|
--- |
|
license: apache-2.0 |
|
--- |
|
# Revenue Assurance and Fraud Management (RAFM) with AI Assistance |
|
|
|
## Project Overview |
|
|
|
This project aims to deliver an RAFM prediction (if that particular telco transaction is fraudulent or not) with a AI model assistance with; <br> |
|
(1) Balanced Random Forest,<br> |
|
The model has trained on semi-synthetic telecom data to predict fraud cases and identify potential anomalies. The goal is to provide proactive revenue management and enhance revenue workflows. |
|
## Data |
|
![Revenue Assurance Data Structure](https://raw.githubusercontent.com/fenar/etc-ai-wrx/main/revenueassurance/data/rev_ass_data.png)<br> |
|
## Results: |
|
![Revenue Assurance Accuracy](https://raw.githubusercontent.com/fenar/etc-ai-wrx/main/revenueassurance/data/rev_ass_models_accuracy.png)<br> |
|
## Steps to Test |
|
(A) Potential Fraud Test: <br> |
|
``` |
|
curl -X POST -H "Content-Type: application/json" -d '{ |
|
"Call_Duration": 300, |
|
"Data_Usage": 10000, |
|
"Sms_Count": 50, |
|
"Roaming_Indicator": 1, |
|
"MobileWallet_Use": 1, |
|
"Plan_Type_prepaid": 1, |
|
"Plan_Type_postpaid": 0, |
|
"Cost": 500, |
|
"Cellular_Location_Distance": 100, |
|
"Personal_Pin_Used": 0, |
|
"Avg_Call_Duration": 50, |
|
"Avg_Data_Usage": 8000 |
|
}' http://localhost:5000/predict |
|
``` |
|
(B) Potential Non-Fraud Test: <br> |
|
``` |
|
curl -X POST -H "Content-Type: application/json" -d '{ |
|
"Call_Duration": 10, |
|
"Data_Usage": 300, |
|
"Sms_Count": 5, |
|
"Roaming_Indicator": 0, |
|
"MobileWallet_Use": 1, |
|
"Plan_Type_prepaid": 1, |
|
"Plan_Type_postpaid": 0, |
|
"Cost": 50, |
|
"Cellular_Location_Distance": 3, |
|
"Personal_Pin_Used": 1, |
|
"Avg_Call_Duration": 12, |
|
"Avg_Data_Usage": 350 |
|
}' http://localhost:5000/predict |
|
``` |