![]() ![]() For instance, we might know that transactions from account x on day y usually do not exceed a certain amount so all transactions that do not fulfill this condition could be flagged as anomalies. One way to tackle this problem could be to devise some hard-coded criteria for ‘normal’ transactions that are based on domain knowledge. Fraudulent activity often deviates from these patterns in some way, providing an entry-point for data-driven methods of fraud detection. Transaction records capture the flow of assets between parties, which, if observed over long periods of time, follows certain patterns. One type of data where anomalies are considered to be of particularly high interest is financial transactions. In many cases, this corresponds to finding data that was created erroneously, or by fraudulent activity. The idea is to find entries that were generated by a different process than the majority of the data. Anomaly Detection and Transaction Data MotivationĪnomaly detection typically refers to the process of identifying outliers in a set of data that is largely composed of ‘normal’ data points. If you prefer watching a video about this rather than reading text, feel free to check out a recording of one of our webinars that covers the contents of this article here. ![]() Finally, we are going to touch the topic of Explanation Models and apply them to the results from the previous step for a qualitative evaluation. Next, we will evaluate the Results and compare them to reasonable baselines. We will discuss the Dataset used, and go through all the steps from Feature Engineering to choosing and building the right Model for the task at hand. In the following article we will discuss the topic of Anomaly Detection and Transaction Data, and why it makes sense to employ an unsupervised machine learning model to detect fraudulent transactions. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |