Determining Predictor Relevance in Loan Prediction Model
Question
You are using classification algorithms in AutoML to train a model to predict whether your customers are expected to take a loan or not.
Your model has predictors as marital status, job and education.
After running ML experiments, you want to find which predictor is most relevant in predicting the target variable.
Which action should you take?
Answers
Explanations
Click on the arrows to vote for the correct answer
A. B. C. D.Answer: D.
Option A is incorrect because by examining the local importance of features can help understandhow each feature contributes to the result of a specific prediction.
Option B is incorrect because auto-featurization instructs AutoML to generate derived features based on the original features of the dataset.
It has no effect on the model explainability.
Option C is incorrect because while global feature importance can be used to understand the relative importance of features in the test dataset, in this case you should look for the feature with the highest global importance.
Option D is CORRECT because global feature importance can be used to understand the relative importance of features.
You should look for features with the highest global importance as the strongest contributors to the predictions.
Reference:
To find which predictor is most relevant in predicting the target variable, we need to evaluate the importance of each feature or predictor. In Azure AutoML, feature importance can be measured using global importance or local importance.
Global Importance: Global importance measures the overall importance of a feature in a dataset. It is determined by examining the feature's effect on the model's performance. This method considers the feature's contribution to the model's accuracy or error rate.
Local Importance: Local importance measures the importance of a feature for a particular instance in the dataset. This method considers the effect of the feature on the predicted value for each instance.
In this scenario, we want to determine the most relevant feature for predicting whether a customer is likely to take a loan or not. Therefore, we should use global importance to evaluate the contribution of each feature in the model's accuracy.
Option A suggests selecting the feature with the highest local importance. However, local importance only measures the importance of a feature for a specific instance, not the overall contribution to the model's accuracy.
Option B suggests enabling auto-featurization, which is a technique that automatically transforms and combines input features to improve model performance. However, this option does not directly answer the question of determining the most relevant feature for predicting the target variable.
Option C suggests selecting the feature with the lowest global importance. This option is incorrect since we are looking for the most relevant feature, which would have the highest global importance, not the lowest.
Option D suggests selecting the feature with the highest global importance. This is the correct option since we want to find the feature that contributes the most to the model's accuracy for predicting the target variable.
Therefore, the correct answer is D: Select the feature with the highest global importance.
