This paper proposes a hybrid approach for solving data clustering problems. This hybrid approach used one of the swarm intelligence algorithms (SIAs): grasshopper optimization algorithm (GOA) due to its robustness and...
详细信息
This paper proposes a hybrid approach for solving data clustering problems. This hybrid approach used one of the swarm intelligence algorithms (SIAs): grasshopper optimization algorithm (GOA) due to its robustness and effectiveness in solving optimization problems. in addition, a local search (LS) strategy is applied to enhance the solution quality and access to optimal dataclustering. The proposed algorithm is divided into two stages, the first of which aims to use GOA to prevent getting trapped in local minima and to find an approximate solution. While the second stage aims by LS to increase LS performance and obtain the best optimal solution. In other words, the proposed algorithm combines the exploitation capability of GOA and the discovery capability of LS, and integrates the merits of both GOA and LS. In addition, 7 well-known datasets that commonly used in several studies are used to validate the proposed technique. The results of the proposed methodology are compared to previous studies;where statistical analysis, for the various algorithms, indicated the superiority of the proposed methodology over other algorithms and its ability to solve this type of problem. (C) 2021 The Authors. Published by Atlantis Press B.V.
dataclustering is one of the most common and challenging problems in the machine learning domain. It requires an efficient method to be addressed. This paper proposed a new version of the Flow Direction Algorithm (FD...
详细信息
dataclustering is one of the most common and challenging problems in the machine learning domain. It requires an efficient method to be addressed. This paper proposed a new version of the Flow Direction Algorithm (FDA) to solve various optimization problems. The proposed method is called FDAOA, which enhanced the performance of the original Flow Direction Algorithm by the arithmetic operators that have been used in the Arithmetic Optimization Algorithm (AOA). The main aim of the proposed FDAOA is to avoid the recognized weaknesses in the original methods;stuck in the local area, premature convergence, and weak equilibrium between the exploration and exploitation search mechanisms. The proposed method is tested on two sets of various problems to validate its performance. In the first set, twenty-three benchmark functions are used, which belong to three categories;seven unimodal functions, six multimodal functions, and ten fixed dimension functions. In the second set, eight common data clustering problems are used to prove the ability of the proposed FDAOA to deal with real-world optimization problems. The results of the proposed method are compared with other well-established methods, and the proposed FDAOA achieved promising output compared to the other methods on various tested problems. The proposed method got the optimal clustering solutions almost in all the tested data clustering problems with clear significant improvements against the other comparative methods.
The Aquila Optimizer (AO) is a newly proposed, highly capable metaheuristic algorithm based on the hunting and search behavior of the Aquila bird. However, the AO faces some challenges when dealing with high-dimension...
详细信息
The Aquila Optimizer (AO) is a newly proposed, highly capable metaheuristic algorithm based on the hunting and search behavior of the Aquila bird. However, the AO faces some challenges when dealing with high-dimensional optimization problems due to its narrow exploration capabilities and a tendency to converge prematurely to local optima, which can decrease its performance in complex scenarios. This paper presents a modified form of the previously proposed AO, the Locality Opposition-Based Learning Aquila Optimizer (LOBLAO), aimed at resolving such issues and improving the performance of tasks related to global optimization and dataclustering in particular. The proposed LOBLAO incorporates two key advancements: the Opposition-Based Learning (OBL) strategy, which enhances solution diversity and balances exploration and exploitation, and the Mutation Search Strategy (MSS), which mitigates the risk of local optima and ensures robust exploration of the search space. Comprehensive experiments on benchmark test functions and data clustering problems demonstrate the efficacy of LOBLAO. The results reveal that LOBLAO outperforms the original AO and several state-of-the-art optimization algorithms, showcasing superior performance in tackling high-dimensional datasets. In particular, LOBLAO achieved the best average ranking of 1.625 across multiple clusteringproblems, underscoring its robustness and versatility. These findings highlight the significant potential of LOBLAO to solve diverse and challenging optimization problems, establishing it as a valuable tool for researchers and practitioners.
This paper addresses the scalability issue in spectral analysis which has been widely used in data management applications. Spectral analysis techniques enjoy powerful clustering capability while suffer from high comp...
详细信息
ISBN:
(纸本)9781424489589
This paper addresses the scalability issue in spectral analysis which has been widely used in data management applications. Spectral analysis techniques enjoy powerful clustering capability while suffer from high computational complexity. In most of previous research, the bottleneck of computational complexity of spectral analysis stems from the construction of pairwise similarity matrix among objects, which costs at least O(n(2)) where n is the number of the data points. In this paper, we propose a novel estimator of the similarity matrix using K-means accumulative consensus matrix which is intrinsically sparse. The computational cost of the accumulative consensus matrix is O(nlogn). We further develop a Non-negative Matrix Factorization approach to derive clustering assignment. The overall complexity of our approach remains O(nlogn). In order to validate our method, we (1) theoretically show the local preserving and convergent property of the similarity estimator, (2) validate it by a large number of real world datasets and compare the results to other state-of-the-art spectral analysis, and (3) apply it to large-scale data clustering problems. Results show that our approach uses much less computational time than other state-of-the-art clustering methods, meanwhile provides comparable clustering qualities. We also successfully apply our approach to a 5-million dataset on a single machine using reasonable time. Our techniques open a new direction for high-quality large-scale data analysis.
In this work, an in-distinguish migratory bird swarm based algorithm is presented and the same applied to the challenging optimal data clustering problems. Many algorithms that are inspired by the nature are giving su...
详细信息
ISBN:
(纸本)9781538633601
In this work, an in-distinguish migratory bird swarm based algorithm is presented and the same applied to the challenging optimal data clustering problems. Many algorithms that are inspired by the nature are giving super-efficient solutions. The migratory behavior of Northern Bald Ibises (Geronticus eremita) for food paves the path to devise a new optimization algorithm termed as NOA. To examine the performance of NOA algorithm, benchmarking is done in two different phases. In the first phase, 23 standard mathematical testing functions are employed to examine the optimization characteristics of NOA. Secondly, solved 10 well-known data cluster problems to test the numerical efficiency of NOA. The obtained experimental results and statistical analysis of these two phases are portrayed in graphical and tabular form. The comparisons have been made with other futuristic algorithms and it proves that the devised NOA optimization algorithm is good at benchmark function optimization problems as well as optimal data clustering problems.
In data mining, clustering is an important data analysis concept. It plays a vital role in extracting the useful hidden knowledge from large input datasets. This unsupervised technique partitions the input dataset int...
详细信息
ISBN:
(纸本)9781538633601
In data mining, clustering is an important data analysis concept. It plays a vital role in extracting the useful hidden knowledge from large input datasets. This unsupervised technique partitions the input dataset into groups called clusters. The data objects mapping is done into clusters such clusters should maintain similarity between the objects within same cluster and dissimilarity between the data objects in different clusters. In this process factors like distance measuring techniques, initial conditions and criterion functions playa key role in finding optimal clusters of data. Many optimization algorithms have come into existence to resolve these types of optimization problems. But still finding optimal clusters is a big challenging task. This work presents hybrid version of the recently devised nature-inspired algorithm i.e. Tornadogenesis Optimization Algorithm (TOA) for solving data clustering problems using BB-BC. We framed this work in two phases wherein the first phase testing for optimization performance on 23 standard mathematical benchmark functions took place, in the second phase numerical ability is tested by applying hybridized Tornadogenesis Optimization Algorithm (HTOA) on 10 real-world data clustering problems. In addition to that various distance measuring techniques used to test the improvement in clustering performance. We portrayed the obtained results in tabular and graphical forms. Various analysis and comparisons have been made and found that the performance of proposed HTOA is good at solving data clustering problems using Euclidean distance measuring technique.
暂无评论