Explainable AI

Hardik Chheda Updated by Hardik Chheda

What is Explainable AI?

Explainable AI exposes business friendly interpretation of machine learning models which makes Tellius more transparent and trust worthy. In other words, it's a breakdown version of predicted value along with the accuracy precision to indicate and validate the model's results for each row in the data. 

Why is it Important?

Many machine learning models act as black boxes for end users. It brings users, who have no background or programming background closer to predictive analytics. This is because machine learning aims to address how black box decisions of AI systems are made. 

How Does it Work?

Explainable AI leverages the ability to use specific techniques to explainable complex models findings in Natural Language. 

Machine learning in classification works by:

  1. Transforming the input feature space into a different representation (feature engineering) and 
  2. Searching for a decision boundary to separate the classes in that representation space. (optimization). 

Note: 1 and 2 together are known as a hierarchical representation learning.

Simple models like linear regression are also inherently explainable since the weights and signs on them are indicators of importance of features and decisions can be explained in terms of those weights.

Decision trees also are explainable, because the various attributes are used to form splits. These splitting decisions can be used as a basis for mining rules that can suffice to some extent as explanations. Similar there are other ways to extract explanations from models.

What can I do with it?

You can validate and check the accuracy and the prediction for each row, within a sample subset of your data. You can also apply the model of your choice that is useful to another dataset (subset) and view explainable predictive analytics on the fly.

You can download the fly and take your results and use it out side of Tellius to continue further analysis.

To get started, create predictive analytics on Tellius and analyze the explanations. 

On the Predict page, when you open the AutoML prediction model, you can view the detailed explanation of the model.

The explanation includes the following information:

  1. The column that you are predicting
  2. The factors that are contributing to the prediction model
  3. The table with rows for the selected column. Select a row to view the prediction
  4. The option to view the prediction for trained data or new data

You can view the information for every model/ algorithm listed in the leaderboard.

How did we do?

Contact