import pandas as pd
import gradio as gr
import json
import pickle
from sklearn.decomposition import PCA

# Load the JSON file
with open('fraud_mean_by_country.json', 'r') as json_file:
    fraud_mean_by_country_dict = json.load(json_file)

# Load the decision tree model
with open("decision_tree_model.pkl", "rb") as file:
    clf = pickle.load(file)

# Load PCA model
with open("pca_model.pkl", "rb") as file:
    pca = pickle.load(file)

# Dummy data for generating choices
df = pd.read_csv('synthetic_financial_data.csv')

# Let's create new column from age column
def group(x):
  if x < 33:
    a = '18-32'
  elif x < 60:
    a = '33-60'
  elif x >= 60:
    a = '60+'
  return a

# Choices for dropdown menus
card_type_choices = [(val, val) for val in df['card_type'].unique()]
location_choices = [(val, val) for val in df['location'].unique()]
purchase_category_choices = [(val, val) for val in df['purchase_category'].unique()]
country_choices = [(val, val) for val in df['country'].unique()]

# Function to predict
def predict(transaction_id, customer_id, merchant_id, amount, transaction_time, card_type, location, purchase_category, customer_age, transaction_description, country):
    try:
        data = pd.DataFrame({
            'amount': [amount],
            'country': [country],
            'customer_age': [customer_age],
            'card_type': [card_type],
            'purchase_category': [purchase_category]
        })
        data = pd.get_dummies(data, columns=['card_type', 'purchase_category'])
        data['country'] = data['country'].map(fraud_mean_by_country_dict)
        data['Age_group'] = data['customer_age'].apply(lambda x: group(x))
        data = pd.get_dummies(data, columns=['Age_group'], drop_first=True, dtype=float)

        columns_to_add = ['customer_age','card_type_MasterCard','card_type_Discover', 'card_type_Visa','purchase_category_Online Shopping', 
                          'purchase_category_Groceries', 'purchase_category_Restaurant','purchase_category_Retail', 'purchase_category_Travel','purchase_category_Gas Station'
                          ,'Age_group_33-60', 'Age_group_60+']

        # Add missing columns with default value of 0
        for column in columns_to_add:
            if column not in data.columns:
                data[column] = 0

        for i in data.columns:
          data[i] = data[i].astype(int)

        # Perform PCA
        X_pca = pca.transform(data[['purchase_category_Groceries', 'purchase_category_Retail', 'purchase_category_Travel', 'Age_group_33-60', 'Age_group_60+']])
        data['pca1'] = X_pca[:, 0]
        data['pca2'] = X_pca[:, 1]

        data.drop(columns=['customer_age', 'card_type_Discover', 'card_type_Visa', 'purchase_category_Groceries', 'purchase_category_Retail', 'purchase_category_Travel', 'Age_group_33-60', 'Age_group_60+'], inplace=True)
        # Make predictions
        data = data[['amount', 'country', 'card_type_MasterCard','purchase_category_Online Shopping', 'purchase_category_Restaurant','pca1', 'pca2']]
        prediction = clf.predict(data)
        result = "Fraud ❌" if prediction[0] == 1 else "Not Fraud ✅"
        # Custom output for specific IDs
        return f'Customer with ID {customer_id} and Transaction ID {transaction_id}, which happened at {transaction_time}, is {result}'
    except Exception as e:
        return str(e)

# Define Gradio interface
inputs = [
    gr.Textbox(label="Transaction ID"),
    gr.Textbox(label="Customer ID"),
    gr.Textbox(label="Merchant ID"),
    gr.Number(label="Amount"),
    gr.Textbox(label="Transaction Date (YYYY-MM-DD)"), 
    gr.Dropdown(choices=card_type_choices, label="Card Type"),  
    gr.Dropdown(choices=location_choices, label="Location"),
    gr.Dropdown(choices=purchase_category_choices, label="Purchase Category"),
    gr.Number(label="Customer Age"),
    gr.Textbox(label="Transaction Description"),
    gr.Dropdown(choices=country_choices, label="Country")
]

# Define Gradio interface
gr.Interface(fn=predict, inputs=inputs, outputs="text",title="Fraud Detection Model",description="Enter details to predict fraud.").launch()