Upload 2 files

eda.py

@@ -0,0 +1,75 @@
+import pandas as pd
+import seaborn as sns
+import matplotlib.pyplot as plt
+import numpy as np
+import plotly.express as px
+from matplotlib.gridspec import GridSpec
+
+def average_sales_by_region(df):
+    """
+    Generate a bar plot for average sales by region.
+    """
+    df_bar = df[['region', 'sales']]
+    df_bar = df_bar.groupby('region').mean().sort_values(by='sales', ascending=False)
+    fig, ax = plt.subplots(figsize=[10, 6])
+    sns.barplot(x=df_bar.index, y='sales', data=df_bar, palette='viridis', ax=ax)
+    ax.set_title('Average Sales Across Different Regions')
+    ax.set_xlabel('Region')
+    ax.set_ylabel('Average Sales')
+    for index, value in enumerate(df_bar['sales']):
+        ax.text(index, value, f"{value:.2f}", ha='center', va='bottom')
+    return fig
+
+def average_sales_and_profit_over_time(df):
+    """
+    Generate a line plot for average sales and profit over time.
+    """
+    df_line = df[['order_date', 'sales', 'profit']].sort_values('order_date')
+    df_line['order_date'] = pd.to_datetime(df_line['order_date'])
+    df_line = df_line.groupby(df_line['order_date'].dt.to_period("M")).mean()
+    df_line.index = df_line.index.to_timestamp()
+    fig, ax = plt.subplots(figsize=[10, 6])
+    ax.plot(df_line.index, 'sales', data=df_line, color='green', label='Avg Sales')
+    ax.plot(df_line.index, 'profit', data=df_line, color='red', label='Avg Profit')
+    ax.legend()
+    ax.set_title('Average Sales and Profit Over Time (Monthly)')
+    ax.set_xlabel('Time')
+    ax.set_ylabel('Value')
+    return fig
+
+def segment_vs_region_distribution(df):
+    """
+    Generate a count plot for segments across different regions.
+    """
+    fig = plt.figure(figsize=(10, 6))
+    sns.countplot(x='segment', data=df, hue='region', palette='viridis')
+    plt.title('Segment vs. Region Distribution')
+    plt.xlabel('Segment')
+    plt.ylabel('Count')
+    plt.legend(title='Region')
+    return fig
+
+def sales_vs_profit_across_segments(df):
+    """
+    Generate a scatter plot comparing sales and profit across different customer segments.
+    """
+    fig, ax = plt.subplots(figsize=(10, 6))
+    sns.scatterplot(x='sales', y='profit', hue='segment', data=df, palette='viridis', size='sales', sizes=(20, 200), ax=ax)
+    ax.set_title('Sales vs. Profit Across Different Customer Segments')
+    ax.set_xlabel('Sales')
+    ax.set_ylabel('Profit')
+    return fig
+
+def category_composition_for_profit_and_sales(df):
+    """
+    Generate pie charts for the composition of category for profit and sales.
+    """
+    df_pie = df.groupby('category').agg({'sales': 'sum', 'profit': 'sum'}).reset_index()
+    fig, axs = plt.subplots(1, 2, figsize=(14, 7))
+    axs[0].pie(df_pie['sales'], labels=df_pie['category'], autopct='%1.1f%%', startangle=140, colors=['#ff9999','#66b3ff','#99ff99','#ffcc99'])
+    axs[0].set_title('Sales Composition by Category')
+    axs[1].pie(df_pie['profit'], labels=df_pie['category'], autopct='%1.1f%%', startangle=140, colors=['#ff9999','#66b3ff','#99ff99','#ffcc99'])
+    axs[1].set_title('Profit Composition by Category')
+    return fig
+
+# Additional EDA functions can be added following the same pattern

prediction.py

@@ -0,0 +1,31 @@
+# prediction.py
+
+import joblib
+import pandas as pd
+
+# Load the pipeline and model
+pipeline = joblib.load('full_pipeline_with_unit_price.pkl')
+model = joblib.load('best_model.pkl')
+
+def make_prediction(features):
+    """
+    Takes an array of original features, transforms it using the pipeline, 
+    and makes a prediction with the model.
+    
+    Parameters:
+    - features: array-like, shape [n_features]
+    
+    Returns:
+    - The predicted value as a float.
+    """
+    # Convert the features array into a DataFrame expected by the pipeline
+    # Note: You need to ensure the columns match those expected by the pipeline
+    features_df = pd.DataFrame([features], columns=['sales', 'quantity', 'discount', 'sub_category'])
+    
+    # Process features through the pipeline
+    processed_features = pipeline.transform(features_df)
+    
+    # Make a prediction
+    prediction = model.predict(processed_features)
+    
+    return prediction[0]  # Assuming we want a single prediction value