Code Editor for app.py

import streamlit as st
import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestRegressor

# Initialize session state
if 'sample_data' not in st.session_state:
    # Create sample historical data
    np.random.seed(42)
    makes = ['Toyota', 'Honda', 'Ford', 'BMW', 'Mercedes']
    models = {
        'Toyota': ['Camry', 'Corolla', 'RAV4'],
        'Honda': ['Civic', 'Accord', 'CR-V'],
        'Ford': ['F-150', 'Explorer', 'Mustang'],
        'BMW': ['3 Series', '5 Series', 'X5'],
        'Mercedes': ['C-Class', 'E-Class', 'GLC']
    }
    
    data = []
    for make in makes:
        for model in models[make]:
            for month in range(1, 13):
                for year in [2021, 2022, 2023]:
                    base_demand = np.random.randint(50, 200)
                    seasonal_factor = 1 + 0.2 * np.sin(2 * np.pi * month / 12)
                    actual_sales = int(base_demand * seasonal_factor * (1 + np.random.normal(0, 0.1)))
                    data.append({
                        'make': make,
                        'model': model,
                        'year': year,
                        'month': month,
                        'sales': actual_sales
                    })
    
    st.session_state.sample_data = pd.DataFrame(data)

# Title and description
st.title('Vehicle Order Recommendation System')
st.markdown("""
This system helps predict recommended order quantities for different vehicle makes and models 
based on historical sales data and seasonal patterns.
""")

# Sidebar filters
st.sidebar.header('Select Vehicle Details')
selected_make = st.sidebar.selectbox('Make', options=sorted(st.session_state.sample_data['make'].unique()))

models_for_make = sorted(st.session_state.sample_data[
    st.session_state.sample_data['make'] == selected_make]['model'].unique())
selected_model = st.sidebar.selectbox('Model', options=models_for_make)

# Prepare data for the selected make/model
filtered_data = st.session_state.sample_data[
    (st.session_state.sample_data['make'] == selected_make) & 
    (st.session_state.sample_data['model'] == selected_model)
].copy()

# Train a simple model
X = filtered_data[['year', 'month']]
y = filtered_data['sales']
model = RandomForestRegressor(n_estimators=100, random_state=42)
model.fit(X, y)

# Show historical data
st.subheader('Historical Sales Data')
st.line_chart(filtered_data.set_index(['year', 'month'])['sales'])

# Make predictions for next 6 months
last_date = pd.to_datetime(f"{filtered_data['year'].max()}-{filtered_data['month'].max()}-01")
future_dates = pd.date_range(start=last_date, periods=7, freq='M')[1:]
future_data = pd.DataFrame({
    'year': future_dates.year,
    'month': future_dates.month
})

predictions = model.predict(future_data)
predictions = np.round(predictions).astype(int)

# Display recommendations
st.subheader('Order Recommendations')
st.markdown(f"Recommended order quantities for **{selected_make} {selected_model}**:")

recommendations_df = pd.DataFrame({
    'Year': future_dates.year,
    'Month': future_dates.month,
    'Recommended Quantity': predictions
})

st.table(recommendations_df)

# Additional insights
st.subheader('Additional Insights')
avg_monthly_sales = filtered_data.groupby('month')['sales'].mean().round(0)
peak_month = avg_monthly_sales.idxmax()
peak_sales = avg_monthly_sales.max()

st.markdown(f"""
- Peak sales month: {peak_month} (average of {peak_sales:.0f} units)
- Average monthly sales: {avg_monthly_sales.mean():.0f} units
- Yearly trend: {'+' if filtered_data.groupby('year')['sales'].mean().is_monotonic_increasing else '-'}
""")

# Confidence disclaimer
st.markdown("""
---
*Note: These recommendations are based on historical data and simple seasonal patterns. 
Actual optimal quantities may vary based on additional factors such as market conditions, 
promotions, and inventory costs.*
""")