Distance Metrics and Clustering Methods for Mixed-type Data

doi:10.1111/insr.12274

This content will become publicly available on June 21, 2019

Title: Distance Metrics and Clustering Methods for Mixed-type Data

In spite of the abundance of clustering techniques and algorithms, clustering mixed interval (continuous) and categorical (nominal and/or ordinal) scale data remain a challenging problem. In order to identify the most effective approaches for clustering mixed–type data, we use both theoretical and empirical analyses to present a critical review of the strengths and weaknesses of the methods identified in the literature. Here, the guidelines on approaches to use under different scenarios are provided, along with potential directions for future research.

Authors:

; Markatou, Marianthi ^[1] ; Ray, Bonnie ^[2]

Univ. at Buffalo, Buffalo, NY (United States)
Arenadotio, New York, NY (United States)

Publication Date:: 2018-06-21

Report Number(s):: SAND-2018-7091J
Journal ID: ISSN 0306-7734; 665360

Grant/Contract Number:: AC04-94AL85000

Type:: Accepted Manuscript

Journal Name:: International Statistical Review

Additional Journal Information:: Journal Name: International Statistical Review; Journal ID: ISSN 0306-7734

Research Org:: Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)

Sponsoring Org:: USDOE National Nuclear Security Administration (NNSA)

Country of Publication:: United States

Language:: English

Subject:: 97 MATHEMATICS AND COMPUTING; discretisation; dummy coding; Gower's distance; k-mean clustering; machine learning; Mahalanobis distance; mixture model; multivariate data analysis; unsupervised learning

OSTI Identifier:: 1459931


                    Foss, Alexander H., Markatou, Marianthi, and Ray, Bonnie. Distance Metrics and Clustering Methods for Mixed-type Data.  United States: N. p.,  
        Web.  doi:10.1111/insr.12274.

Copy to clipboard


                    Foss, Alexander H., Markatou, Marianthi, & Ray, Bonnie. Distance Metrics and Clustering Methods for Mixed-type Data.  United States.  doi:10.1111/insr.12274.

Copy to clipboard


                    Foss, Alexander H., Markatou, Marianthi, and Ray, Bonnie. 2018.  
        "Distance Metrics and Clustering Methods for Mixed-type Data".  United States.  
        doi:10.1111/insr.12274.

Copy to clipboard


                    
@article{osti_1459931,

  title        = {Distance Metrics and Clustering Methods for Mixed-type Data},

  author       = {Foss, Alexander H. and Markatou, Marianthi and Ray, Bonnie},

  abstractNote = {In spite of the abundance of clustering techniques and algorithms, clustering mixed interval (continuous) and categorical (nominal and/or ordinal) scale data remain a challenging problem. In order to identify the most effective approaches for clustering mixed–type data, we use both theoretical and empirical analyses to present a critical review of the strengths and weaknesses of the methods identified in the literature. Here, the guidelines on approaches to use under different scenarios are provided, along with potential directions for future research.},

  doi          = {10.1111/insr.12274},

  journal      = {International Statistical Review},

  number       = ,

  volume       = ,

  place        = {United States},

  year         = {2018},

  month        = {6}

}

Copy to clipboard

Free Publicly Accessible Full Text
This content will become publicly available on June 21, 2019
Publisher's Version of Record
10.1111/insr.12274
https://doi.org/10.1111/insr.12274

Works referenced in this record:

Estimating the number of clusters in a data set via the gap statistic
journal, May 2001

Tibshirani, Robert; Walther, Guenther; Hastie, Trevor
Journal of the Royal Statistical Society: Series B (Statistical Methodology), Vol. 63, Issue 2, p. 411-423
DOI: 10.1111/1467-9868.00293

Similar Records

Similar records in DOE PAGES and OSTI.GOV collections:

Comparison of integrated clustering methods for accurate and stable prediction of building energy consumption data

Journal Article Hsu, David September 2015 - Applied Energy

Clustering methods are often used to model energy consumption for two reasons. First, clustering is often used to process data and to improve the predictive accuracy of subsequent energy models. Second, stable clusters that are reproducible with respect to non-essential changes can be used to group, target, and interpret observed subjects. However, it is well known that clustering methods are highly sensitive to the choice of algorithms and variables. This can lead to misleading assessments of predictive accuracy and mis-interpretation of clusters in policymaking. This paper therefore introduces two methods to the modeling of energy consumption in buildings: clusterwise regression,more »« less
Cited by 18Full Text Available
The clustering of galaxies in the completed SDSS-III Baryon Oscillation Spectroscopic Survey: Double-probe measurements from BOSS galaxy clustering and Planck data – towards an analysis without informative priors

Journal Article Pellejero-Ibanez, Marco ; Chuang, Chia -Hsun ; Rubino-Martin, J. A. ; ... March 2016 - Monthly Notices of the Royal Astronomical Society

Here, we develop a new methodology called double-probe analysis with the aim of minimizing informative priors in the estimation of cosmological parameters. We extract the dark-energy-model-independent cosmological constraints from the joint data sets of Baryon Oscillation Spectroscopic Survey (BOSS) galaxy sample and Planck cosmic microwave background (CMB) measurement. We measure the mean values and covariance matrix of {R, l _a, Ω _bh ², n _s, log(A _s), Ω _k, H(z), D _A(z), f(z)σ ₈(z)}, which give an efficient summary of Planck data and 2-point statistics from BOSS galaxy sample, where R = √Ω _mH ² ₀, and l _a =more »« less
Cited by 3Full Text Available
Methods and Challenges in the Determination of Molecular Weight Metrics of Bio-oils

Journal Article Harman-Ware, Anne E. ; Ferrell, Jack R. August 2018 - Energy and Fuels

The analyses of thermochemically-derived bio-oil properties and composition are challenging due to the diversity of compounds present and the reactivity of the oils. There are currently a variety of techniques used and no standard method established for the analysis of the molecular weight distribution, weight average molecular weight (Mw) and other molecular weight metrics of bio-oils. This review focuses on the challenges and variation in methodologies employed for the analysis of bio-oils on the basis of molecular weight, particularly by gel permeation chromatography (GPC). GPC is the most practical means for determination of molecular weight metrics of bio-oils but needsmore »« less
Mixed-Potential NO _x and NH ₃ Sensors Fabricated by Commercial Manufacturing Methods [Mixed-Potential Sensors NO _x and NH ₃ Fabricated by Commercial Manufacturing Methods]

Journal Article Kreller, Cortney R. ; Spernjak, Dusan ; Li, Wenxia ; ... August 2014 - ECS Transactions (Online)

Meeting EPA Tier 3 emissions reduction requirements while simultaneously increasing fuel economy to meet new CAFE standards will require optimization of advanced combustion strategies and emissions control technologies. There is an immediate need for suitable exhaust gas sensor technologies to monitor internal combustion engine tailpipe emissions and to control and maintain efficient operation of the engine and exhaust after treatment systems. NH ₃, NO _x, and HC sensors could enable onboard diagnostics and combustion control in lean-burn engines, analogous to the role of O ₂ sensors in stoichiometric burn engines. Commercial manufacturing methods have been used to fabricate self-heated mixed-potentialmore »« less
Cited by 4Full Text Available
Marker-free registration of forest terrestrial laser scanner data pairs with embedded confidence metrics

Journal Article Van Aardt, Jan ; Romanczyk, Paul ; van Leeuwen, Martin ; ... April 2016 - IEEE Transactions on Geoscience and Remote Sensing

Terrestrial laser scanning (TLS) has emerged as an effective tool for rapid comprehensive measurement of object structure. Registration of TLS data is an important prerequisite to overcome the limitations of occlusion. However, due to the high dissimilarity of point cloud data collected from disparate viewpoints in the forest environment, adequate marker-free registration approaches have not been developed. The majority of studies instead rely on the utilization of artificial tie points (e.g., reflective tooling balls) placed within a scene to aid in coordinate transformation. We present a technique for generating view-invariant feature descriptors that are intrinsic to the point cloud datamore »« less
Cited by 5Full Text Available

Access journal articles and accepted manuscripts resulting from DOE research funding

Title: Distance Metrics and Clustering Methods for Mixed-type Data

Access journal articles and accepted manuscripts
resulting from DOE research funding