTM models, respectively.

Fields of Study & Subjects

Metaheuristic Optimization Algorithms Research Simulation Techniques and Applications Distributed Control Multi-Agent Systems

Publication details

Journal / source
International Conference on Communication Networks and Machine Learning (CNML)
Pages
1277–1283
Publication date
30 January 2026
Language
EN