Thorough data analysis and profiling process
In the previous section, you were introduced to the concept of data-centric XAI in which we discussed three important aspects of data-centric XAI: analyzing data volume, data consistency, and data purity. You might already be aware of some of the methods of data analysis and data profiling that we are going to learn in this section. But we are going to assume that we already have a trained ML model and, now, we are working toward explaining the model's decision-making process by adopting data-centric approaches.
The need for data analysis and profiling processes
In Chapter 2, Model Explainability Methods, when we discussed knowledge extraction using exploratory data analysis (EDA), we discovered that this was a pre-hoc analysis process, in which we try to understand the data to form relevant hypotheses. As data scientists, these initial hypotheses are important as they allow us to take the necessary steps to build a better model...