33365_Poster for firefly optimization algorithm
- 1. Firefly Optimization Algorithm for Feature Selection Seminar Guide : Mr. Vinit Tribhuvan Unnati Rathi (33365) Abstract โ The modified Firefly Algorithm (FFA) improves feature selection in machine learning by efficiently reducing dataset dimensionality, balancing both classification accuracy and feature reduction. โ This algorithm draws inspiration from the natural flashing behavior of fireflies, and the modified version incorporates a quasi-reflection learning method to overcome limitations of the standard FFA. โ Validation across various datasets shows that the modified FFA outperforms other optimization techniques, such as Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). โ The modified FFA excels in selecting relevant features while maintaining high classification accuracy, making it a valuable tool for machine learning tasks. Algorithm Methods Introduction References โขRagab, M. (2024). Hybrid firefly particle swarm optimisation algorithm for feature selection problems. Expert Systems, 41(7), e13363. โขIbrahim, H. T., Mazher, W. J., & Yaseen, Z. F. (2024). Hybrid Feature Selection Approach Based on Firefly Algorithm and Simulated Annealing for Cancer Datasets. University of Thi-Qar Journal for Engineering Sciences, 14(1), 1-9. โขBezdan, Timea, et al. "Feature selection by firefly algorithm with improved initialization strategy." 7th conference on the engineering of computer based systems. 2021. โขXu, H., Yu, S., Chen, J., & Zuo, X. (2018). An improved firefly algorithm for feature selection in classification. Wireless Personal Communications, 102, 2823-2834. โขEmary, E., Zawbaa, H. M., Ghany, K. K. A., Hassanien, A. E., & Parv, B. (2015, September). Firefly optimization algorithm for feature selection. In Proceedings of the 7th balkan conference on informatics conference (pp. 1- 7). โขJohari, N. F., Zain, A. M., Noorfa, M. H., & Udin, A. (2013). Firefly algorithm for optimization problem. Applied Mechanics and Materials, 421, 512-517. โขYang, X. S., & He, X. (2013). Firefly algorithm: recent advances and applications. International journal of swarm intelligence, 1(1), 36-50. โขFeature Selection Techniques in Machine Learning: Geeks for geeks โข Initialization: The quasi-reflection learning method ensures diverse initial solutions, allowing better coverage of the feature space and minimizing premature convergence. โข Attraction and Movement: Fireflies move towards brighter (better) solutions, with attraction being proportional to the brightness difference. This mechanism explores the search space to find optimal feature subsets. โข Fitness Function: It evaluates the quality of feature subsets based on two main criteriaโ classification accuracy and feature reduction. This ensures a balance between model performance and computational efficiency. โข Convergence: The algorithm iterates through the search space, refining feature subsets until an optimal or near-optimal solution is reached. Advancements like adaptive parameter control further improve its convergence. In today's data-driven world, extracting key insights from large datasets is critical but challenging due to irrelevant or redundant features. Feature selection helps streamline this process by identifying the most important features while maintaining or improving model accuracy. Inspired by the natural flashing of fireflies, the Firefly Optimization Algorithm (FFA) offers an innovative solution. The modified FFA, enhanced with quasi-reflection learning, excels in efficiently selecting optimal feature subsets, boosting both performance and reducing computational complexity. Conclusion โข This presentation highlights the advancements made in the modified Firefly Algorithm (FFA) for feature selection in machine learning. By incorporating the quasi-reflection learning technique, the enhanced FFA significantly improves exploration of the feature space, overcoming premature convergence and enabling the precise identification of optimal feature subsets. โข When compared to traditional methods like the original FFA, Particle Swarm Optimization (PSO), and Genetic Algorithms (GA), the modified FFA consistently delivers superior results in terms of both accuracy and computational efficiency. Its versatility is evident across a wide range of applications, including healthcare, finance, cybersecurity, and IoT, making it a highly adaptable tool in the machine learning ecosystem. โข In conclusion, the modified FFA proves to be a powerful and efficient approach for feature selection, surpassing conventional techniques. Its ability to handle complex, high-dimensional datasets with precision solidifies its standing as an invaluable resource for enhancing model performance in a variety of real-world scenarios. โข Improved Search Efficiency: The integration of quasi-reflection learning enhances the exploration phase, allowing the algorithm to search the feature space more effectively. This reduces the chances of getting trapped in local optima and improves the precision of selected feature subsets. โข Balanced Feature Selection: By optimizing both classification accuracy and feature reduction, the modified FFA strikes an ideal balance between retaining relevant features and minimizing dataset dimensionality. This ensures a compact yet highly informative feature set, boosting model performance while reducing computational overhead. โข Versatility Across Domains: The modified FFA has demonstrated superior performance across a wide range of applications, including healthcare, finance, cybersecurity, and IoT. Its ability to handle diverse datasets and different types of optimization problems makes it highly adaptable to various real-world scenarios. โข Outperformance of Traditional Methods: When compared to traditional algorithms like Particle Swarm Optimization (PSO) and Genetic Algorithms (GA), the modified FFA consistently achieves higher classification accuracy and computational efficiency, making it a preferred choice for feature selection tasks. โข Scalability for High-Dimensional Data: The algorithm is particularly effective for high- dimensional datasets, where traditional methods often struggle. Its capability to handle large-scale data without significant performance loss makes it suitable for complex machine learning problems. Advantages