Category Archives: PhD Thesis

The objective of this study is to delineate the challenges associated with addressing complex optimization issues, with a specific focus on metaheuristic algorithms. A comprehensive investigation was undertaken to explore the principles and categories of these algorithms to gain a deeper understanding of the issues they present and develop effective strategies to overcome them. To challenge these issues, the study explores metaheuristic algorithms, which are known for their effectiveness in solving such problems. However, these algorithms often struggle with getting stuck in local optima and maintaining a balance between exploration and exploitation. Additionally, they exhibit poor searchability and exploitation performance.

To address these challenges, this research work introduces three different algorithms: a modified version of child drawing development optimization MCDDO, a hybrid algorithm combining child drawing development optimization with harmony search CDDO-HS, and a novel metaheuristic called the social psychology interaction behavior algorithm SPIBA, inspired by human social psychology interactions.

The performance of these algorithms is evaluated using various benchmark test functions, including classical and CEC-C06 2019 benchmark functions. Statistical methods, such as ranking and the Wilcoxon rank-sum test, are used to compare the results of these algorithms with the original algorithms, CDDO, HS, and other popular algorithms.

In the beginning, two different approaches were proposed, namely MCDDO and CDDO-HS. The main objective of both techniques is to overcome the issues that the CDDO faces. The CDDO is an example of a human-based metaheuristic approach that may encounter challenges such as getting trapped in local optima, demonstrating suboptimal performance in the exploration phase, and experiencing stagnation in the nearest optimal solution.

The first proposed MCDDO incorporates four key mechanisms: iterative pattern memory PM updating during the exploitation phase, where new experiences are compared with the child's current drawings; a change in the primary rule employed during the exploitation phase; parameter tuning to strike a balance between exploration and exploitation phases; and preservation of the best solution obtained in each iteration and comparing new solutions with the best solution during the exploration phase. Following the completion of the evaluation, the statistical findings indicate a consistent superiority of the proposed approach over standard algorithms, as evident in both average and p-value results. Specifically, out of the nineteen classical test functions and ten CEC-2019 benchmark test functions, the proposed approach demonstrated better performance in thirteen and nine instances, respectively. These results were then compared with those obtained from the JAYA, SCA, ChOA, DA, GPSO, and BOA algorithms. The comparative analysis confirmed that the proposed approach outperformed all other metaheuristic algorithms in four out of the ten CEC-2019 benchmark test functions.

The second proposed method, CDDO-HS, represents a hybridization between CDDO and HS and integrates two crucial mechanisms. Firstly, it relocates the PM to the algorithm's core, updating it with each iteration using the HS algorithm. Secondly, it establishes the PM size at 80% of the overall population, aiming for optimal exploration. After the evaluation, the statistical results reveal that the hybridization approach consistently outperforms standard algorithms in both average and p-value outcomes. Specifically, in comparison with CDDO, it achieves better results in eleven out of nineteen classical test functions and all functions from the CEC-2019 benchmark. When compared with HS, the hybrid approach excels in sixteen out of nineteen classical test functions and seven out of ten CEC-2019 benchmark test functions. These results were then pitted against the ChOA, BOA, FOX, GWO-WOA, WOA-BAT, and DCSO algorithms. The study proved beyond a reasonable doubt that the suggested method is better than all other metaheuristic algorithms in six of the ten CEC-2019 test functions.

In the subsequent phase, SPIBA, an innovative metaheuristic optimization algorithm inspired by social psychology interaction behavior and social interaction—processes involving the stimulus or response of two or more individuals—was developed. These fundamental ideas have been easily incorporated into SPIBA's core, which operates as a single-object and population-based algorithm. SPIBA's performance was compared to that of the ChOA, BOA, FOX, GWO-WOA, WOA-BAT, and DCSO algorithms. The exploration and convergence measures were utilized to assess its success. Their analytical results definitively indicated that the proposed approach beat all other metaheuristic algorithms in six of ten CEC-2019 benchmark test functions.

Additionally, SPIBA was applied to equipment real-world engineering and applied science challenges, specifically in pressure vessel design and the analysis of the pathological IgG fraction in the nervous system. When working in pressure vessel design and compared with eight other algorithms—WOA, GWO, FDO, CFDO, WOAGWO, KMGWO, RFSO, and MFDO—SPIBA appeared as the top-performing algorithm. It showed an average solution quality of 6.01E-05 and the lowest standard deviation of 2.00E-04, guaranteeing the ahead position. In the context of the "Nervous System's Pathological IgG Fraction" application problem, a comparison between SPIBA and Leo revealed a significant improvement in the proposed algorithm's performance.

The evolutionary sophistication method solves optimization problems; however, its effectiveness and scalability can be challenged as problem complexity increases. Population-based evolutionary metaheuristic algorithms heavily rely on operators that determine their overall performance. These operators enhance exploration and exploitation, crucial for effective search and optimization. The research introduces the crossover operator, Lagrangian Problem Crossover (LPX), to boost evolutionary algorithms' performance in tackling new optimization problems. Additionally, it presents Lagrange Elementary Optimization (LEO), a single-objective algorithm where LPX plays a significant role.

The crossover operator in population-based algorithms is crucial for selecting suitable solutions in optimization processes. Its efficiency saves time, minimizes errors, and reduces costs in engineering applications. The initial phase of the study presents an overview of the current crossover methods utilized in engineering operations and problem representation. Furthermore, presenting LPX, it is a fresh and inventive hybrid technique that draws inspiration from the principles of the Lagrangian Dual Function (LDF). Experimental evaluations compare LPX with other standards such as Simulated Binary Crossover (SBX), Blended Crossover (BX), and Qubit-Crossover (Qubit-X) in real-coded crossovers. The results indicate that LPX generally outperforms other methods and shows comparable performance in remaining cases. Specifically, in TF7, LPX demonstrates superior performance and shorter computation time across all three random values compared to Mean (α=0.2) at 0.0048, Standard Deviation (α=0.2) at 0.0031, and time computation (α=0.2) at 143.005 units. Statistical analysis validates the significance and reliability of LPX compared to other crossover standards.

In the second phase of the research, a novel evolutionary method named Leo is introduced. Leo is inspired by the accurate vaccination process that utilizes the human blood albumin quotient.  Leo utilizes a self-adaptive approach, evolving intelligent agents through gene crossover based on fitness function values. The algorithm's accuracy and precision are extensively validated through rigorous testing on diverse benchmark functions, including both traditional and CECC06 2019 benchmarks. Leo's performance is benchmarked against well-known algorithms like Dragonfly, Genetic Algorithm, Practical Swarm Optimization, and others across multiple functions. A comprehensive comparison evaluates Leo's effectiveness and efficiency in solving optimization problems against these established algorithms. In optimizing multimodal test functions (TF8-TF13), particularly TF11, the proposed approach outperformed other algorithms, with an average TF11 value of (2.7393E-08). Notably, across the composite test functions (TF14-TF19), the proposed method exhibited consistently high performance compared to the base algorithms. The statistical analysis supports the research conclusions, and real-world applications of Leo are also showcased. The stability of Leo is confirmed using standard metrics for exploration and exploitation.

In agriculture farming, pests and diseases are the most imperative factor that affects cucumber leaves. Farmers around the globe are currently facing difficulty in recognizing various cucumber leaf diseases. Unfortunately, current manual techniques to diagnose and detect cucumber leaf diseases consumes a large amount of human resources, subjective, laborious and exhibits poor real-time performance. Therefore, there is a demanding need for an effective algorithm that enables the diagnosis of cucumber leaf diseases and pests. This dissertation intends to propose and improve a model using deep learning techniques for the diagnosis and detection of cucumber leaf disease. Throughout the study, various challenges and issues have been identified, necessitating resolution. Foremost, reliable public dataset for real-world scenarios involving cucumber leaf disease images are currently lacking. Secondly, there is a need for an efficient convolutional neural network (CNN) algorithm to effectively balance the trade-off between classifying cucumber leaf diseases and performance. Thirdly, you only look once (YOLOv5) model has raised concerns related to time consumption, storage complexity, low detection accuracy, and limited ability to detect small symptom diseases.

Thus, a new dataset of cucumber leaf disease and pest has been constructed that includes two pests (spider, and leaf miner), two fungal diseases (downy mildew, powdery mildew), one viral disease, and healthy class leaves in a real-world scenario. The dataset has a total of 4868 images. Furthermore, this PhD dissertation focuses on proposing a new CNN algorithm with tuning of hyper-parameters to optimize the model’s performance that comprises image enhancement, feature extraction, and classification. Data augmentation was used to enlarge the datasets and reduce overfitting, while CNN layers were employed to automatically extract features. Then, five cucumber leaf diseases and one healthy leaf are classified. Moreover, an improved YOLOv5 model for precise detection of cucumber leaf disease and pest symptoms was developed. With the aim of reducing the model's size, modifications were applied to the model's hyper-parameters. Additionally, the BottleneckCSP module replaced the C3 module in both the backbone and neck network sections. The detection impact was notably enhanced through the reductions in parameters, number of layers, and computations; in addition to that, the improved model demonstrates the ability to detect even small leaf disease and pest spots. Furthermore, the integration of the convolutional block attention module (CBAM) into both the enhanced and standard YOLOv5l models further demonstrates the effectiveness of the proposed model.

The study evaluated the effectiveness of the proposed CNN model by comparing it to pre-trained models (AlexNet, Inception-V3, and ResNet-50). The experimental results confirmed that the proposed CNN algorithm outperformed the other algorithms in recognizing cucumber disease and healthy leaves, based on both datasets with and without data augmentation. The proposed CNN achieves a recognition accuracy of 98.19% with the augmented self-made dataset and 100% with cucumber plant disease dataset. Furthermore, The experimental results of the detection system indicated that the improved YOLOv5 model achieved a mean average precision (mAP) of 80.10%, along with precision and recall rates of 73.8% and 73.9%, respectively. In a comparative analysis, the improved YOLOv5 model demonstrated superior performance to the original YOLOv5l, YOLOv5n, YOLOv5s, YOLOv5m, and YOLOv5x networks. It also achieved significant reductions in storage complexity, decreasing from 92.8 MB to 13.6 MB, and in training time, reducing from 4 hours and 41 minutes to 2 hours and 58 minutes.