Ada-WHIPS: Explaining AdaBoost Classification with Applications in the Health Sciences

Abstract

Background Computer Aided Diagnostics (CAD) can support medical practitioners to make critical decisions about their patients' disease conditions. Practitioners require access to the chain of reasoning behind CAD to build trust in the CAD advice and to supplement their own expertise. Yet, CAD systems might be based on black box machine learning (ML) models and high dimensional data sources (electronic health records, MRI scans, cardiotocograms, etc). These foundations make interpretation and explanation of the CAD advice very challenging. This challenge is recognised throughout the machine learning research community. eXplainable Artificial Intelligence (XAI) is emerging as one of the most important research areas of recent years because it addresses the interpretability and trust concerns of critical decision makers, including those in clinical and medical practice. Methods In this work, we focus on AdaBoost, a black box ML model that has been widely adopted in the CAD literature. We address the challenge -- to explain AdaBoost classification -- with a novel algorithm that extracts simple, logical rules from AdaBoost models. Our algorithm, Adaptive-Weighted High Importance Path Snippets (Ada-WHIPS), makes use of AdaBoost's adaptive classifier weights. Using a novel formulation, Ada-WHIPS uniquely redistributes the weights among individual decision nodes of the internal decision trees (DT) of the AdaBoost model. Then, a simple heuristic search of the weighted nodes finds a single rule that dominated the model's decision. We compare the explanations generated by our novel approach with the state of the art in an experimental study. We evaluate the derived explanations with simple statistical tests of well-known quality measures, precision and coverage, and a novel measure stability that is better suited to the XAI setting.Results Experiments on 9 CAD-related data sets showed that Ada-WHIPS explanations consistently generalise better (mean coverage 15%-68%) than the state of the art while remaining competitive for specificity (mean precision 80%-99%). A very small trade-off in specificity is shown to guard againstover-fitting which is a known problem in the state of the art methods.Conclusions The experimental results demonstrate the benefits of using our novel algorithm for explaining CAD AdaBoost classifiers widely found in the literature. Our tightly coupled, AdaBoost-specific approach outperforms model-agnostic explanation methods and should be considered by practitioners looking for an XAI solution for this class of models.

Establishment and validation of a torsade de pointes prediction model based on human iPSC‑derived cardiomyocytes

“…Considering that a precise classification is difficult to achieve with only one model, the combination of multiple models was used in the current study. Thus, LR and AdaBoost (a black box model widely used in computer-aided diagnosis), were chosen as the sub-models according to the classification performance of each model (45).…”

Explainable AI for clinical and remote health applications: a survey on tabular and time series data

“…Results obtained by analysing the percentage of clinical adherence confirmed that almost all the rules discovered by the model are also the most accepted statements among experts. Hatwell et al (2020) proposed a novel approach to explain the AdaBoost (AB) classification algorithm, a black-box model widely used in the CAD literature. The new algorithm, named Adaptive-Weighted High Importance Path Snippets (Ada-WHIPS), has been designed to provide accurate and highly-interpretable disease detection by extracting simple and logical rules from the AB model.…”

Ethnicity influences phenotype and clinical outcomes: Comparing a South American with a North American inflammatory bowel disease cohort

“…[14] To further confirm the RF results, we compared the results of several supervised ML models without tuning (super vector machine, logistic regression, AdaBoost, multilayer perceptron (MLP), and linear discriminant analysis). [12,[15][16][17][18] The data were split into 2 groups, 80% for training and 20% for testing, as in our RF. Additionally, we developed a generalized linear model to predict the cohort (Chilean or North American) using R (version 4.1.1; 2021-08-10, RStudio, Boston, MA) and dplyr, mlogit, tidyr libraries.…”