Blame for app.py

import streamlit as st
import pandas as pd
import numpy as np
import joblib
from datetime import datetime

# 1. Load the model and features
model = joblib.load('linear_regression_model.pkl')
features = joblib.load('model_features.pkl')

st.set_page_config(page_title="Store Sales Forecaster", layout="centered")

st.title("📈 Store Sales Forecasting")
st.markdown("""
This app predicts daily sales for a retail store in Ecuador using a **Linear Regression** model.
It accounts for promotions, oil prices, and special events like holidays and paydays.
""")

st.sidebar.header("User Input Parameters")

def user_input_features():
    # Date Input
    date = st.sidebar.date_input("Select Date", datetime(2017, 8, 16))
    
    # Store and Family Input
    store_nbr = st.sidebar.number_input("Store Number", min_value=1, max_value=54, value=1)
    onpromotion = st.sidebar.number_input("Items on Promotion", min_value=0, value=0)
    
    # Simplified dropdowns (we'll map these to the codes the model expects)
    # Note: In a production app, you'd map these to the exact category codes used in training
    family_code = st.sidebar.slider("Product Family Code", 0, 32, 10)
    holiday_code = st.sidebar.slider("Holiday Type Code", 0, 6, 0)
    store_type_code = st.sidebar.slider("Store Type Code", 0, 4, 0)
    
    # Static or Derived Features
    oil_price = st.sidebar.number_input("Current Oil Price (dcoilwtico)", value=45.0)
    
    # Deriving date-based features
    month = date.month
    day_of_week = date.weekday()
    is_payday = 1 if (date.day == 15 or date.day == pd.Period(date, freq='D').days_in_month) else 0
    earthquake = 0 # Currently 0 as the event has passed
    
    data = {
        'store_nbr': store_nbr,
        'family_code': family_code,
        'onpromotion': onpromotion,
        'dcoilwtico': oil_price,
        'month': month,
        'day_of_week': day_of_week,
        'holiday_code': holiday_code,
        'is_payday': is_payday,
        'earthquake': earthquake,
        'store_code': store_type_code
    }
    
    return pd.DataFrame(data, index=[0])

input_df = user_input_features()

st.subheader("Input Summary")
st.write(input_df)

# Prediction
if st.button("Predict Sales"):
    # Reorder columns to match the trained model's features list
    input_df = input_df[features]
    prediction = model.predict(input_df)
    
    # Apply clip to ensure no negative sales
    result = np.clip(prediction[0], 0, None)
    
    st.success(f"### Estimated Sales: {result:.2f} units")
    
    st.info("Note: This prediction is based on a baseline Linear Regression model trained on historical data from 2013-2017.")