CRAFTER: A Tree-Ensemble Clustering Algorithm for Static Datasets with Mixed Attributes and High Dimensionality

Lin, Sangdi; Azarnoush, Bahareh; Runger, George C.

doi:10.1109/tkde.2018.2807444

Cited by 12 publications

(7 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Lin et al [101] present a tree-ensembles clustering algorithm, CRAFTER, for clustering high-dimensional mixed datasets. First, a random subset of data points is drawn and the random forests clustering algorithm [162] is applied.…”

Section: E Othermentioning

confidence: 99%

Survey of State-of-the-Art Mixed Data Clustering Algorithms

2019

View full text Add to dashboard Cite

Mixed data comprises both numeric and categorical features, and mixed datasets occur frequently in many domains, such as health, finance, and marketing. Clustering is often applied to mixed datasets to find structures and to group similar objects for further analysis. However, clustering mixed data are challenging because it is difficult to directly apply mathematical operations, such as summation or averaging, to the feature values of these datasets. In this paper, we present a taxonomy for the study of mixed data clustering algorithms by identifying five major research themes. We then present the state-of-the-art review of the research works within each research theme. We analyze the strengths and weaknesses of these methods with pointers for future research directions. At last, we present an in-depth analysis of the overall challenges in this field, highlight open research questions, and discuss guidelines to make progress in the field. INDEX TERMS Categorical features, clustering, mixed datasets, numeric features. I. INTRODUCTION Clustering is an unsupervised machine learning technique used to group unlabeled data into clusters that contain data points that are 'similar' to each other and 'dissimilar' from those in other clusters [1], [2]. Many clustering algorithms can only handle data that contain either numeric or categorical feature values [3], [4]. Numeric features can take real values, such as height, weight, and distance. Categorical features represent data that can be divided into a fixed number of categories, such as color, race, sex, profession, and blood group. Clustering algorithms group data points into clusters using some notion of 'similarity', which can be as simple as the Euclidean distance. To compute the similarity between numeric feature values, mathematical operations (such as distances, angles, summation, or mean) are applied to them. Distance-based similarity measures are mostly used for numeric data points. Generally, categorical feature values are not inherently ordered (for example, the categorical values, red and blue). It is not possible to directly compute the distance between two categorical feature values. Therefore, computing distance-based similarity measures for categorical data is a challenging task [5]. Nevertheless, several methods The associate editor coordinating the review of this manuscript and approving it for publication was Haruna Chiroma.

show abstract

Section: E Othermentioning

confidence: 99%

Survey of State-of-the-Art Mixed Data Clustering Algorithms

2019

View full text Add to dashboard Cite

show abstract

“…Clustering is the method of categorising data into groups or clusters such that objects within a cluster have a high degree of similarity to one another but are quite different from objects in other clusters. The term cluster analysis itself encompasses a number of different algorithms and methods (Tree Clustering [Lin et al (2018), Liu et al (2005), Freeman (2006), Ahmed et al (2011), Lv et al (2018b), Freeman (2007), WANG et al (2009), Buttrey & Whitaker (2015), Qiu & Li (2021), Jothi et al (2015), Page (1974), Vathy-Fogarassy et al (2005), Miller & Rose (1994)], Block Clustering, k-Means Clustering Wilkin & Huang (2007) and EM algorithms) for grouping objects of similar kind into respective categories, graph-based clustering (Bai et al (2017)), hierarchical clustering (Köhn & Hubert (2014)), model-based clustering (Fraley & Raftery (1998), Fraley & Raftery (1999), Fraley & Raftery (2002)); Lloyd's K-means clustering and the progressive greedy K-means clustering (Wilkin & Huang (2007)).…”

Section: On Clustering and The Random Potts Modelsmentioning

confidence: 99%

Potts Clustering with Complete Shrinkage: a novel clustering methodology with its Software Package pottscompleteshrinkage in Python

Alahassa¹

2021

Preprint

View full text Add to dashboard Cite

We have modified the Potts Model Swendsen-Wang algorithm to insert some clusters constraints by applying a modified agglomerative clustering approach (Kurita, 1991). We have called the induced Potts Model, the Potts Clustering with Complete Shrinkage (PCCS), under the Python package pottscompleteshrinkage deployed on PyPi Index under its current release. In this approach, we deal with the increasing number of small clusters generated in a given partition by merging all small clusters of size ≤ h with their closest cluster in terms of minimal distance respectively, where h is an integer greater or equal to 2. The algorithm uses a technique in which distances of all pairs of observations are stored. Then the nearest cluster (with size ≥ h) is given by the cluster with the closest node in terms of minimal distance to the cluster to be merged using complete linkage. This approach is truly effective as it helps to control the clusters size, and we have found empirical evidence of Chi-Square and Gamma density curves for the constrained cluster size distribution of PCCS, when applied to some datasets taken from the multiple-output benchmark datasets available in the Mulan project website (Tsoumakas et al., 2020). We add a last framework based on Frequency of frequency distribution (FoF) to find the conditional bonds distribution given the clusters size constraints which results in an intractable distribution for large datasets, but its computation framework is a land of rich mathematical developments.

show abstract

“…The performance of KMCMD algorithm with initKmix algorithm was also compared with different state of the art initialization methods; Wu's initialization [41], Cao's initialization [14], Khan and Ahmad's initialization [28] and Ini Entropy [31]. KMCMD with initKmix algorithm is also compared with recently published CRAFTER [30] algorithm. The results of various clustering methods are presented in Table 4.…”

Section: Categorical Datasetsmentioning

confidence: 99%

“…The results of various clustering methods are presented in Table 4. The results for the other clustering algorithms were taken from the papers [28,30,31]. Except for Soybean-small dataset, KMCMD algorithm with initKmix outperformed other clustering methods.…”

Section: Categorical Datasetsmentioning

confidence: 99%

initKmix -- A Novel Initial Partition Generation Algorithm for Clustering Mixed Data using k-means-based Clustering

Ahmad,

Khan

2019

Preprint

View full text Add to dashboard Cite

Mixed datasets consist of both numeric and categorical attributes. Various K-means-based clustering algorithms have been developed to cluster these datasets. Generally, these algorithms use random partition as a starting point, which tend to produce different clustering results in different runs. This inconsistency of clustering results may lead to unreliable inferences from the data. A few initialization algorithms have been developed to compute initial partition for mixed datasets; however, they are either computationally expensive or they do not produce consistent clustering results in different runs. In this paper, we propose, initKmix, a novel approach to find initial partition for K-means-based clustering algorithms for mixed datasets. The initKmix is based on the experimental observations that (i) some data points in a dataset remain in the same clusters created by k-means-based clustering algorithm irrespective of the choice of initial clusters, and (ii) individual attribute information can be used to create initial clusters. In initKmix method, a k-means-based clustering algorithm is run many times, in each run one of the attribute is used to produce initial partition. The clustering results of various runs are combined to produce initial partition. This initial partition is then be used as a seed to a k-means-based clustering algorithm to cluster mixed data. The initial partitions produced by initKmix are always fixed, do not change over different runs or by changing the order of the data objects. Experiments with various categorical and mixed datasets showed that initKmix produced accurate and consistent results, and outperformed random initialization and other state-

show abstract

CRAFTER: A Tree-Ensemble Clustering Algorithm for Static Datasets with Mixed Attributes and High Dimensionality

Cited by 12 publications

References 19 publications

Survey of State-of-the-Art Mixed Data Clustering Algorithms

Survey of State-of-the-Art Mixed Data Clustering Algorithms

Potts Clustering with Complete Shrinkage: a novel clustering methodology with its Software Package pottscompleteshrinkage in Python

initKmix -- A Novel Initial Partition Generation Algorithm for Clustering Mixed Data using k-means-based Clustering

Contact Info

Product

Resources

About