WMF Fraud Detection API Frontend

API SPECS

Transaction

Resource Definition

Base Path	/api/v1/transaction
Functionality	Validates transactions as fraud/not fraud
Verb	POST

Request Definition

parameter	description
contrib_id	int - contribution_id, used for logging
name	string - donor name
financial_type_id	int - financial_type_id
payment_instrument_id	int - payment_instrument_id
total_amount	float - donation amount
currency	string - abbrv. currency info
usd_amount	float - amount in usd
gateway	string - payment gateway
payment_method	string - mode of payment
country	string - abbrv. country name
utm_medium	string - utm_medium
utm_campaign	string - utm_campaign truncated to first 3 letters
avs_filter	float - avs_filter values
cvv_filter	float - cvv_filter values
country_filter	float - country_filter_values
email_domain_filter	float - email_domain_filter values
utm_filter	float - utm_filter values
ip_filter	float - ip_velocity_filter values
minfraud_filter	float - minfraud scores
receive_date	string - receive date in mysql format

Examples

cURL

curl -X POST \
  -d 'http://localhost:5000/api/v1/transaction?contrib_id=1234567&name=Jon%20Big%20Birdman&financial_type_id=9&payment_instrument_id=16&total_amount=15.99&currency=USD&usd_amount=15.99&gateway=globalcollect&payment_method=cc&country=US&utm_medium=Waystogive&utm_campaign=C18&avs_filter=0&cvv_filter=0&country_filter=0&email_domain_filter=0&utm_filter=0&ip_filter=0&minfraud_filter=0.10000000149011612&receive_date=2015-03-15%2018:04:55'

PHP

<?php

$curl = curl_init();

curl_setopt_array($curl, array(
  CURLOPT_PORT => "5000",
  CURLOPT_URL => "http://localhost:5000/api/v1/transaction",
  CURLOPT_RETURNTRANSFER => true,
  CURLOPT_ENCODING => "",
  CURLOPT_MAXREDIRS => 10,
  CURLOPT_TIMEOUT => 30,
  CURLOPT_HTTP_VERSION => CURL_HTTP_VERSION_1_1,
  CURLOPT_CUSTOMREQUEST => "POST",
  CURLOPT_POSTFIELDS => array(
  	'contrib_id' => 1234567,
  	'name' => "Jane Doe",
  	'financial_type_id' => 9,
  	'payment_instrument_id' => 15,
  	'receive_date' => "2018-08-01 02:12:42",
  	'total_amount' => 3.00,
  	'currency' => "USD",
    'usd_amount' => 3.00,
  	'gateway' => "globalcollect",
  	'payment_method' => "cc",
  	'country' => "SE",
  	'utm_medium' => "sidebar",
  	'utm_campaign' => "C13",
  	'avs_filter' => 50,
  	'cvv_filter' => 0,
  	'email_domain_filter' => 0,
  	'utm_filter' => 0,
  	'ip_filter' => 0,
  	'minfraud_filter' => 0.05000000074505806,
  	'country_filter' => 0,
  );,
));

$response = curl_exec($curl);
$err = curl_error($curl);

curl_close($curl);

if ($err) {
  echo "cURL Error #:" . $err;
} else {
  echo $response;
}

Response

response variable	description
prediction	int - 1 for fraud, 0 for not fraud
proba_score	float - probability score of txn being fraud

Example

{
    "prediction": 1,
    "proba_score": 0.985672492234017
}

Installation and usage

• Install the required python libraries using pip install. Strongly recommended that you do that in a virtualenv as there are dependencies on old scikit libraries.
• Make sure the private folder is populated with the required files. In order to generate them, follow - https://github.com/saurabhbatra96/wmf-fraud-pipeline#pipeline-steps---importing-a-freshupdated-version-of-the-model-into-the-api
• Move into the api version directory - $ cd v1.
• Start the api with $ python main.py
• The api should be ready to receive POST requests at http://localhost:5000/api/v1/transaction