Practical Data Science using Python - Logistic Regression - Model Optimization 2

The video tutorial discusses the process of determining the optimum probability threshold for predictions in a logistic regression model, focusing on a telecom churn prediction case. It explores the tradeoffs between accuracy, sensitivity, specificity, precision, and recall, and explains how to choose the right threshold based on data balance. The tutorial also covers making predictions on test data using the chosen threshold and recalculating metrics to evaluate model performance.

What is the initial optimum probability threshold discussed in the video?

• 0.2

• 0.3

• 0.5

• 0.4

Which metric is NOT recalculated after changing the probability threshold?

• F1 Score

• Precision

• Specificity

• Sensitivity

What is the purpose of the precision-recall curve in the context of this video?

• To calculate the accuracy of the model

• To find the optimum probability threshold

• To determine the best model architecture

• To visualize the data distribution

Why is there a dilemma between choosing a threshold of 0.3 and 0.4?

• Due to conflicting results from different metrics

• Because of a lack of data

• Due to a software error

• Because both thresholds yield the same results

What factor influences the decision to use the precision-recall curve threshold?

• Moderate class imbalance

• Equal class distribution

• High class imbalance

• Low data quality

What is the new probability threshold determined by the precision-recall tradeoff?

• 0.5

• 0.4

• 0.3

• 0.2

What is the final accuracy score achieved after applying the new threshold?

• 77.9%

• 76%

• 85%

• 80%

Which function is used to scale the test data?

• MinMaxScaler

• RobustScaler

• Normalizer

• StandardScaler

What is the purpose of the predict_proba function in this context?

• To calculate probability values

• To scale the data

• To visualize the data

• To clean the data

What is the sensitivity value achieved with the final model?

• 58%

• 65%

• 75%

• 85%