Ensemble Clustering via Co-Association Matrix Self-Enhancement

Jia, Yuheng; Tao, Sirui; Wang, Ran; Wang, Yongheng

doi:10.1109/tnnls.2023.3249207

Cited by 10 publications

(2 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, we adopt adaptive weights to achieve it. Next, we introduce the CA matrix enhancement assumption A = C + E in the ECCMS model (Jia et al 2023), which considers that the CA matrix in real data is the sum of pure CA matrix and noise matrix. We embed it into the high-order model with the following formula:…”

Section: Our Proposed Methodologymentioning

confidence: 99%

Enhancing Ensemble Clustering with Adaptive High-Order Topological Weights

Xu,

Li,

Duan

2024

AAAI

View full text Add to dashboard Cite

Ensemble clustering learns more accurate consensus results from a set of weak base clustering results. This technique is more challenging than other clustering algorithms due to the base clustering result set's randomness and the inaccessibility of data features. Existing ensemble clustering methods rely on the Co-association (CA) matrix quality but lack the capability to handle missing connections in base clustering. Inspired by the neighborhood high-order and topological similarity theories, this paper proposes a topological ensemble model based on high-order information. Specifically, this paper compensates for missing connections by mining neighborhood high-order connection information in the CA matrix and learning optimal connections with adaptive weights. Afterward, the learned excellent connections are embedded into topology learning to capture the topology of the base clustering. Finally, we incorporate adaptive high-order connection representation and topology learning into a unified learning framework. To our knowledge, this is the first ensemble clustering work based on topological similarity and high-order connectivity relations. Extensive experiments on multiple datasets demonstrate the effectiveness of the proposed method. The source code of the proposed approach is available at https://github.com/ltyong/awec.

show abstract

Section: Our Proposed Methodologymentioning

confidence: 99%

Enhancing Ensemble Clustering with Adaptive High-Order Topological Weights

Xu,

Li,

Duan

2024

AAAI

View full text Add to dashboard Cite

show abstract

“…Then, the framework applies the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering algorithm [17], [18] and the Kernel Density Clustering (KDC) algorithm [19], [20] to split the data into clusters. Then, the output clusters are fed to the co-association matrix to check data similarity [21]- [24]. After that, the AML-CTP extracts a data sample with the same probability distribution as the original population.…”

Section: Introductionmentioning

confidence: 99%

Machine Learning-Based Cellular Traffic Prediction Using Data Reduction Techniques

Nashaat,

Mohammed,

Abdel-Mageid

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Estimating and analyzing traffic patterns become essential in managing Quality of Service (QoS) metrics while assessing internet data traffic in cellular networks. Cellular network planners frequently apply various approaches to predict network traffic. However, most existing studies focus on using the available local data to jointly build prediction models, facing data security challenges and time complexity, especially with multi-dimensional datasets. Therefore, this paper proposes a framework to handle traffic prediction with the considerable potential of Machine Learning (ML) algorithms. An Adaptive Machine Learning-based Cellular Traffic Prediction (AML-CTP) framework is presented to select a suitable ML algorithm for multi-dimensional datasets. Its objective is to streamline and speed up the selection of an appropriate model for predicting network traffic load. The framework employs two density-based clustering algorithms to categorize similar nearby traffic into various clusters, considering data similarity and convergence. Additionally, it assesses data quality and homogeneity by training models with data samples from each cluster to accurately determine the most suitable machine learning model. The optimal model is selected from four supervised predicting algorithms, reducing training time and hardware complexity. Two case studies from a popular telecommunication equipment corporation in Egypt are implemented using reallife cellular traffic with multi-dimensional features. The case studies show that the framework can help reduce the computational cost of training the model and reduce the risk of overfitting. The experimental results show that selecting the best prediction model training could save up to 85% of computational time compared to two state-of-the-art techniques while achieving an accuracy of 98.8%.

show abstract