A Method to Find Optimum Number of Clusters Based on Fuzzy Silhouette on Dynamic Data Set

Subbalakshmi, Chatti; Krishna, G. Rama; Rao, S. Krishna Mohan; Pulipati, Venkateswara Rao

doi:10.1016/j.procs.2015.02.030

Cited by 61 publications

(33 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One of the main problems of the k-Means method is how to determine the optimal number of clusters k. Research by Subbalakshmi et al [15] have shown that the accuracy of the k-Means method can be higher, if appropriate in selecting the initial value and number of clusters [2,13].…”

Section: K-means Methodsmentioning

confidence: 99%

Clustering optimization in RFM analysis Based on k-Means

Gustriansyah

Suhandi

Antony

2020

IJEECS

View full text Add to dashboard Cite

RFM stands for Recency, Frequency, and Monetary. RFM is a simple but effective method that can be applied to market segmentation. RFM analysis is used to analyze customer’s behavior which consists of how recently the customers have purchased (recency), how often customer’s purchases (frequency), and how much money customers spend (monetary). In this study, RFM analysis has been used for product segmentation is to be arrayed in terms of recent sales (R), frequent sales (F), and the total money spent (M) using the data mining method. This study has proposed a new procedure for RFM analysis (in product segmentation) using the k-Means method and eight indexes of validity to determine the optimal number of clusters namely Elbow Method, Silhouette Index, Calinski-Harabasz Index, Davies-Bouldin Index, Ratkowski Index, Hubert Index, Ball-Hall Index, and Krzanowski-Lai Index, which can improve the objectivity and similarity of data in product segmentation so that it can improve the accuracy of the stock management process. The evaluation results showed that the optimal number of clusters for the k-Means method applied in the RFM analysis consists of three clusters (segmentation) with a variance value of 0.19113.

show abstract

Section: K-means Methodsmentioning

confidence: 99%

Clustering optimization in RFM analysis Based on k-Means

Gustriansyah

Suhandi

Antony

2020

IJEECS

View full text Add to dashboard Cite

show abstract

“…There are many unsupervised statistical methods that have been used to determine an ideal number of clusters, such as Bayesian statistics (Senthilnath et al 2017) and the Hierarchical method (Chen et al 2005, Corstanje et al 2016, Grafius et al 2018). The elbow and silhouette methods (Subbalakshmi et al 2015, Rial et al 2017, Scharsich et al 2017, Wang et al 2017 a , b , Pascucci et al 2018) were used in this study because of their ease of interpretation and reasonable processing time.…”

Section: Methodsmentioning

confidence: 99%

“…The silhouette method provides a graphical representation of how well observations belong to natural clusters in the data (Subbalakshmi et al 2015). This method calculates the difference between the minimum Euclidean distance between cluster centroid and observation and the average distance within the cluster (Wang et al 2017).…”

Section: Methodsmentioning

confidence: 99%

“…A hierarchical and nested clustering tree is constructed by calculating the similarity between the different clusters (Wang et al 2017). An overall clustering quality is given between 1 and −1 for good clustering and poor clustering solutions, respectively (Subbalakshmi et al 2015). The clustering quality is based on similarity (shorter Euclidean distances between observation and cluster centroid) and dissimilarity (larger nearest neighbor distances between clusters; Pascucci et al 2018).…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Environmental controls on Landsat‐derived phenoregions across an East African megatransect

2020

View full text Add to dashboard Cite

Semiarid and savanna‐type (SAST) ecosystems in East Africa have unique plant species compositions and characteristics that make quantifying this biome's seasonality and interannual variability difficult. Phenoregion classification offers a way to use seasonality of vegetation growth to help understand the phenological spatial patterns of complex landscapes. Here, we used Normalized Difference Vegetation Index (NDVI) time series from Landsat 8 to map phenoregions in scenes centered around national parks from Mt. Kenya National Park (Kenya) to Limpopo National Park (Mozambique). We then assessed whether landscape‐scale controls on phenology are consistent across the region or whether they vary across this latitudinal gradient. We compared our phenoregion maps to MODIS Land Cover and geology, and we used multinomial logistic regression to determine the role that elevation, slope, aspect, and geology play in driving phenological differences. The sites' phenoregions showed no unique land cover composition, suggesting that MODIS land cover does not capture the subtle variations identified in phenoregion analysis. Multinomial logistic regression showed that geographic trend (x‐ and y‐directions) was a strong predictor in four of the five landscapes and that, depending on the scene, geology, elevation, or aspect was a strong secondary predictor. Using seasonality of the NDVI time series to generate phenoregions provides different and, in some cases, more ecologically relevant information, compared to vegetation maps that use only land cover from a single season or time period to generate ecoregions. With a significant population of humans and animals that live in and depend on SAST ecosystems, it is important to better understand vegetation processes and factors that affect them as climate change becomes an increasingly pertinent issue in dry systems.

show abstract

“…IC assesses the quality of clustering solution by reducing the complexity of the modular architecture. The Silhouette (Sil) index is also a quantitative evaluation for clustering validity and can recommend the 'appropriate' number of clusters [18,42]. Moreover, the DSM direct clustering [43] is a traditional DSM partitioning algorithm to acquire the clustering result with elements' rearrangement.…”

Section: Research On a Benchmark Problemmentioning

confidence: 99%

Determining optimal granularity level of modular product with hierarchical clustering and modularity assessment

Wang

Yin

2019

J Braz. Soc. Mech. Sci. Eng.

View full text Add to dashboard Cite

Modular product architecture is beneficial for the product maintenance, upgrade, components' concurrent design through team work, and achieving the mass customization eventually. Recent researches tend to group elements into modules as a flat map, but this is inconsistent with the nested composition in product final assembly. Finding the hierarchical modular partition for elements of a product and obtaining its optimal granularity level are still necessary to ease the partition and combination for sub-design tasks. We use the design structure matrix (DSM) to represent the relationships among elements of a product. The hierarchical clustering functions such as pdist and linkage within MATLAB software are utilized to form the hierarchical dendrogram for DSM elements, and the modularity index Q is applied to assess the modular clustering results. The partitions are acquired by various distance threshold values in the vertical axis of the hierarchical dendrogram. The 'cityblock' and 'average' are found as preferred parameters for pdist and linkage functions, respectively, from loop test experiments. Combining the modularity formula in a research literature, a hierarchical modular architecture design methodology is proposed in this work to determine the optimal granularity level of a modular product and to provide solutions for designer selection. Case studies and comparisons in a benchmark problem and a concrete spraying machine illustrate the effectiveness and efficiency of the proposed method. It is characterized by being able to find stable modular partition results for both float and binary DSM elements and satisfying the recommended modular number rule simultaneously.

show abstract

A Method to Find Optimum Number of Clusters Based on Fuzzy Silhouette on Dynamic Data Set

Cited by 61 publications

References 7 publications

Clustering optimization in RFM analysis Based on k-Means

Clustering optimization in RFM analysis Based on k-Means

Environmental controls on Landsat‐derived phenoregions across an East African megatransect

Determining optimal granularity level of modular product with hierarchical clustering and modularity assessment

Contact Info

Product

Resources

About