Assessing and Comparing Interpretability Techniques for Artificial Neural Networks Breast Cancer Classification

ABSTRACT

Breast cancer is the most common type of cancer among women. Thankfully, early detection and treatment improvements helped decrease the number of deaths. Data Mining techniques have always assisted BC tasks whether it is screening, diagnosis, prognosis, treatment, monitoring, and/or management. Nowadays, the use of Data Mining is witnessing a new era. In fact, the main objective is no longer to replace humans but to enhance their capabilities, which is why Artificial Intelligence is now referred to as Intelligence Augmentation. In this context, interpretability is used to help domain experts learn new patterns and machine learning experts debug their models. This paper aims to investigate three black-boxes interpretation techniques: Feature Importance, Partial Dependence Plot, and LIME when applied to two types of feed-forward Artificial Neural Networks: Multilayer perceptrons, and Radial Basis Function Network, trained on the Wisconsin Original dataset for breast cancer diagnosis. Results showed that local LIME explanations were instance-level interpretations that came in line with the global interpretations of the other two techniques. Global/local interpretability techniques can thus be combined to define the trustworthiness of a black-box model.

KEYWORDS:

• Interpretability
• explainability
• breast cancer
• diagnosis
• LIME
• Partial Dependence Plot
• features importance

Acknowledgments

The authors would like to thank the Moroccan Ministry of Higher Education and Scientific Research, Digital Development Agency (ADD), CNRST, and UM6P for their support.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was conducted under the research project ‘Machine Learning based Breast Cancer Diagnosis and Treatment’, Alkhawarizmi Program, 2020-2022.

Notes on contributors

Hajar Hakkoum

Hajar Hakkoum is a PhD student part of the Software Project Management team at the National School of Computer Science and Systems Analysis (ENSIAS), University Mohammed V in Rabat, Morocco. She received a software engineering degree in 2019 from ENSIAS.

Ali Idri

Ali Idri is a Full Professor at the Computer Science and Systems Analysis School (ENSIAS, University Mohammed V, Rabat, Morocco). He received his Master and Doctorate of 3rd Cycle in Computer Science from the University of Mohamed V in 1994 and 1997 respectively. He received his Ph.D. in Cognitive and Computer Sciences from the University of Quebec at Montreal in 2003. He is the head of the Software Project Management Research Team since 2010. He published more than 200 papers in well-recognized journals and conferences. His research interests include medical informatics, machine learning and software engineering.

Ibtissam Abnane

Ibtissam Abnane is an assistant professor in the National School of Computer Science and Systems Analysis (ENSIAS), University Mohammed V in Rabat, Morocco. She received a PhD in computer science in 2018 from ENSIAS, University Mohammed V. She has several active research areas such as machine learning, data science, software engineering and health informatics. She is a member of several national and international research projects aiming to improve health services in Morocco with the use of machine learning algorithms and Information and Communication Technologies.