Fraud_​Detection_​Isolation_​Forest_​Training

0 ×

This workflow uses the KNIME H2O Machine Learning Integration to train an isolation forest model for fraud detection, a method particularly effective for anomaly detection, to spot potential fraud in transactional data. Starting with data intake and preprocessing, it segments the data into training and validation sets. We then use the H2O Isolation Forest Learner to train the model on the 2/3 of 'good' transaction patterns. The trained model is validated against the training set from the split, with its performance measured and the model saved for use in the deployment workflow.

Here are the steps taken to train the mode;
1. Read Training Data
2. Data Preprocessing: Partition data into a training, test, and validation set
3. Train/Validate the model and Evaluate results
4. Save model for deployment
5. Evaluate Results

URL: Kaggle Dataset https://www.kaggle.com/datasets/mlg-ulb/creditcardfraud

Nodes

• Rule Engine2 ×
• Table to H2O2 ×
• CSV Reader1 ×
• Component Input1 ×
• Component Output1 ×
• Show all 14 nodes

Extensions

• FeatureKNIME Base nodes
• FeatureKNIME H2O Machine Learning Integration
• FeatureKNIME JavaScript Views (Labs)
• FeatureKNIME Javasnippet

Links

• No links available