الفهرس | Only 14 pages are availabe for public view |
Abstract Abstract In recent years, the exponential growth in computing power and the existence of massive datasets integrated with the algorithms’ improvement led to an unparalleled surge of interest in machine learning topic. Nowadays, machine learning algorithms are successfully applied in a wide range of domains. It has been employed for regression, classification, dimensional reduction, especially for high-dimensional datasets and clustering. In fact, machine learning algorithms played a remarkable role in huge parts of our daily life such as anomaly detection, medical diagnosis, email/spam filtering, web searches, credit card fraud detection, financial analysis, and many more. Although machine learning proved its efficiency in many fields, it has several problems, which can be categorized as follows: feature selection, classification, regression, rule extraction, clustering, contrast enhancement of images and parameters setting. Recently, nature-inspired meta-heuristic algorithms have become influential and powerful in many applications. They have been used as efficient tools to deal with machine learning problems. Swarm intelligence algorithms are a class of meta-heuristic optimization algorithms. However, swarm intelligent algorithms have been shown to optimize a wide range of problems successfully; some of these algorithms often suffer from premature convergence and entrapment in a local optimum, especially in complex optimization problems. Therefore, several studies have been proposed in literature to boost the performance of these algorithms and to overcome these problems. The main goals of the thesis are to introduce enhanced versions of swarm intelligence algorithms and to apply the proposed algorithms on different optimization problems. In the present study, we are mainly focused on boosting the performance of the Salp Swarm Algorithm (SSA), and Grasshopper Optimization Algorithm (GOA), where three different enhancing methods are used. These methods are the principles of chaos theory, the random walk of l´evy-flight method and the principles of quantum computing. Also, four optimization problems are considered. These problems are global optimization problem, feature selection optimization problem, contrast enhancement of images optimization problem and finally, parameters setting optimization problem. Each one of the proposed algorithms is applied to solve two optimization problems. Meanwhile, all the proposed algorithms are applied to the global optimization problem. SSA is one of the recently proposed algorithms driven by the simulation behavior of salps. However, like most of the meta-heuristic algorithms, it suffered from stagnation in local optima and low convergence rate. Recently, chaos theory has been successfully applied to solve these problems. In this thesis, a novel hybrid approach based on SSA and the chaos theory is proposed. The proposed Chaotic Salp Swarm Algorithm (CSSA) is applied to global optimization and feature selection optimization problems, where fourteen unimodal and multimodal global benchmark optimization problems and twenty benchmark datasets are adopted. Additionally, ten different chaotic maps are employed iii to enhance the convergence rate and resulting precision. Simulation results showed that the proposed CSSA is a promising algorithm. Also, the results showed the capability of CSSA in finding an optimal feature subset, which maximizes the classification accuracy, while minimizing the number of selected features. Moreover, the results showed that logistic chaotic map is the optimal map of the used ten, which can significantly boost the performance of original SSA. Additionally, a modified GOA based on the random walk of L´evy-flight method, called as (LevyGOA) is proposed. GOA is one of the recently meta-heuristic optimization algorithms. Although, GOA has shown good performance, it still has demerits respect to low precision, slow convergence and easily stuck at local minima. The experimental results showed that LevyGOA able to provide a better trade-off between exploitation and exploration, which makes LevyGOA faster and more robust than GOA. LevyGOA is further compared with other meta-heuristic optimization algorithms and the basic GOA for solving two optimization problems. These problems are global optimization problem and parameters optimization of SVM, where two global benchmark functions and six well-known benchmark datasets are used. The experimental results showed that LevyGOA could significantly improve the performance of GOA. The results demonstrated that LevyGOA outperforms the other algorithms on a majority of the benchmark functions and benchmark datasets. Finally, a new hybrid approach based on quantum computing and SSA, called Quantum Salp Swarm Algorithm (QSSA) is proposed. The proposed QSSA is used to boost the performance of SSA through finding the optimal trade-off between exploitation and exploration. QSSA relies on embedding the integration of the quantum operators and interference in the optimization process of SSA and the quantum representation of the search space. The proposed QSSA is applied to global optimization and contrast enhancement of images optimization problems. It is tested on twenty-nine global benchmark functions and eight benchmark images. The simulation results showed that employing the principles of quantum can significantly boost the performance of SSA. Also, the performance of QSSA is compared with SSA and other recent and well-known optimization algorithms. The results on global benchmark datasets demonstrated the capability of QSSA to find the global optima for most of unimodal and multimodal benchmark functions, especially with complex search space. Moreover, the results revealed that the proposed image enhancement based QSSA is robust and can distinctly enhance the contrast of the color images. To the best of our knowledge, this thesis is first to propose the hybridization of chaos theory with SSA, random walk of l´evy-flight method with GOA, and the principles of quantum computing with SSA. Additionally, this thesis is first to apply the hybridization between chaos theory and SSA on feature selection optimization problem, the hybridization between l´evy-flight method and GOA on parameter setting optimization problem, and the hybridization between quantum theory and SSA on contrast enhancement optimization problem. |