An ecommerce company has unlabeled historical transaction CSV data in S3 and needs to detect anomalous transactions (fraud). Which combination of models should they use to detect anomalies in this unlabeled dataset? (Choose two.)

The company has unlabeled data, so unsupervised learning models are required. Random Cut Forest (RCF) is an unsupervised algorithm specifically designed for anomaly detection in streaming or batch data. It identifies data points that are statistically different from the rest. IP Insights is also an unsupervised anomaly detection algorithm that learns normal IP address usage patterns and flags deviations, which is highly relevant for detecting fraudulent transactions based on IP addresses. K-nearest neighbors (k-NN) can be used for anomaly detection in an unsupervised manner by identifying outliers as points with large distances to their neighbors, but RCF is often more efficient for large datasets. Linear learner with a logistic function and XGBoost are supervised learning algorithms, meaning they require labeled data for training, which is not available in this scenario.