Detect Fraud

app.py

@@ -0,0 +1,94 @@
+import gradio as gr
+import pandas as pd
+import pickle
+from typing import List, Dict, Any
+
+# CSS styles (unchanged)
+css = """
+body, html {
+    height: 100%;
+    margin: 0;
+    color: white;
+    background-color: black;
+}
+header {
+    background: url('fraude.png') no-repeat top left;
+    background-size: 120px; /* Adjust this value to the desired size of your image */
+    padding-top: 130px; /* Adjust this value to provide space for the image */
+}
+"""
+
+# Load model and configurations
+def load_pickle(filename: str) -> Any:
+    with open(filename, 'rb') as f:
+        return pickle.load(f)
+
+model = load_pickle('model/modelo_proyecto_final2.pkl')
+ohe_columns = load_pickle('model/categories_ohe_without_fraudulent.pickle')
+bins_order = load_pickle('model/saved_bins_order.pickle')
+bins_transaction = load_pickle('model/saved_bins_transaction.pickle')
+
+def predict(order_amount: float, order_state: str, payment_method_registration_failure: bool,
+            payment_method_type: str, payment_method_provider: str, payment_method_issuer: str,
+            transaction_amount: float, transaction_failed: bool, email_domain: str,
+            email_provider: str, customer_ip_address_simplified: str, same_city: str) -> str:
+    # Create input DataFrame
+    data = {
+        "orderAmount": [order_amount],
+        "orderState": [order_state],
+        "paymentMethodRegistrationFailure": [payment_method_registration_failure],
+        "paymentMethodType": [payment_method_type],
+        "paymentMethodProvider": [payment_method_provider],
+        "paymentMethodIssuer": [payment_method_issuer],
+        "transactionAmount": [transaction_amount],
+        "transactionFailed": [transaction_failed],
+        "emailDomain": [email_domain],
+        "emailProvider": [email_provider],
+        "customerIPAddressSimplified": [customer_ip_address_simplified],
+        "sameCity": [same_city]
+    }
+    df = pd.DataFrame(data)
+    
+    # Fill null values with modes or medians
+    for column in df.columns:
+        if pd.api.types.is_numeric_dtype(df[column]):
+            df[column].fillna(df[column].median(), inplace=True)
+        else:
+            df[column].fillna(df[column].mode().iloc[0], inplace=True)
+    
+    # Apply binning and one-hot encoding
+    df["orderAmount"] = pd.cut(df["orderAmount"].astype(float), bins=bins_order, include_lowest=True)
+    df["transactionAmount"] = pd.cut(df["transactionAmount"].astype(int), bins=bins_transaction, include_lowest=True)
+    df_encoded = pd.get_dummies(df).reindex(columns=ohe_columns, fill_value=0)
+
+    # Prediction
+    prediction = model.predict(df_encoded)
+    type_of_fraud = int(prediction[0])
+    responses = ["OK", "FRAUD DETECTED", "Warning"]
+    return responses[type_of_fraud] if type_of_fraud in [0, 1, 2] else "Error parsing value"
+
+# Configure Gradio interface
+interface = gr.Interface(
+    fn=predict,
+    inputs=[
+        gr.Number(label="Order Amount"),
+        gr.Dropdown(choices=["pending", "fulfilled", "failed"], label="Order State"),
+        gr.Checkbox(label="Payment Method Registration Failure"),
+        gr.Dropdown(choices=["card", "bitcoin", "paypal"], label="Payment Method Type"),
+        gr.Dropdown(choices=['JCB 16 digit', 'VISA 16 digit', 'Diners Club / Carte Blanche', 'Mastercard', 'American Express', 'Maestro', 'Discover', 'Voyager', 'VISA 13 digit', 'JCB 15 digit'], label="Payment Method Provider"),
+        gr.Dropdown(choices=['Citizens First Banks', 'Solace Banks', 'Vertex Bancorp', 'His Majesty Bank Corp.', 'Bastion Banks', 'Her Majesty Trust', 'Fountain Financial Inc.', 'Grand Credit Corporation', 'weird', 'Bulwark Trust Corp.', 'Rose Bancshares'], label="Payment Method Issuer"),
+        gr.Number(label="Transaction Amount"),
+        gr.Checkbox(label="Transaction Failed"),
+        gr.Dropdown(choices=["info","com","biz","org"], label="Email Domain"),
+        gr.Dropdown(choices=["yahoo","gmail","hotmail","other"], label="Email Provider"),
+        gr.Dropdown(choices=["only_letters", "digits_and_letters"], label="Customer IP Address Simplified"),
+        gr.Dropdown(choices=["yes", "no"], label="Same City")
+    ],
+    outputs="text",
+    title="API for Fraud Detection",
+    description="APP to predict if a transaction is fraudulent.",
+    css=css
+)
+
+if __name__ == "__main__":
+    interface.launch()

model/categories_ohe_without_fraudulent.pickle

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dbeb4b297ccfe265a6cceb5283089d2276a835ec3735a39e17b698952983fa8
+size 1972

model/modelo_proyecto_final2.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e90f9a99d271fa2abfe3e52fe8c26efb7741652be995b53da3503f5510ada398
+size 3996578

model/saved_bins_order.pickle

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:552fdcdb92034ea7aebdb6d837f1dcef67f643d8a9c409074d425789a5cd0317
+size 174

model/saved_bins_transaction.pickle

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23fd3e29980e60fa786b5819d59e4803e6817a2cd8c16fbe610db359ba413fe3
+size 166

requirements.txt

@@ -0,0 +1 @@
+Cython>=0.29