Spaces:

EdwinHB
/

david

Runtime error

david / main.py

Create main.py

27b0e2e verified over 2 years ago

4.51 kB

import gradio as gr
import pandas as pd
from prophet import Prophet
import matplotlib.pyplot as plt

def forecast_sales(file):
    # Load the sales data
    sales_data = pd.read_csv(file.name)

    # Convert date column to datetime
    sales_data['date'] = pd.to_datetime(sales_data['date'])

    # Prepare the data for Prophet with additional regressors
    prophet_data = sales_data[['date', 'sales', 'orders', 'goog_upper_spend', 'fb_upper', 'search_spend', 'brand_spend', 'shopping_spend', 'fb_spend', 'pinterest_spend', 'tiktok', 'RTB_Spend', 'venta_buen_fin', 'venta_hot_sale']].rename(columns={'date': 'ds', 'sales': 'y'})

    # Fill missing values with the mean of the respective columns
    prophet_data.fillna(prophet_data.mean(), inplace=True)

    # Initialize the Prophet model with additional regressors
    model = Prophet()

    # Add additional regressors
    regressors = ['orders', 'goog_upper_spend', 'fb_upper', 'search_spend', 'brand_spend', 'shopping_spend', 'fb_spend', 'pinterest_spend', 'tiktok', 'RTB_Spend', 'venta_buen_fin', 'venta_hot_sale']
    for regressor in regressors:
        model.add_regressor(regressor)

    # Fit the model
    model.fit(prophet_data)

    # Create a dataframe for future dates with additional regressors
    future = model.make_future_dataframe(periods=30)
    future = future.merge(sales_data[['date'] + regressors].set_index('date'), left_on='ds', right_index=True, how='left')
    future.fillna(future.mean(), inplace=True)

    # Predict the future sales
    forecast = model.predict(future)

    # Plot the forecast
    fig1, ax1 = plt.subplots()
    model.plot(forecast, ax=ax1)
    ax1.set_title('Sales Forecast with Additional Regressors')
    ax1.set_xlabel('Date')
    ax1.set_ylabel('Sales')

    # Merge the forecasted sales with the actual sales data for comparison
    comparison_df = pd.merge(forecast[['ds', 'yhat']], prophet_data, left_on='ds', right_on='ds', how='left')
    comparison_df.rename(columns={'yhat': 'forecasted_sales', 'y': 'actual_sales'}, inplace=True)

    # Resample the comparison dataframe to monthly frequency and calculate the mean sales for each month
    monthly_comparison_df = comparison_df.resample('M', on='ds').mean()

    # Filter the monthly comparison dataframe to show only the last 12 months
    last_12_months_df = monthly_comparison_df.tail(12)

    # Calculate the percentage difference between actual and forecasted sales for the last 12 months
    last_12_months_df['percentage_difference'] = ((last_12_months_df['actual_sales'] - last_12_months_df['forecasted_sales']) / last_12_months_df['actual_sales']) * 100

    # Drop rows with NaN values in 'actual_sales'
    last_12_months_df = last_12_months_df.dropna(subset=['actual_sales'])

    # Plot the monthly comparison with bar charts and the percentage difference on top
    fig2, ax2 = plt.subplots(figsize=(14, 7))
    bar_width = 15  # Set the bar width

    # Plot actual sales bars
    bars_actual = ax2.bar(last_12_months_df.index, last_12_months_df['actual_sales'], 
                          width=bar_width, label='Actual Sales', color='black', alpha=0.7)

    # Plot forecasted sales bars
    bars_forecasted = ax2.bar(last_12_months_df.index + pd.DateOffset(days=15), 
                              last_12_months_df['forecasted_sales'], width=bar_width, 
                              label='Forecasted Sales', color='blue', alpha=0.7)

    # Add the percentage difference on top of the bars
    for bar, diff in zip(bars_actual, last_12_months_df['percentage_difference']):
        height = bar.get_height()
        ax2.text(bar.get_x() + bar.get_width() / 2.0, height, f'{diff:.2f}%', ha='center', va='bottom', color='red')

    ax2.set_title('Monthly Actual vs Forecasted Sales with Percentage Difference (Last 12 Months)')
    ax2.set_xlabel('Date')
    ax2.set_ylabel('Sales')
    ax2.legend()

    # Save the comparison dataframe to a CSV
    last_12_months_df.to_csv('forecast_comparison.csv')

    return fig1, fig2, 'forecast_comparison.csv'

# Define the Gradio interface
interface = gr.Interface(
    fn=forecast_sales,
    inputs=gr.File(label="Upload CSV File"),
    outputs=[
        gr.Plot(label="Sales Forecast"),
        gr.Plot(label="Monthly Comparison"),
        gr.File(label="Comparison CSV")
    ],
    title="Sales Forecasting with Prophet",
    description="Upload a CSV file with sales data to forecast future sales and compare with actual sales."
)

# Launch the interface
interface.launch(debug=True)

	import gradio as gr
	import pandas as pd
	from prophet import Prophet
	import matplotlib.pyplot as plt

	def forecast_sales(file):
	# Load the sales data
	sales_data = pd.read_csv(file.name)

	# Convert date column to datetime
	sales_data['date'] = pd.to_datetime(sales_data['date'])

	# Prepare the data for Prophet with additional regressors
	prophet_data = sales_data[['date', 'sales', 'orders', 'goog_upper_spend', 'fb_upper', 'search_spend', 'brand_spend', 'shopping_spend', 'fb_spend', 'pinterest_spend', 'tiktok', 'RTB_Spend', 'venta_buen_fin', 'venta_hot_sale']].rename(columns={'date': 'ds', 'sales': 'y'})

	# Fill missing values with the mean of the respective columns
	prophet_data.fillna(prophet_data.mean(), inplace=True)

	# Initialize the Prophet model with additional regressors
	model = Prophet()

	# Add additional regressors
	regressors = ['orders', 'goog_upper_spend', 'fb_upper', 'search_spend', 'brand_spend', 'shopping_spend', 'fb_spend', 'pinterest_spend', 'tiktok', 'RTB_Spend', 'venta_buen_fin', 'venta_hot_sale']
	for regressor in regressors:
	model.add_regressor(regressor)

	# Fit the model
	model.fit(prophet_data)

	# Create a dataframe for future dates with additional regressors
	future = model.make_future_dataframe(periods=30)
	future = future.merge(sales_data[['date'] + regressors].set_index('date'), left_on='ds', right_index=True, how='left')
	future.fillna(future.mean(), inplace=True)

	# Predict the future sales
	forecast = model.predict(future)

	# Plot the forecast
	fig1, ax1 = plt.subplots()
	model.plot(forecast, ax=ax1)
	ax1.set_title('Sales Forecast with Additional Regressors')
	ax1.set_xlabel('Date')
	ax1.set_ylabel('Sales')

	# Merge the forecasted sales with the actual sales data for comparison
	comparison_df = pd.merge(forecast[['ds', 'yhat']], prophet_data, left_on='ds', right_on='ds', how='left')
	comparison_df.rename(columns={'yhat': 'forecasted_sales', 'y': 'actual_sales'}, inplace=True)

	# Resample the comparison dataframe to monthly frequency and calculate the mean sales for each month
	monthly_comparison_df = comparison_df.resample('M', on='ds').mean()

	# Filter the monthly comparison dataframe to show only the last 12 months
	last_12_months_df = monthly_comparison_df.tail(12)

	# Calculate the percentage difference between actual and forecasted sales for the last 12 months
	last_12_months_df['percentage_difference'] = ((last_12_months_df['actual_sales'] - last_12_months_df['forecasted_sales']) / last_12_months_df['actual_sales']) * 100

	# Drop rows with NaN values in 'actual_sales'
	last_12_months_df = last_12_months_df.dropna(subset=['actual_sales'])

	# Plot the monthly comparison with bar charts and the percentage difference on top
	fig2, ax2 = plt.subplots(figsize=(14, 7))
	bar_width = 15 # Set the bar width

	# Plot actual sales bars
	bars_actual = ax2.bar(last_12_months_df.index, last_12_months_df['actual_sales'],
	width=bar_width, label='Actual Sales', color='black', alpha=0.7)

	# Plot forecasted sales bars
	bars_forecasted = ax2.bar(last_12_months_df.index + pd.DateOffset(days=15),
	last_12_months_df['forecasted_sales'], width=bar_width,
	label='Forecasted Sales', color='blue', alpha=0.7)

	# Add the percentage difference on top of the bars
	for bar, diff in zip(bars_actual, last_12_months_df['percentage_difference']):
	height = bar.get_height()
	ax2.text(bar.get_x() + bar.get_width() / 2.0, height, f'{diff:.2f}%', ha='center', va='bottom', color='red')

	ax2.set_title('Monthly Actual vs Forecasted Sales with Percentage Difference (Last 12 Months)')
	ax2.set_xlabel('Date')
	ax2.set_ylabel('Sales')
	ax2.legend()

	# Save the comparison dataframe to a CSV
	last_12_months_df.to_csv('forecast_comparison.csv')

	return fig1, fig2, 'forecast_comparison.csv'

	# Define the Gradio interface
	interface = gr.Interface(
	fn=forecast_sales,
	inputs=gr.File(label="Upload CSV File"),
	outputs=[
	gr.Plot(label="Sales Forecast"),
	gr.Plot(label="Monthly Comparison"),
	gr.File(label="Comparison CSV")
	],
	title="Sales Forecasting with Prophet",
	description="Upload a CSV file with sales data to forecast future sales and compare with actual sales."
	)

	# Launch the interface
	interface.launch(debug=True)