Evaluating Tree Explanation Methods for Anomaly Reasoning: A Case Study of SHAP TreeExplainer and TreeInterpreter

Sharma, Pulkit; Mirzan, Shezan Rohinton; Bhandari, Apurva; Pimpley, Anish; Eswaran, Abhiram; Srinivasan, Soundar; Shao, Liqun

doi:10.1007/978-3-030-65847-2_4

Cited by 15 publications

(12 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The SHAP value method is a means of interpreting the output of a given machine learning model. Moreover, SHAP can ensure consistency and, when applied to a particular element, can be quantified as the mean of all the significant contributions to the prediction across all possible permutations of the elements …”

Section: Methodsmentioning

confidence: 99%

“…Moreover, SHAP can ensure consistency and, when applied to a particular element, can be quantified as the mean of all the significant contributions to the prediction across all possible permutations of the elements. 52 SHAP is implemented as a Python toolkit for interpretable machine learning and provides a set of Shapley values for every feature in the data set. 53 SHAP determines the effect of each of the input features in terms of positive and negative contribution with a SHAP value for each observation.…”

Section: Shapley Additive Explanations (Shaps)mentioning

confidence: 99%

See 1 more Smart Citation

Optimizing Photodegradation Rate Prediction of Organic Contaminants: Models with Fine-Tuned Hyperparameters and SHAP Feature Analysis for Informed Decision Making

Schossler,

Ojo,

2023

ACS EST Water

View full text Add to dashboard Cite

Titanium dioxide (TiO 2 ) is commonly used as a photocatalyst for treating water contaminants. However, experimentally determining the photodegradation kinetics is both expensive and time-consuming. In this study, the authors developed gradient booster regressor (GBR) and extreme gradient boosted (XGBoost) decision tree models using the BayesSearchCV and hyperopt libraries, respectively, to fine-tune the hyperparameters. The goal was to estimate photodegradation rate constants for organic contaminants and investigate the relationship between input variables and their characteristics to enhance model performance. To assess the models' predictive capabilities, the researchers employed quantitative approaches such as the coefficient of determination (R 2 ), mean absolute error (MAE), and root mean squared error (RMSE). Additionally, they utilized the SHAP technique, which applies cooperative game theory to estimate photodegradation rates and rank the importance of features, to provide interpretability to the results. The study identified enhanced models that outperformed other models in the literature, with the XGBoost model combined with the TPE algorithm performing the best, achieving an R 2 of 0.891, an MAE of 0.100, and an RMSE of 0.157. Furthermore, the SHAP technique provided insights into the features predicting photodegradation rates and their correlations.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Shapley Additive Explanations (Shaps)mentioning

confidence: 99%

Optimizing Photodegradation Rate Prediction of Organic Contaminants: Models with Fine-Tuned Hyperparameters and SHAP Feature Analysis for Informed Decision Making

Schossler,

Ojo,

2023

ACS EST Water

View full text Add to dashboard Cite

show abstract

“…In this section, we describe a method in which local SHAP feature importance can be verified. This method as suggested by Sharma et al 32 to perturb a random attribute by a small percentage and find the absolute difference between the SHAP values of the attributes of each example before and after perturbation. The highest absolute difference must be in the perturbed attribute.…”

Section: Validation Of Shap Methodologymentioning

confidence: 99%

Development of Gradient Boosting Machines for Estimation of Total and Dynamic Liquid Holdup in Trickle Bed Reactor

Bhosale

Pandya

Chatterjee

et al. 2023

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

Prediction of liquid holdup is of significance in designing and in evaluating the performance of trickle bed contactors. The present work focuses on the development of Gradient Boosting Machines (GBM) for the prediction of total and dynamic liquid holdup in trickle bed reactors. A comprehensive data set of 394 data points of total liquid holdup and 416 data points of dynamic liquid holdup curated from open literature is used in this study. We built GBM models with the input data sets containing 11 governing variables. GBM provided excellent predictions for both data sets. We have also compared the GBM predictions with that of the Random Forest (RF) and Artificial Neural Networks (ANN) predictions. As GBM provided the best performance, we further employed SHAP (SHapley Additive exPlanations) with GBM black box models to get local and global interpretability. Also, we have used SHAP to identify informative subsets of governing variables. The work shall pave the way for use of GBM in prediction of hydrodynamic parameters in multiphase systems.

show abstract

“…• Boruta: evaluates the importance of each variable with respect to a permuted version of it to determine its relevance (Kursa et al, 2010). • Shap RFE: is a modified version of RFE that identifies more robustly the importance of each variable using SHAP (see Section 2.6) (Lundberg et al, 2019;Sharma et al, 2020).…”

Section: Step 5: Model Improvementmentioning

confidence: 99%

“…We propose using SHAP (SHapley Additive exPlanations) (Lundberg et al, 2019), because this method allows us to estimate the contribution of each variable to individual predictions in a robust, consistent, and locally accurate way (Lundberg et al, 2018). It uses an optimized procedure for tree-based algorithms allowing interpreting and debugging the resultant model (Sharma et al, 2020;Yoshida, 2020).…”

Section: Step 6: Explaining the Modelmentioning

confidence: 99%

A methodology to design, develop, and evaluate machine learning models for predicting dropout in school systems: the case of Chile

et al. 2023

View full text Add to dashboard Cite

School dropout is a structural problem which permanently penalizes students and society in areas such as low qualification jobs, higher poverty levels and lower life expectancy, lower pensions, and higher economic burden for governments. Given these high consequences and the surge of the problem due to COVID-19 pandemic, in this paper we propose a methodology to design, develop, and evaluate a machine learning model for predicting dropout in school systems. In this methodology, we introduce necessary steps to develop a robust model to estimate the individual risk of each student to drop out of school. As advancement from previous research, this proposal focuses on analyzing individual trajectories of students, incorporating the student situation at school, family, among other levels, changes, and accumulation of events to predict dropout. Following the methodology, we create a model for the Chilean case based on data available mostly through administrative data from the educational system, and according to known factors associated with school dropout. Our results are better than those from previous research with a relevant sample size, with a predictive capability 20% higher for the actual dropout cases. Also, in contrast to previous work, the including non-individual dimensions results in a substantive contribution to the prediction of leaving school. We also illustrate applications of the model for Chilean case to support public policy decision making such as profiling schools for qualitative studies of pedagogic practices, profiling students’ dropout trajectories and simulating scenarios.

show abstract

Evaluating Tree Explanation Methods for Anomaly Reasoning: A Case Study of SHAP TreeExplainer and TreeInterpreter

Cited by 15 publications

References 14 publications

Optimizing Photodegradation Rate Prediction of Organic Contaminants: Models with Fine-Tuned Hyperparameters and SHAP Feature Analysis for Informed Decision Making

Optimizing Photodegradation Rate Prediction of Organic Contaminants: Models with Fine-Tuned Hyperparameters and SHAP Feature Analysis for Informed Decision Making

Development of Gradient Boosting Machines for Estimation of Total and Dynamic Liquid Holdup in Trickle Bed Reactor

A methodology to design, develop, and evaluate machine learning models for predicting dropout in school systems: the case of Chile

Contact Info

Product

Resources

About