XPlainIT is about opening the “black box” decision making of machine learning algorithms so that decisions are transparent and understandable. The users of this capability to explain decisions are data-scientists, end-users, company personnel, regulatory authorities, or indeed any stakeholder who has a valid remit to ask questions about the decision making of such systems. The focus of XplainIT is on Deep Learning Models for structured data.
XPlainIT is a Web-based tool which offers a series of functionalities for explaining the decision-making process of Deep Learning models applied to structured data in a user-friendly manner, including:
- Three cutting-edge methods: LRP (mainly), LIME and SHAP.
- Two scenarios: Telecom Customer Churn prediction and Credit Card Fraud Detection.
- Three types of Analysis: Local (Individual), Global (Exploring Patterns) and New Record (Exploring New Inputs).
- Explanation based on intuitive and interactive Charts: Heatmaps, Bar plots, Range plots, Box plots.
- Explanation based on Text and Infoboxes.
The tool is relevant to any company or organisation who needs to discover how explainability can explain the decision making of a deep neural network (in this case 1D-CNN). This can help them in understanding certain issues in a system, such as bias, critical decision features, unnecessary features that can slow down processing and increasing equipment costs, and/or more specifically enforcing or taking care of GDPR.
- Dr Susan Mckeever, Technological University Dublin,
- Dr Ihsan Ullah, Technological University Dublin,
- Mr André Ríos, Technological University Dublin,
- Mr Viabhav Gala, Technological University Dublin