Srikumar Krishnamoorthy
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Srikumar Krishnamoorthy has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
Avdhoot Kishore Saple
Infosys Technologies Ltd, Bangalore, India
1 shared publicationKishore Varma Indukuri
SETLabs, Infosys Technologies Limited
1 shared publicationManish Kumar
Purdue University, West Lafayette, IN
1 shared publicationNilesh Ranade
Infosys Technologies Limited
1 shared publicationP. Radha Krishna
SETLabs, Infosys Technologies Limited
1 shared publicationPrahalad Haldhoderi Achutharao
Infosys Technologies Ltd, Bangalore, India
1 shared publication
Latest Publications
- Open Access
- 27 May 2026
- Published by Elsevier in SSRN Electronic Journal
Interpretability Myopia: Governance Fitness in Financial Risk Models
Interpretable machine learning models are routinely favored for high-stakes deployment on the basis of predictive accuracy and architectural simplicity. This implicitly presumes that legibility alone suffices for trustworthy decision support. We argue this presumption is systematically myopic. Computational transparency does not guarantee that a model is expressive enough to encode the risk structures that governed decision environments demand. Legibility and governance fitness are orthogonal properties. Conflating them has yielded models that are transparent about what they capture yet structurally incapable of representing what they miss. We formalize this critique through a three-tier evaluation framework applied to six classifier families across four financial benchmark datasets. The first tier assesses discrimination, operational utility under stochastic portfolio simulation, and three distinct dimensions of stability. These are structural stability of learned representations, functional stability of predictive outputs, and individual group membership stability across resampled conditions. Existing work overwhelmingly restricts attention to functional stability, ignoring whether model architecture and group-level assignments persist under perturbation. The second tier profiles deployment suitability across robustness, parsimony, reproducibility, calibration, scenario resilience, and decision-policy adaptability. The third constructs a governance stress test probing six structural failure classes: subgroup heterogeneity, risk amplification, proxy discrimination, trajectory asymmetry, conditional non-monotonicity, and protective factor failure. Bootstrap-validated experiments confirm that marginal-additive and recursive-partitioning architectures exhibit these deficiencies as inherent consequences of their design. No single model dominates all three tiers. Ensembles lead on utility, linear models on stability, and pattern-based methods uniquely expose governance failures as auditable, first-class outputs. Governance fitness remains a critical yet underdeveloped axis of model selection in regulated domains. Ensembles engineered for structural coverage consistently outperform individual classifiers in meeting this requirement.
- 27 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessInterpretability Myopia: Governance Fitness in Financial Risk Models
Interpretable machine learning models are routinely favored for high-stakes deployment on the basis of predictive accuracy and architectural simplicity. This implicitly presumes that legibility alone suffices for trustworthy decision support. We argue this presumption is systematically myopic. Computational transparency does not guarantee that a model is expressive enough to encode the risk structures that governed decision environments demand. Legibility and governance fitness are orthogonal properties. Conflating them has yielded models that are transparent about what they capture yet structurally incapable of representing what they miss. We formalize this critique through a three-tier evaluation framework applied to six classifier families across four financial benchmark datasets. The first tier assesses discrimination, operational utility under stochastic portfolio simulation, and three distinct dimensions of stability. These are structural stability of learned representations, functional stability of predictive outputs, and individual group membership stability across resampled conditions. Existing work overwhelmingly restricts attention to functional stability, ignoring whether model architecture and group-level assignments persist under perturbation. The second tier profiles deployment suitability across robustness, parsimony, reproducibility, calibration, scenario resilience, and decision-policy adaptability. The third constructs a governance stress test probing six structural failure classes: subgroup heterogeneity, risk amplification, proxy discrimination, trajectory asymmetry, conditional non-monotonicity, and protective factor failure. Bootstrap-validated experiments confirm that marginal-additive and recursive-partitioning architectures exhibit these deficiencies as inherent consequences of their design. No single model dominates all three tiers. Ensembles lead on utility, linear models on stability, and pattern-based methods uniquely expose governance failures as auditable, first-class outputs. Governance fitness remains a critical yet underdeveloped axis of model selection in regulated domains. Ensembles engineered for structural coverage consistently outperform individual classifiers in meeting this requirement.
- Open Access
- 6 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 126088-126107
Interpretable Classifier Models for Decision Support Using High Utility Gain Patterns
Ensemble models such as gradient boosting and random forests are proven to offer the best predictive performance on a wide variety of supervised learning problems. The high performance of these black box models, however, comes at a cost of model interpretability. They are also inadequate to meet regulatory demands and explainability needs of organizations. The model interpretability in high performance black-box models is achieved with the help of posthoc explainable models such as Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP). This paper presents an alternate intrinsic classifier model that extracts a class of higher order patterns and embeds them into an interpretable learning model. More specifically, the proposed model extracts novel High Utility Gain (HUG) patterns that capture higher order interactions, transforms the model input data into a new space, and applies interpretable classifier methods on the transformed space. We conduct rigorous experiments on forty benchmark binary and multi-class classification datasets to evaluate the proposed model against the state-of-the-art ensemble and interpretable classifier models. The proposed model was comprehensively assessed on three key dimensions: (1) quality of predictions using classifier measures such as accuracy, F 1 , AUC , H-measure, and logistic loss, (2) computational performance on large and high-dimensional data, and (3) interpretability aspects. The HUG-based learning model was found to deliver performance comparable to that of the state-of-the-art ensemble models. Our model was also found to achieve 2-40% (45%) prediction quality (interpretability) improvements with significantly lower computational requirements over other interpretable classifier models. Furthermore, we present case studies in finance and healthcare domains and generate oneand two-dimensional HUG profiles to illustrate the interpretability aspects of our HUG models. The proposed solution offers an alternate approach to build high performance and transparent machine learning classifier models. We hope that our ML solution help organizations meet their growing regulatory and explainability needs.
- 6 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 126088-126107
Open AccessInterpretable Classifier Models for Decision Support Using High Utility Gain Patterns
Ensemble models such as gradient boosting and random forests are proven to offer the best predictive performance on a wide variety of supervised learning problems. The high performance of these black box models, however, comes at a cost of model interpretability. They are also inadequate to meet regulatory demands and explainability needs of organizations. The model interpretability in high performance black-box models is achieved with the help of posthoc explainable models such as Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP). This paper presents an alternate intrinsic classifier model that extracts a class of higher order patterns and embeds them into an interpretable learning model. More specifically, the proposed model extracts novel High Utility Gain (HUG) patterns that capture higher order interactions, transforms the model input data into a new space, and applies interpretable classifier methods on the transformed space. We conduct rigorous experiments on forty benchmark binary and multi-class classification datasets to evaluate the proposed model against the state-of-the-art ensemble and interpretable classifier models. The proposed model was comprehensively assessed on three key dimensions: (1) quality of predictions using classifier measures such as accuracy, F 1 , AUC , H-measure, and logistic loss, (2) computational performance on large and high-dimensional data, and (3) interpretability aspects. The HUG-based learning model was found to deliver performance comparable to that of the state-of-the-art ensemble models. Our model was also found to achieve 2-40% (45%) prediction quality (interpretability) improvements with significantly lower computational requirements over other interpretable classifier models. Furthermore, we present case studies in finance and healthcare domains and generate oneand two-dimensional HUG profiles to illustrate the interpretability aspects of our HUG models. The proposed solution offers an alternate approach to build high performance and transparent machine learning classifier models. We hope that our ML solution help organizations meet their growing regulatory and explainability needs.
- 17 August 2022
- Published by Elsevier in Computers & Operations Research
- Vol. 148
A two-stage integer programming model considering transaction equivalence for privacy preservation
- 17 August 2022
- Published by Elsevier in Computers & Operations Research
- Vol. 148
A two-stage integer programming model considering transaction equivalence for privacy preservation
- 1 March 2019
- Published by Elsevier in Expert Systems with Applications
- Vol. 117, 148-165
Mining top-k high utility itemsets with effective threshold raising strategies
- 1 March 2019
- Published by Elsevier in Expert Systems with Applications
- Vol. 117, 148-165
Mining top-k high utility itemsets with effective threshold raising strategies
- 19 January 2019
- Published by Springer Nature
- p. 47-74
A Comparative Study of Top-K High Utility Itemset Mining Methods
- 19 January 2019
- Published by Springer Nature
- p. 47-74
A Comparative Study of Top-K High Utility Itemset Mining Methods
- 1 April 2018
- Published by Elsevier in Knowledge-Based Systems
- Vol. 145, 1-14
Efficiently mining high utility itemsets with negative unit profits
- 1 April 2018
- Published by Elsevier in Knowledge-Based Systems
- Vol. 145, 1-14
Efficiently mining high utility itemsets with negative unit profits
- Top Cited Papers
- 1 December 2017
- Published by Elsevier in Expert Systems with Applications
- Vol. 90, 168-183
HMiner: Efficiently mining high utility itemsets
- 1 December 2017
- Published by Elsevier in Expert Systems with Applications
- Vol. 90, 168-183
Top Cited PapersHMiner: Efficiently mining high utility itemsets
- 24 November 2017
- Published by Springer Nature in Knowledge and Information Systems
- Vol. 56 (2) , 373-394
Sentiment analysis of financial news articles using performance indicators
- 24 November 2017
- Published by Springer Nature in Knowledge and Information Systems
- Vol. 56 (2) , 373-394
Sentiment analysis of financial news articles using performance indicators
- 1 May 2015
- Published by Elsevier in Expert Systems with Applications
- Vol. 42 (7) , 3751-3759
Linguistic features for review helpfulness prediction
- 1 May 2015
- Published by Elsevier in Expert Systems with Applications
- Vol. 42 (7) , 3751-3759
Linguistic features for review helpfulness prediction
- 1 April 2015
- Published by Elsevier in Expert Systems with Applications
- Vol. 42 (5) , 2371-2381
Pruning strategies for mining high utility itemsets
- 1 April 2015
- Published by Elsevier in Expert Systems with Applications
- Vol. 42 (5) , 2371-2381
Pruning strategies for mining high utility itemsets