How to Open a Black Box Classifier for Tabular Data

被引:6
|
作者
Walters, Bradley [1 ]
Ortega-Martorell, Sandra [1 ]
Olier, Ivan [1 ]
Lisboa, Paulo J. G. [1 ]
机构
[1] Liverpool John Moores Univ, Sch Comp Sci & Math, Liverpool L3 2AF, England
关键词
ANOVA; Shapley values; self-explaining neural networks; generalised additive models; interpretability;
D O I
10.3390/a16040181
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
A lack of transparency in machine learning models can limit their application. We show that analysis of variance (ANOVA) methods extract interpretable predictive models from them. This is possible because ANOVA decompositions represent multivariate functions as sums of functions of fewer variables. Retaining the terms in the ANOVA summation involving functions of only one or two variables provides an efficient method to open black box classifiers. The proposed method builds generalised additive models (GAMs) by application of L1 regularised logistic regression to the component terms retained from the ANOVA decomposition of the logit function. The resulting GAMs are derived using two alternative measures, Dirac and Lebesgue. Both measures produce functions that are smooth and consistent. The term partial responses in structured models (PRiSM) describes the family of models that are derived from black box classifiers by application of ANOVA decompositions. We demonstrate their interpretability and performance for the multilayer perceptron, support vector machines and gradient-boosting machines applied to synthetic data and several real-world data sets, namely Pima Diabetes, German Credit Card, and Statlog Shuttle from the UCI repository. The GAMs are shown to be compliant with the basic principles of a formal framework for interpretability.
引用
收藏
页数:26
相关论文
共 50 条
  • [41] BIOCHEMISTRY Enzyme's black box cracked open
    Sherman, David H.
    NATURE, 2009, 461 (7267) : 1068 - 1069
  • [42] Force Open: Lightweight black box file repair
    Wüst K.
    Tsankov P.
    Radomirović S.
    Torabi Dashti M.
    Digital Investigation, 2017, 20 : S75 - S82
  • [43] Cracking Open the Black Box of Genetic Ancestry Testing
    Donovan, Joan
    Pasquetto, Irene
    Pierre, Jennifer
    PROCEEDINGS OF THE 51ST ANNUAL HAWAII INTERNATIONAL CONFERENCE ON SYSTEM SCIENCES (HICSS), 2018, : 1731 - 1740
  • [44] The immunobiology of the mammalian epididymis: The black box is now open!
    Guiton R.
    Henry-Berger J.
    Drevet J.R.
    Basic and Clinical Andrology, 23 (1)
  • [45] Model risk - daring to open up the black box
    Aggarwal, A.
    Beck, M. B.
    Cann, M.
    Ford, T.
    Georgescu, D.
    Morjaria, N.
    Smith, A.
    Taylor, Y.
    Tsanakas, A.
    Witts, L.
    Ye, I.
    BRITISH ACTUARIAL JOURNAL, 2016, 21 (02) : 229 - 296
  • [46] The events of early pregnancy: prying open the black box
    Jukic, A. M.
    Wilcox, A. J.
    ULTRASOUND IN OBSTETRICS & GYNECOLOGY, 2012, 40 (06) : 617 - 618
  • [47] IS ERP THE RIGHT KEY TO OPEN THE BLACK-BOX
    KARMOS, G
    CSEPE, V
    BEHAVIORAL AND BRAIN SCIENCES, 1990, 13 (02) : 245 - 245
  • [48] Open tools for quantitative anonymization of tabular phenotype data: literature review
    Haber, Anna C.
    Sax, Ulrich
    Prasser, Fabian
    BRIEFINGS IN BIOINFORMATICS, 2022, 23 (06)
  • [49] Low-cost and efficient prediction hardware for tabular data using tiny classifier circuits
    Iordanou, Konstantinos
    Atkinson, Timothy
    Ozer, Emre
    Kufel, Jedrzej
    Aligada, Grace
    Biggs, John
    Brown, Gavin
    Lujan, Mikel
    NATURE ELECTRONICS, 2024, 7 (05) : 405 - 413