Investigation of Applications of Jaya Algorithm on Clustering and Classification

Koay, Lek Sheng (2023) Investigation of Applications of Jaya Algorithm on Clustering and Classification. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text Full Text -Koay Lek Sheng.pdf Restricted to Registered users only Download (2MB)

Abstract

Malaysia as a main global palm oil export, requires a high yield of oil palm for the extraction of palm oil. However, several bottlenecks have been faced in the oil palm industry. To solve the problems faced, solutions were provided for identifying the ripeness of oil palm fruit. Besides that, to ensure a more efficient planting process, clustering is done to ease the path planning of drones. In this project, a Jaya Algorithm based clustering is proposed to solve the problem faced such auto-grading and path planning. This project is expected to improve and enhance the oil palm industry in Malaysia. Clustering is a machine learning method that separates data points or objects into groups based on their similarities or features. Clustering has been widely used in various fields such as telecommunication, agriculture, and other fields. Through studies, it has shown that clustering has played an important role in path planning, image classification and auto-grading. Besides that, previous works on clustering have been studied and more knowledge and procedures of clustering for various engineering applications. The Jaya method is a parameter-free method that requires only the population size and number of iterations as parameters. Unlike other algorithms which require algorithm-specific parameters, improper tuning of algorithm-specific parameters will cause additional computational efforts. Jaya Algorithm is not affected as it does not require algorithm-specific parameters. From various studies, it has shown that Jaya Algorithm has the capability to solve various problems whether it is constrained or unconstrained problems as it can be hybridized with other algorithms or techniques. From the studies, it also shows that the Jaya Algorithm is able to improve various systems as its greedy search to find the best solution and avoid the worst solution for every iteration. Since Jaya Algorithm is able to perform clustering with better performance than other algorithms, Jaya Algorithm will be implemented to optimize the classifier for auto-grading of oil palm fruits. Furthermore, Jaya Algorithm will be applied as one of the optimization methods of clustering for path planning of UAVs. Therefore, a Jaya Algorithm Based Clustering Method is proposed for meeting the objectives. In this system, MATLAB is chosen as the software development tools for its high performance in developing algorithms. Data normalisation and data partitioning is done for the ease of training of the system. The Jaya Algorithm is used in MATLAB for clustering based on extracted and normalized image features. The data is split into 70% training and 30% testing sets, and the accuracy is evaluated using a confusion matrix. The algorithm is also applied to cluster coordinates using various data initialization methods, and the fitness of the clustering is compared depending on the approach used. The Jaya Algorithm is a promising tool for classifying oil palm fruit ripeness based on color and density features, but further improvements are needed to better distinguish overripe fruits. To enhance its performance, future developments should focus on improving feature selection and providing more training data for all classes. Meanwhile, for clustering coordinates, the min-max thresholding technique is an effective data initialization strategy, while the binning method is not recommended. Setting a distance restriction between centroids and data points can increase cluster quality and identify the optimal number of clusters. However, further effort is needed to improve the clustering coordinates of the system. Overall, these techniques have the potential to lead to more efficient and effective agricultural practices.