Which data mining technique is applied when the objective is to predict a known outcome, such as whether a claim will settle or go to litigation?

Which data mining technique is applied when the objective is to predict a known outcome, such as whether a claim will settle or go to litigation?

• A. Classification
• B. Machine learning
• C. Cluster analysis
• D. Association rule learning

Answer: (A)

Classification is the technique when you want to predict a known outcome that falls into predefined categories, such as whether a claim will settle or go to litigation. It’s a supervised learning approach: you train a model on labeled examples where the outcome is known, and then apply the model to new cases to assign them to a category. For a binary outcome like settle versus go to litigation, classification methods—think logistic regression, decision trees, or random forests—are well suited because they learn the relationship between claim features (like claim type, severity, policy limits, prior history) and the upcoming outcome. The other options don’t fit this predictive goal as directly. Cluster analysis is unsupervised and groups cases by similarity without using known outcomes. Association rule learning discovers relationships between items or attributes (for example, which features tend to co-occur) rather than predicting a specific future label. Broad machine learning describes the field, but without specifying the predictive task, it’s too general for this particular goal.

Classification is the technique when you want to predict a known outcome that falls into predefined categories, such as whether a claim will settle or go to litigation. It’s a supervised learning approach: you train a model on labeled examples where the outcome is known, and then apply the model to new cases to assign them to a category. For a binary outcome like settle versus go to litigation, classification methods—think logistic regression, decision trees, or random forests—are well suited because they learn the relationship between claim features (like claim type, severity, policy limits, prior history) and the upcoming outcome.

The other options don’t fit this predictive goal as directly. Cluster analysis is unsupervised and groups cases by similarity without using known outcomes. Association rule learning discovers relationships between items or attributes (for example, which features tend to co-occur) rather than predicting a specific future label. Broad machine learning describes the field, but without specifying the predictive task, it’s too general for this particular goal.