app.py

import gradio as gr
import pandas as pd
import pickle



# Load the trained model from data.pkl
def load_model():
  with open('data.pkl', 'rb') as file:
    model = pickle.load(file)
  return model

# Define the prediction function using the loaded model
def predict_user_profile(inputs):
  # Preprocess the input data
  

  # Create a DataFrame from the user input dictionary
  df = pd.DataFrame.from_dict([inputs])

  # Select the relevant feature columns used during model training
  feature_columns_to_use = ['statuses_count', 'followers_count', 'friends_count',
                           'favourites_count', 'listed_count', 'lang_code']
  df_features = df[feature_columns_to_use]

  # Load the pre-trained model
  model = load_model()

  # Make predictions using the loaded model
  prediction = model.predict(df_features)

  # Return the predicted class label (0 for fake, 1 for genuine)
  return "Genuine" if prediction[0] == 1 else "Fake"

# Define the Gradio interface
inputs = [
  gr.Textbox(label="statuses_count"),
  gr.Textbox(label="followers_count"),
  gr.Textbox(label="friends_count"),
  gr.Textbox(label="favourites_count"),
  gr.Textbox(label="listed_count"),
  gr.Textbox(label="name"),
  gr.Textbox(label="Language"),
]

outputs = gr.Textbox(label="Prediction")

# Create the Gradio interface
interface = gr.Interface(fn=predict_user_profile, inputs=inputs, outputs=outputs,
                         title='User Profile Classifier',
                         description='Predict whether a user profile is genuine or fake.')
interface.launch(share=True)