An Improved K-means Algorithm Based on Multiple Clustering and Density

Ling, Yulong; Zhang, Xiao

doi:10.1145/3457682.3457695

Cited by 4 publications

(4 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An additional issue is the instability of the assignment of clusters [16]. To overcome instability, [24] combines density and multiple clustering. This solution improves the running time and stability of the clustering by choosing the centroids according to the furthest distance and the highest density principle.…”

Section: Literature Reviewmentioning

confidence: 99%

See 1 more Smart Citation

Enhancing K-means Clustering Results with Gradient Boosting: A Post-Processing Approach

Alzakan,

Almousa,

Almarzoqi

et al. 2024

IJACSA

View full text Add to dashboard Cite

As the volume and complexity of data continue to grow exponentially, finding efficient and accurate clustering algorithms has become crucial for many applications. Kmeans clustering is a widely used unsupervised machine learning technique for data analysis and pattern recognition. Despite its popularity, k-means suffers from certain limitations, such as sensitivity to initial conditions, difficulty in determining the optimal number of clusters, and the potential for misclassification. This research paper proposes an enhanced approach for improving the accuracy and performance of the k-means clustering algorithm by incorporating post-processing techniques using a gradient boosting algorithm. The proposed method comprises training the gradient boosting model on the labeled training set, i.e., the samples with correct cluster assignments obtained from the kmeans algorithm, to predict the correct cluster assignments for the misclassified samples in the testing set. This results in refined cluster assignments for the testing set. The k-means algorithm is only used initially to cluster the data and obtain initial cluster assignments. The effectiveness of the proposed approach is validated through experiments on several benchmark datasets, and the results show a significant improvement in clustering accuracy and robustness compared to the standard k-means algorithm. The proposed approach has the potential to enhance the performance of k-means in various real-world applications and domains.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

“…This solution improves the running time and stability of the clustering by choosing the centroids according to the furthest distance and the highest density principle. However, solutions that use just density have a high time complexity [24].…”

Section: Literature Reviewmentioning

confidence: 99%

Enhancing K-means Clustering Results with Gradient Boosting: A Post-Processing Approach

Alzakan,

Almousa,

Almarzoqi

et al. 2024

IJACSA

View full text Add to dashboard Cite

show abstract

“…e sample data acquisition topology of a sparse scattered multisensor array network is established using the Voronoi polygon topology, and the data structure model is examined. According to the sample data structure analysis of sparse scattered multisensor array, the data distribution imbalance control algorithm is used for optimization control [9] as shown in Figure 2.…”

Section: Analysis Of Data Structure Of Sensor Array Samplingmentioning

confidence: 99%

Sample Density Clustering Method Considering Unbalanced Data Distribution

Wang

2022

Mobile Information Systems

View full text Add to dashboard Cite

The data distribution of the multidimensional array sensor is unbalanced in data sample collection. To improve the clustering ability of data samples, a data density clustering method of sparse scattered points and multisensor array sensor samples based on the analysis of unbalanced data distribution characteristics is proposed. The sparse scattered multisensor array network’s sample data collection structure is created using the Voronoi polygon topology. By analyzing the unbalanced parameters between data classes and reconstructing the characteristic space of data sample sequence, the time series of sample data collected by sparse scattered multisensor array is reorganized, and the statistical characteristic quantity and high-order cumulant of sample data collected by sparsely scattered multisensor array are extracted. Combined with the learning algorithm of unbalanced data distribution sample feature fusion, the fuzzy clustering of sample data information flow collected by sparse scattered multisensor array elements is realized. According to the feature clustering and convergence analysis, the sparse scattered feature detection method is adopted to realize the data density clustering and data structure optimization configuration of sparse scattered multisensor array elements. The test results show that the method in this paper has good convergence, strong spectrum expansion ability, and low error rate of data clustering when collecting samples with sparse scattered points and multisensor arrays.

show abstract

“…K-means algorithm is highly dependent on the selection of initial center points, so it is susceptible to the influence of outliers, which leads to clustering into local optimal solutions. For this reason, some papers have proposed improved k-means algorithms [6,7] to solve the initial center points selection problem of traditional k-means algorithm, and these algorithms have good results on small data.…”

Section: Introductionmentioning

confidence: 99%

Parallel MS-Kmeans clustering algorithm based on MapReduce

Wang

2022

Preprint

View full text Add to dashboard Cite

Aimed at the problems of initial points selection, outliers influence and cluster instability of traditional K-means algorithm in big data clustering, an MS-Kmeans algorithm based on MapReduce framework was proposed. The algorithm selected multiple non-outlier points as the candidate center points, and mean shifted the candidate center points, used the maximum-minimum principle to select K initial center points from the candidate center points, and then executed the K-means algorithm to find the final center points. In order to improve the running speed of the algorithm for clustering large data sets, the algorithm used MapReduce framework to implement parallel computing on Hadoop platform. The experimental results showed that the parallel MS-Kmeans algorithm is feasible in big data clustering, and the algorithm performs well in terms of performance, speed and stability

show abstract

An Improved K-means Algorithm Based on Multiple Clustering and Density

Cited by 4 publications

References 7 publications

Enhancing K-means Clustering Results with Gradient Boosting: A Post-Processing Approach

Enhancing K-means Clustering Results with Gradient Boosting: A Post-Processing Approach

Sample Density Clustering Method Considering Unbalanced Data Distribution

Parallel MS-Kmeans clustering algorithm based on MapReduce

Contact Info

Product

Resources

About