Questions
The following are questions that will consolidate and deepen your knowledge of fraud detection:
- What is a Gaussian Distribution?
- The algorithm in our fraud detection system requires something really important to be fed into it, before generating probabilities? What is that?
- Why is the selection of an error term (Epsilon) such a big deal in detecting outliers and identifying the correct false positives and false negatives?
- Why is fraud detection not exactly a classification problem?
- Fraud detection is essentially an anomaly identification problem. Can you name two properties that define anomaly identification?
- Can you think of other applications that can leverage anomaly identification or outlier detection?
- Why is cross-validation so important?
- Why is our fraud detection problem not a supervised learning problem?
- Can you name a couple of ways to optimize the Gaussian Distribution algorithm?
- Sometimes, our results may not be satisfactory because the algorithm failed to identify certain samples...
