Project name
Anomaly_Detection
Description
This is a Microsoft-Fabric-powered solution, geared towards pin-pointing anomalies in financial institutional data. The main objective is built around detection and automated alerting (no predictions happening). Our solution uses advanced anomaly detection to flag unusual activity as it happens—before the damage is done. It’s lightweight, and integrates seamlessly into the financial institutions' existing systems, ensuring robust protection without slowing down services. From a high-level POV, once the transactions data is accessed, our solution runs the Machine Learning model, which detects anomalies, uses OpenAI's GPT-4o-mini to analyze and rank the anomalies in levels of (high/medium/low), generates an explanation for the deductions along with the alert message that contains the transaction info. After this, the high-risk anomalies are sorted out, added to text files that are automatically uploaded to Azure Blob storage. We have configured Power Automate to monitor the blob storage and additions or modifications of files triggers the email sending steps or even teams message(depending on what works for the team), prompting risk analysts to act on these anomalies.
Project Repository URL
https://github.com/SamuelNw/Anomaly_Detection
Project video
https://danielwangere-my.sharepoint.com/:v:/g/personal/danielwangere_danielwangere_onmicrosoft_com/Eb5JsWnHzQJBnEdE0uWb2-oBbXSdhgDazgl-ukoiImvHng?e=OXRyfl&nav=eyJyZWZlcnJhbEluZm8iOnsicmVmZXJyYWxBcHAiOiJTdHJlYW1XZWJBcHAiLCJyZWZlcnJhbFZpZXciOiJTaGFyZURpYWxvZy1MaW5rIiwicmVmZXJyYWxBcHBQbGF0Zm9ybSI6IldlYiIsInJlZmVycmFsTW9kZSI6InZpZXcifX0%3D
Team members
@SamuelNw, @DanWangere, @CynthiaKiplagat
Registration status.
Project name
Anomaly_Detection
Description
This is a Microsoft-Fabric-powered solution, geared towards pin-pointing anomalies in financial institutional data. The main objective is built around detection and automated alerting (no predictions happening). Our solution uses advanced anomaly detection to flag unusual activity as it happens—before the damage is done. It’s lightweight, and integrates seamlessly into the financial institutions' existing systems, ensuring robust protection without slowing down services. From a high-level POV, once the transactions data is accessed, our solution runs the Machine Learning model, which detects anomalies, uses OpenAI's GPT-4o-mini to analyze and rank the anomalies in levels of (high/medium/low), generates an explanation for the deductions along with the alert message that contains the transaction info. After this, the high-risk anomalies are sorted out, added to text files that are automatically uploaded to Azure Blob storage. We have configured Power Automate to monitor the blob storage and additions or modifications of files triggers the email sending steps or even teams message(depending on what works for the team), prompting risk analysts to act on these anomalies.
Project Repository URL
https://github.com/SamuelNw/Anomaly_Detection
Project video
https://danielwangere-my.sharepoint.com/:v:/g/personal/danielwangere_danielwangere_onmicrosoft_com/Eb5JsWnHzQJBnEdE0uWb2-oBbXSdhgDazgl-ukoiImvHng?e=OXRyfl&nav=eyJyZWZlcnJhbEluZm8iOnsicmVmZXJyYWxBcHAiOiJTdHJlYW1XZWJBcHAiLCJyZWZlcnJhbFZpZXciOiJTaGFyZURpYWxvZy1MaW5rIiwicmVmZXJyYWxBcHBQbGF0Zm9ybSI6IldlYiIsInJlZmVycmFsTW9kZSI6InZpZXcifX0%3D
Team members
@SamuelNw, @DanWangere, @CynthiaKiplagat
Registration status.