Types of model explainability methods
There are different approaches that you can use to provide model explainability. Certain techniques are specific to a model, and certain approaches are applied to the input and output of the model. In this section, we will discuss the different types of methods used to explain ML models:
- Knowledge extraction methods: Extracting key insights and statistical information from the data during Exploratory Data Analysis (EDA) and post-hoc analysis is one way of providing model-agnostic explainability. Often, statistical profiling methods are applied to extract the mean and median values, standard deviation, or variance across the different data points, and certain descriptive statistics are used to estimate the expected range of outcomes.
Similarly, other insights using correlation heatmaps, decomposition trees, and distribution plots are also used to observe any relationships between the features to explain the model's results. For...