Towards Refined Classifications Driven by SHAP Explanations

in Holzinger, Andreas; Kieseberg, Peter; Tjoa, A. Min (Eds.) et al Machine Learning and Knowledge Extraction (2022, August 11)

Machine Learning (ML) models are inherently approximate; as a result, the predictions of an ML model can be wrong. In applications where errors can jeopardize a company's reputation, human experts often ... [more ▼]

Machine Learning (ML) models are inherently approximate; as a result, the predictions of an ML model can be wrong. In applications where errors can jeopardize a company's reputation, human experts often have to manually check the alarms raised by the ML models by hand, as wrong or delayed decisions can have a significant business impact. These experts often use interpretable ML tools for the verification of predictions. However, post-prediction verification is also costly. In this paper, we hypothesize that the outputs of interpretable ML tools, such as SHAP explanations, can be exploited by machine learning techniques to improve classifier performance. By doing so, the cost of the post-prediction analysis can be reduced. To confirm our intuition, we conduct several experiments where we use SHAP explanations directly as new features. In particular, by considering nine datasets, we first compare the performance of these "SHAP features" against traditional "base features" on binary classification tasks. Then, we add a second-step classifier relying on SHAP features, with the goal of reducing false-positive and false-negative results of typical classifiers. We show that SHAP explanations used as SHAP features can help to improve classification performance, especially for false-negative reduction. [less ▲]

On the Suitability of SHAP Explanations for Refining Classifications

in In Proceedings of the 14th International Conference on Agents and Artificial Intelligence (ICAART 2022) (2022, February)

In industrial contexts, when an ML model classifies a sample as positive, it raises an alarm, which is subsequently sent to human analysts for verification. Reducing the number of false alarms upstream in ... [more ▼]

In industrial contexts, when an ML model classifies a sample as positive, it raises an alarm, which is subsequently sent to human analysts for verification. Reducing the number of false alarms upstream in an ML pipeline is paramount to reduce the workload of experts while increasing customers’ trust. Increasingly, SHAP Explanations are leveraged to facilitate manual analysis. Because they have been shown to be useful to human analysts in the detection of false positives, we postulate that SHAP Explanations may provide a means to automate false-positive reduction. To confirm our intuition, we evaluate clustering and rules detection metrics with ground truth labels to understand the utility of SHAP Explanations to discriminate false positives from true positives. We show that SHAP Explanations are indeed relevant in discriminating samples and are a relevant candidate to automate ML tasks and help to detect and reduce false-positive results. [less ▲]

A Comparison of Pre-Trained Language Models for Multi-Class Text Classification in the Financial Domain

in Companion Proceedings of the Web Conference 2021 (WWW '21 Companion), April 19--23, 2021, Ljubljana, Slovenia (2021, April 19)

SARS-CoV-2 Transmission in Educational Settings During an Early Summer Epidemic Wave in Luxembourg

E-print/Working paper (2020)

Background: The role of schools and children in the transmission of SARS-CoV-2 remains to be determined. Following a first wave in spring and gradual easing of lockdown, Luxembourg experienced an early ... [more ▼]

Background: The role of schools and children in the transmission of SARS-CoV-2 remains to be determined. Following a first wave in spring and gradual easing of lockdown, Luxembourg experienced an early second epidemic wave before the start of summer school holidays on 15th July. This provided the opportunity to study the role of school-age children and school settings in SARS-CoV-2 transmission. More specifically, we compared the incidence in school-age children, teachers and the general working population, and estimated the number of secondary transmissions occurring at schools using contact tracing data. Findings: While SARS-CoV-2 incidence was much higher in adults aged 20 and above than in children aged 0 to 19 during the first wave in spring, no significant difference was found during the second wave in early summer. The incidence during the second wave was similar for pupils, teachers and the general working population. Based on a total of 424 reported confirmed COVID-19 cases in school-age children and teachers, we estimate that 179 index cases caused 49 secondary transmissions in schools. While some small clusters of mainly student-to-student transmission within the same class were identified, we did not observe any large outbreaks with multiple generations of infection. Interpretation: Transmission of SARS-CoV-2 within Luxembourg schools was limited during the early summer epidemic wave in 2020. Precautionary measures including physical distancing as well as easy access to testing, systematic contact tracing appears to have been successful in mitigating transmission within educational settings. Funding Statement: LV is supported by the Luxembourg National Research Fund grant COVID-19/2020- 1/14701707/REBORN, LM is supported by Luxembourg National Research Fund grant COVID19/14863306/PREVID, PW is supported by the European Research Council (ERC-CoG 863664). Declaration of Interests: No competing interests. Ethics Approval Statement: The Health Directorate has the legal permission to process patient confidential information for national surveillance of communicable diseases in general and contact tracing for the COVID-19 pandemic and individual patient consent is not required. [less ▲]

Challenges Towards Production-Ready Explainable Machine Learning

in Veiber, Lisa; Allix, Kevin; Arslan, Yusuf (Eds.) et al Proceedings of the 2020 USENIX Conference on Operational Machine Learning (OpML 20) (2020, July)

Machine Learning (ML) is increasingly prominent in or- ganizations. While those algorithms can provide near perfect accuracy, their decision-making process remains opaque. In a context of accelerating ... [more ▼]

Machine Learning (ML) is increasingly prominent in or- ganizations. While those algorithms can provide near perfect accuracy, their decision-making process remains opaque. In a context of accelerating regulation in Artificial Intelligence (AI) and deepening user awareness, explainability has become a priority notably in critical healthcare and financial environ- ments. The various frameworks developed often overlook their integration into operational applications as discovered with our industrial partner. In this paper, explainability in ML and its relevance to our industrial partner is presented. We then dis- cuss the main challenges to the integration of ex- plainability frameworks in production we have faced. Finally, we provide recommendations given those challenges. [less ▲]