app.py

# Import necessary libraries
import numpy as np
import joblib  # For loading the serialized model
import pandas as pd  # For data manipulation
from flask import Flask, request, jsonify  # For creating the Flask API
import os

# Initialize the Flask application
revenue_forecast_predictor_api = Flask("Superkart Revenue Forecast Predictor")

# Load the trained machine learning model
model_path = os.path.join(os.path.dirname(__file__), "superkart_best_pipeline.joblib")
model = joblib.load(model_path)

# Define a route for the home page (GET request)
@revenue_forecast_predictor_api.get('/')
def home():
    return "Welcome to the Superkart Revenue Forecast Predictor!"

# Define an endpoint for single prediction (POST request)
@revenue_forecast_predictor_api.post('/v1/predict')
def predict_revenue_single():
    """
    Handles POST requests to the '/v1/predict' endpoint.
    Expects JSON payload with product-store details and returns predicted sales.
    """
    # Get JSON data
    property_data = request.get_json()

    # Create DataFrame from JSON
    df = pd.DataFrame([property_data])

    # Keep IDs for reference (if present)
    product_id = property_data.get("Product_Id", None)
    store_id = property_data.get("Store_Id", None)

    # Drop IDs before prediction
    features_df = df.drop(columns=["Product_Id", "Store_Id"], errors="ignore")

    # Make prediction
    predicted_sales = model.predict(df)[0]
    predicted_sales = round(float(predicted_sales), 2)

    # Build response
    response = {
        "Predicted_Revenue": predicted_sales
    }
    if product_id and store_id:
        response["Product_Id"] = product_id
        response["Store_Id"] = store_id

    return jsonify(response)

# Define an endpoint for batch prediction (POST request)
@revenue_forecast_predictor_api.post('/v1/predictbatch')
def predict_revenue_batch():
    """
    Handles POST requests to the '/v1/predictbatch' endpoint.
    Expects a CSV file containing multiple product-store records.
    Returns predicted revenue for all rows in the file.
    """
    # Get uploaded CSV file
    file = request.files['file']
    df = pd.read_csv(file)

    # Keep IDs for reference (if available)
    ids = None
    if "Product_Id" in df.columns and "Store_Id" in df.columns:
        ids = df["Product_Id"].astype(str) + "_" + df["Store_Id"].astype(str)

    # Drop IDs before prediction
    features_df = df.drop(columns=["Product_Id", "Store_Id"], errors="ignore")

    # Make predictions
    predicted_sales = model.predict(features_df).tolist()
    predicted_sales = [round(float(val), 2) for val in predicted_sales]

    # Build response
    if ids is not None:
        output_dict = dict(zip(ids, predicted_sales))
    else:
        output_dict = {"Row_" + str(i): val for i, val in enumerate(predicted_sales)}

    return jsonify(output_dict)

# Run the Flask application
if __name__ == '__main__':
    revenue_forecast_predictor_api.run(debug=True)