Notes

1. The following metrics were calculated using the confusion matrix:

Accuracy = (148 + 71) / (148 + 71 + 17 + 32)

= 81.7%

Recall = (71) / (71 + 32)

= 68.9%

Precision = (71) / (71 + 17)

= 80.7%


2. Based on the Feature Importance plot, the 10 most important features (in order) are as follows:

Sex, Fare, PassengerID, Age, Pclass, median(group_1), mean(group_1), max(group_1), SibSp, std(group_1)


3a. Are younger passengers more likely to survive?

Younger passengers are more lilkely to survive. Blue dots (representing lower feature value, or smaller age) have higher SHAP value and more impact on the model output compared to other dots. Blue dots (younger passengers) are more likely to result in survival. Red dots can be seen to have lower SHAP values, so they are less likely to result in survival.

3b. Are males more likely to survive then females?

Males are less likely to survive than females. Females = 0 are represented by the blue dots. The blue dots have a high SHAP value and have more impact on the model output compared to red dots (Males). This means that females are more likely to result in a survivor output than males.

3c. Did passengers who paid a high fare improve their chance at survival?

Passengers who paid a higher fare are more likely to survive. Blue dots (representing passengers who paid lower fare) can be seen to have a lower SHAP value (less impact on model output). Thus, they are less likely to survive. Purple and red dots (higher fares) can be seen to the right which indicates higher SHAP value, more impact on model output, and higher chance of survival.