customer / app.py
entropy25's picture
Update app.py
e26889f verified
Raw History Blame Contribute Delete
51.1 kB
# -*- coding: utf-8 -*- import gradio as gr import pandas as pd import numpy as np import matplotlib.pyplot as plt import seaborn as sns from sklearn.model_selection import train_test_split, cross_val_score from sklearn.ensemble import RandomForestClassifier from sklearn.metrics import classification_report, confusion_matrix, accuracy_score, roc_auc_score, precision_recall_curve import plotly.express as px import plotly.graph_objects as go from plotly.subplots import make_subplots import plotly.io as pio from datetime import datetime, timedelta import io import base64 import warnings from typing import Optional, Tuple, Dict, Any warnings.filterwarnings('ignore') # Try importing optional dependencies try: import xgboost as xgb XGBOOST_AVAILABLE = True except ImportError: XGBOOST_AVAILABLE = False try: from reportlab.lib.pagesizes import letter, A4 from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer, Table, TableStyle from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle from reportlab.lib.units import inch from reportlab.lib import colors REPORTLAB_AVAILABLE = True except ImportError: REPORTLAB_AVAILABLE = False # Business configuration BUSINESS_CONFIG = { 'churn_threshold_days': 90, 'high_risk_probability': 0.7, 'rfm_quantiles': 5, 'min_customers_for_model': 10 } # UI color scheme COLORS = { 'primary': '#6366f1', 'success': '#10b981', 'warning': '#f59e0b', 'danger': '#ef4444', 'purple': '#8b5cf6', 'pink': '#ec4899', 'blue': '#3b82f6', 'indigo': '#6366f1' } class DataProcessor: """Handles data loading, validation, and preprocessing""" @staticmethod def load_and_validate(file) -> Tuple[Optional[pd.DataFrame], str]: """Load and validate CSV file""" if file is None: return None, "Please upload a CSV file" try: df = pd.read_csv(file.name) # Flexible column mapping column_mapping = DataProcessor._map_columns(df.columns) if not column_mapping: return None, f"Required columns not found. Available: {list(df.columns)}" df = df.rename(columns=column_mapping) # Clean and validate data initial_rows = len(df) df = DataProcessor._clean_data(df) final_rows = len(df) if final_rows == 0: return None, "No valid data after cleaning" status = f"Data loaded successfully! {final_rows} records from {df['customer_id'].nunique()} customers" if initial_rows != final_rows: status += f" ({initial_rows - final_rows} invalid rows removed)" return df, status except Exception as e: return None, f"Error loading data: {str(e)}" @staticmethod def _map_columns(columns) -> Dict[str, str]: """Map CSV columns to standard names""" required = ['customer_id', 'order_date', 'amount'] mapping = {} column_variations = { 'customer_id': ['customer', 'cust_id', 'id', 'customerid', 'client_id', 'customer_id'], 'order_date': ['date', 'order_date', 'orderdate', 'purchase_date', 'transaction_date'], 'amount': ['revenue', 'value', 'price', 'total', 'sales', 'order_value', 'amount'] } for req_col in required: found = False for col in columns: col_lower = col.lower().strip() if col_lower == req_col or any(var in col_lower for var in column_variations[req_col]): mapping[col] = req_col found = True break if not found: return {} return mapping @staticmethod def _clean_data(df: pd.DataFrame) -> pd.DataFrame: """Clean and prepare data""" df = df.copy() df['customer_id'] = df['customer_id'].astype(str) df['order_date'] = pd.to_datetime(df['order_date'], errors='coerce') df['amount'] = pd.to_numeric(df['amount'], errors='coerce') # Remove invalid rows df = df.dropna(subset=['customer_id', 'order_date', 'amount']) df = df[df['amount'] > 0] # Remove negative amounts return df class RFMAnalyzer: """Handles RFM analysis and customer metrics calculation""" @staticmethod def calculate_rfm_metrics(df: pd.DataFrame) -> pd.DataFrame: """Calculate RFM metrics for customers""" current_date = df['order_date'].max() + timedelta(days=1) customer_metrics = df.groupby('customer_id').agg({ 'order_date': ['max', 'count', 'min'], 'amount': ['sum', 'mean', 'std', 'min', 'max'] }) # Flatten column names customer_metrics.columns = [ 'last_order_date', 'frequency', 'first_order_date', 'monetary', 'avg_order_value', 'std_amount', 'min_amount', 'max_amount' ] # Calculate additional features customer_metrics['recency_days'] = (current_date - customer_metrics['last_order_date']).dt.days customer_metrics['customer_lifetime_days'] = ( customer_metrics['last_order_date'] - customer_metrics['first_order_date'] ).dt.days customer_metrics['std_amount'] = customer_metrics['std_amount'].fillna(0) customer_metrics['customer_lifetime_days'] = customer_metrics['customer_lifetime_days'].fillna(0) return customer_metrics.reset_index() class CustomerSegmenter: """Handles customer segmentation based on RFM analysis""" @staticmethod def perform_segmentation(customer_metrics: pd.DataFrame) -> pd.DataFrame: """Segment customers using RFM scores""" df = customer_metrics.copy() # Calculate RFM scores if len(df) >= BUSINESS_CONFIG['rfm_quantiles']: try: df['R_Score'] = pd.qcut(df['recency_days'], BUSINESS_CONFIG['rfm_quantiles'], labels=[5,4,3,2,1], duplicates='drop') df['F_Score'] = pd.qcut(df['frequency'], BUSINESS_CONFIG['rfm_quantiles'], labels=[1,2,3,4,5], duplicates='drop') df['M_Score'] = pd.qcut(df['monetary'], BUSINESS_CONFIG['rfm_quantiles'], labels=[1,2,3,4,5], duplicates='drop') except ValueError: # Fallback for small datasets df['R_Score'] = pd.cut(df['recency_days'], bins=BUSINESS_CONFIG['rfm_quantiles'], labels=[5,4,3,2,1], include_lowest=True) df['F_Score'] = pd.cut(df['frequency'], bins=BUSINESS_CONFIG['rfm_quantiles'], labels=[1,2,3,4,5], include_lowest=True) df['M_Score'] = pd.cut(df['monetary'], bins=BUSINESS_CONFIG['rfm_quantiles'], labels=[1,2,3,4,5], include_lowest=True) else: df['R_Score'] = 3 df['F_Score'] = 3 df['M_Score'] = 3 # Convert to numeric and handle NaN for col in ['R_Score', 'F_Score', 'M_Score']: df[col] = pd.to_numeric(df[col], errors='coerce').fillna(3).astype(int) # Apply segmentation logic df['Segment'] = df.apply(CustomerSegmenter._assign_segment, axis=1) df['Churn_Risk'] = df.apply(CustomerSegmenter._assign_risk_level, axis=1) return df @staticmethod def _assign_segment(row) -> str: """Assign customer segment based on RFM scores""" r, f, m = row['R_Score'], row['F_Score'], row['M_Score'] if r >= 4 and f >= 4 and m >= 4: return 'Champions' elif r >= 3 and f >= 3 and m >= 3: return 'Loyal Customers' elif r >= 3 and f >= 2: return 'Potential Loyalists' elif r >= 4 and f <= 2: return 'New Customers' elif r <= 2 and f >= 3: return 'At Risk' elif r <= 2 and f <= 2 and m >= 3: return 'Cannot Lose Them' elif r <= 2 and f <= 2 and m <= 2: return 'Lost Customers' else: return 'Others' @staticmethod def _assign_risk_level(row) -> str: """Assign churn risk level""" segment = CustomerSegmenter._assign_segment(row) if segment in ['Lost Customers', 'At Risk']: return 'High' elif segment in ['Others', 'Cannot Lose Them']: return 'Medium' else: return 'Low' class ChurnPredictor: """Handles churn prediction model training and inference""" def __init__(self): self.model = None self.feature_importance = None self.model_metrics = {} def train_model(self, customer_metrics: pd.DataFrame) -> Tuple[bool, str, Dict]: """Train churn prediction model""" if len(customer_metrics) < BUSINESS_CONFIG['min_customers_for_model']: return False, f"Insufficient data for training (minimum {BUSINESS_CONFIG['min_customers_for_model']} customers required)", {} # Prepare features feature_cols = [ 'recency_days', 'frequency', 'monetary', 'avg_order_value', 'std_amount', 'min_amount', 'max_amount', 'customer_lifetime_days' ] X = customer_metrics[feature_cols] y = (customer_metrics['recency_days'] > BUSINESS_CONFIG['churn_threshold_days']).astype(int) # Check for sufficient class diversity if y.nunique() < 2: return False, "Cannot train model: all customers have the same churn status", {} # Train-test split X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.2, random_state=42, stratify=y ) # Select and train model if XGBOOST_AVAILABLE: try: self.model = xgb.XGBClassifier(random_state=42, eval_metric='logloss') model_name = "XGBoost Classifier" except: self.model = RandomForestClassifier(random_state=42, n_estimators=100) model_name = "Random Forest Classifier" else: self.model = RandomForestClassifier(random_state=42, n_estimators=100) model_name = "Random Forest Classifier" self.model.fit(X_train, y_train) # Evaluate model y_pred = self.model.predict(X_test) y_pred_proba = self.model.predict_proba(X_test)[:, 1] accuracy = accuracy_score(y_test, y_pred) auc_score = roc_auc_score(y_test, y_pred_proba) # Cross-validation cv_scores = cross_val_score(self.model, X, y, cv=5, scoring='roc_auc') # Feature importance self.feature_importance = pd.DataFrame({ 'feature': feature_cols, 'importance': self.model.feature_importances_ }).sort_values('importance', ascending=False) self.model_metrics = { 'accuracy': accuracy, 'auc_score': auc_score, 'cv_mean': cv_scores.mean(), 'cv_std': cv_scores.std(), 'model_name': model_name, 'n_features': len(feature_cols), 'n_samples': len(X_train) } return True, "Model trained successfully", self.model_metrics def predict(self, customer_metrics: pd.DataFrame) -> pd.DataFrame: """Make churn predictions""" if self.model is None: return customer_metrics feature_cols = [ 'recency_days', 'frequency', 'monetary', 'avg_order_value', 'std_amount', 'min_amount', 'max_amount', 'customer_lifetime_days' ] X = customer_metrics[feature_cols] predictions = self.model.predict_proba(X)[:, 1] result = customer_metrics.copy() result['churn_probability'] = predictions result['predicted_churn'] = (predictions > BUSINESS_CONFIG['high_risk_probability']).astype(int) return result class VisualizationEngine: """Handles all chart creation and visualization""" @staticmethod def create_segment_chart(customer_data: pd.DataFrame): """Create customer segment distribution chart""" segment_counts = customer_data['Segment'].value_counts().reset_index() segment_counts.columns = ['Segment', 'Count'] fig = px.pie( segment_counts, values='Count', names='Segment', title='Customer Segment Distribution', hole=0.4, color_discrete_sequence=list(COLORS.values()) ) fig.update_traces(textposition='inside', textinfo='percent+label') fig.update_layout(height=400, title={'x': 0.5, 'xanchor': 'center'}) return fig @staticmethod def create_rfm_scatter(customer_data: pd.DataFrame): """Create RFM analysis scatter plot""" fig = px.scatter( customer_data, x='recency_days', y='frequency', size='monetary', color='Segment', title='RFM Customer Behavior Matrix', labels={ 'recency_days': 'Days Since Last Purchase', 'frequency': 'Purchase Frequency', 'monetary': 'Total Revenue' }, color_discrete_sequence=list(COLORS.values()) ) fig.update_layout(height=400, title={'x': 0.5, 'xanchor': 'center'}) return fig @staticmethod def create_churn_chart(customer_data: pd.DataFrame, has_predictions: bool = False): """Create churn risk visualization""" if has_predictions and 'churn_probability' in customer_data.columns: fig = px.histogram( customer_data, x='churn_probability', nbins=20, title='Churn Probability Distribution', labels={'churn_probability': 'Churn Probability', 'count': 'Number of Customers'}, color_discrete_sequence=[COLORS['primary']] ) fig.add_vline(x=BUSINESS_CONFIG['high_risk_probability'], line_dash="dash", line_color=COLORS['danger'], annotation_text="High Risk Threshold") else: risk_counts = customer_data['Churn_Risk'].value_counts().reset_index() risk_counts.columns = ['Risk_Level', 'Count'] colors_map = {'High': COLORS['danger'], 'Medium': COLORS['warning'], 'Low': COLORS['success']} fig = px.bar( risk_counts, x='Risk_Level', y='Count', title='Customer Churn Risk Distribution', color='Risk_Level', color_discrete_map=colors_map ) fig.update_layout(showlegend=False) fig.update_layout(height=400, title={'x': 0.5, 'xanchor': 'center'}) return fig @staticmethod def create_revenue_trend(df: pd.DataFrame, time_granularity='month', customer_filter='all'): """Create revenue trend visualization with filters""" df_copy = df.copy() # Filter by customer if specified if customer_filter != 'all' and customer_filter: df_copy = df_copy[df_copy['customer_id'] == customer_filter] if df_copy.empty: # Return empty chart with message fig = go.Figure() fig.add_annotation(text=f"No data found for customer {customer_filter}", xref="paper", yref="paper", x=0.5, y=0.5, showarrow=False) fig.update_layout(title=f"Revenue Trend - {customer_filter}", height=400) return fig # Group by time granularity if time_granularity == 'day': df_copy['time_period'] = df_copy['order_date'].dt.date title_suffix = "Daily" elif time_granularity == 'week': df_copy['time_period'] = df_copy['order_date'].dt.to_period('W') title_suffix = "Weekly" elif time_granularity == 'year': df_copy['time_period'] = df_copy['order_date'].dt.to_period('Y') title_suffix = "Yearly" else: # default to month df_copy['time_period'] = df_copy['order_date'].dt.to_period('M') title_suffix = "Monthly" revenue_data = df_copy.groupby('time_period')['amount'].sum().reset_index() revenue_data['time_period'] = revenue_data['time_period'].astype(str) # Create title if customer_filter == 'all': title = f"{title_suffix} Revenue Trends - All Customers" else: title = f"{title_suffix} Revenue Trends - {customer_filter}" fig = px.line( revenue_data, x='time_period', y='amount', title=title, labels={'amount': 'Revenue ($)', 'time_period': 'Time Period'} ) fig.update_traces(line_color=COLORS['primary'], line_width=3) fig.update_layout(height=400, title={'x': 0.5, 'xanchor': 'center'}) return fig @staticmethod def create_feature_importance_chart(feature_importance: pd.DataFrame): """Create feature importance chart""" fig = px.bar( feature_importance.head(8), x='importance', y='feature', orientation='h', title='Feature Importance Analysis', labels={'importance': 'Importance Score', 'feature': 'Features'}, color='importance', color_continuous_scale='viridis' ) fig.update_layout( height=500, showlegend=False, plot_bgcolor='white', paper_bgcolor='white', title={'x': 0.5, 'xanchor': 'center'}, yaxis={'categoryorder': 'total ascending'} ) return fig class ReportGenerator: """Handles report generation""" @staticmethod def generate_pdf_report(customer_data: pd.DataFrame, model_metrics: Dict) -> bytes: """Generate PDF report""" if not REPORTLAB_AVAILABLE: raise ImportError("PDF generation requires ReportLab library") buffer = io.BytesIO() doc = SimpleDocTemplate(buffer, pagesize=A4, rightMargin=72, leftMargin=72, topMargin=72, bottomMargin=18) styles = getSampleStyleSheet() story = [] # Title title_style = ParagraphStyle('CustomTitle', parent=styles['Title'], fontSize=24, spaceAfter=30, alignment=1) story.append(Paragraph("B2B Customer Analytics Report", title_style)) story.append(Spacer(1, 12)) # Executive summary story.append(Paragraph("Executive Summary", styles['Heading2'])) total_customers = len(customer_data) total_revenue = customer_data['monetary'].sum() avg_revenue = customer_data['monetary'].mean() summary_text = f""" This comprehensive analysis covers {total_customers:,} customers with combined revenue of ${total_revenue:,.2f}. The average customer value is ${avg_revenue:,.2f}. Customer segmentation and churn risk assessment have been performed using advanced RFM analysis and machine learning techniques. """ story.append(Paragraph(summary_text, styles['Normal'])) story.append(Spacer(1, 20)) # Segment distribution story.append(Paragraph("Customer Segmentation Overview", styles['Heading2'])) segment_dist = customer_data['Segment'].value_counts() segment_data = [] segment_data.append(['Segment', 'Count', 'Percentage']) for segment, count in segment_dist.items(): percentage = (count / total_customers) * 100 segment_data.append([segment, str(count), f"{percentage:.1f}%"]) segment_table = Table(segment_data) segment_table.setStyle(TableStyle([ ('BACKGROUND', (0, 0), (-1, 0), colors.grey), ('TEXTCOLOR', (0, 0), (-1, 0), colors.whitesmoke), ('ALIGN', (0, 0), (-1, -1), 'CENTER'), ('FONTNAME', (0, 0), (-1, 0), 'Helvetica-Bold'), ('FONTSIZE', (0, 0), (-1, 0), 14), ('BOTTOMPADDING', (0, 0), (-1, 0), 12), ('BACKGROUND', (0, 1), (-1, -1), colors.beige), ('GRID', (0, 0), (-1, -1), 1, colors.black) ])) story.append(segment_table) story.append(Spacer(1, 20)) # Model performance (if available) if model_metrics: story.append(Paragraph("Churn Prediction Model Performance", styles['Heading2'])) model_text = f""" Model Type: {model_metrics['model_name']}<br/> Accuracy: {model_metrics['accuracy']:.1%}<br/> AUC Score: {model_metrics['auc_score']:.3f}<br/> Cross-validation Score: {model_metrics['cv_mean']:.3f} ± {model_metrics['cv_std']:.3f}<br/> Features Used: {model_metrics['n_features']}<br/> Training Samples: {model_metrics['n_samples']} """ story.append(Paragraph(model_text, styles['Normal'])) # Build and return PDF doc.build(story) pdf_bytes = buffer.getvalue() buffer.close() return pdf_bytes class B2BCustomerAnalytics: """Main analytics orchestrator""" def __init__(self): self.raw_data = None self.customer_metrics = None self.churn_predictor = ChurnPredictor() self.has_trained_model = False def load_data(self, file) -> Tuple[str, str, Optional[pd.DataFrame]]: """Load and process data""" self.raw_data, status = DataProcessor.load_and_validate(file) if self.raw_data is not None: # Calculate RFM metrics self.customer_metrics = RFMAnalyzer.calculate_rfm_metrics(self.raw_data) # Perform segmentation self.customer_metrics = CustomerSegmenter.perform_segmentation(self.customer_metrics) # Generate dashboard dashboard_html = self._generate_dashboard() preview_data = self._prepare_preview_data() return status, dashboard_html, preview_data return status, "", None def train_churn_model(self) -> Tuple[str, Optional[Any]]: """Train churn prediction model""" if self.customer_metrics is None: return "No data available. Please upload data first.", None success, message, metrics = self.churn_predictor.train_model(self.customer_metrics) if success: self.has_trained_model = True # Update predictions self.customer_metrics = self.churn_predictor.predict(self.customer_metrics) results_html = self._format_model_results(metrics) chart = VisualizationEngine.create_feature_importance_chart( self.churn_predictor.feature_importance ) return results_html, chart return f"Model training failed: {message}", None def get_visualizations(self) -> Tuple[Any, Any, Any, Any]: """Get all visualizations""" if self.customer_metrics is None: return None, None, None, None segment_chart = VisualizationEngine.create_segment_chart(self.customer_metrics) rfm_chart = VisualizationEngine.create_rfm_scatter(self.customer_metrics) churn_chart = VisualizationEngine.create_churn_chart( self.customer_metrics, self.has_trained_model ) revenue_chart = VisualizationEngine.create_revenue_trend(self.raw_data) return segment_chart, rfm_chart, churn_chart, revenue_chart def get_revenue_chart_with_filters(self, time_granularity='month', customer_filter='all'): """Get revenue chart with time and customer filters""" if self.raw_data is None: return None return VisualizationEngine.create_revenue_trend( self.raw_data, time_granularity, customer_filter ) def get_customer_list(self): """Get list of customer IDs for dropdown""" if self.raw_data is None: return [] return ['all'] + sorted(self.raw_data['customer_id'].unique().tolist()) def get_customer_table(self) -> Optional[pd.DataFrame]: """Get formatted customer table""" if self.customer_metrics is None: return None columns = ['customer_id', 'Segment', 'Churn_Risk', 'recency_days', 'frequency', 'monetary', 'avg_order_value'] if 'churn_probability' in self.customer_metrics.columns: columns.append('churn_probability') self.customer_metrics['churn_probability'] = ( self.customer_metrics['churn_probability'] * 100 ).round(1) table_data = self.customer_metrics[columns].copy() table_data['monetary'] = table_data['monetary'].round(2) table_data['avg_order_value'] = table_data['avg_order_value'].round(2) # Rename columns for display display_names = { 'customer_id': 'Customer ID', 'Segment': 'Segment', 'Churn_Risk': 'Risk Level', 'recency_days': 'Recency (Days)', 'frequency': 'Frequency', 'monetary': 'Total Spent ($)', 'avg_order_value': 'Avg Order ($)', 'churn_probability': 'Churn Probability (%)' } table_data = table_data.rename(columns=display_names) return table_data.head(50) def get_customer_insights(self, customer_id: str) -> str: """Get detailed customer insights""" if self.customer_metrics is None or not customer_id: return "Please enter a valid customer ID" customer_data = self.customer_metrics[ self.customer_metrics['customer_id'] == customer_id ] if customer_data.empty: return f"Customer {customer_id} not found" customer = customer_data.iloc[0] return self._format_customer_profile(customer) def generate_report(self) -> bytes: """Generate PDF report""" if self.customer_metrics is None: raise ValueError("No data available for report generation") return ReportGenerator.generate_pdf_report( self.customer_metrics, self.churn_predictor.model_metrics ) def _generate_dashboard(self) -> str: """Generate dashboard HTML""" total_customers = len(self.customer_metrics) total_revenue = self.customer_metrics['monetary'].sum() avg_order_value = self.customer_metrics['avg_order_value'].mean() high_risk_customers = (self.customer_metrics['Churn_Risk'] == 'High').sum() segment_dist = self.customer_metrics['Segment'].value_counts() return f""" <div style="display: flex; flex-wrap: wrap; gap: 1rem; margin-bottom: 2rem;"> <div style="flex: 1; min-width: 200px; background: linear-gradient(135deg, #3b82f6, #1d4ed8); padding: 1.5rem; border-radius: 12px; color: white; text-align: center;"> <h3 style="margin: 0 0 0.5rem 0; font-size: 0.9rem; opacity: 0.9;">Total Customers</h3> <div style="font-size: 2.5rem; font-weight: bold;">{total_customers:,}</div> </div> <div style="flex: 1; min-width: 200px; background: linear-gradient(135deg, #10b981, #047857); padding: 1.5rem; border-radius: 12px; color: white; text-align: center;"> <h3 style="margin: 0 0 0.5rem 0; font-size: 0.9rem; opacity: 0.9;">Total Revenue</h3> <div style="font-size: 2.5rem; font-weight: bold;">${total_revenue/1000000:.1f}M</div> </div> <div style="flex: 1; min-width: 200px; background: linear-gradient(135deg, #8b5cf6, #6d28d9); padding: 1.5rem; border-radius: 12px; color: white; text-align: center;"> <h3 style="margin: 0 0 0.5rem 0; font-size: 0.9rem; opacity: 0.9;">Avg Order Value</h3> <div style="font-size: 2.5rem; font-weight: bold;">${avg_order_value:.0f}</div> </div> <div style="flex: 1; min-width: 200px; background: linear-gradient(135deg, #ef4444, #dc2626); padding: 1.5rem; border-radius: 12px; color: white; text-align: center;"> <h3 style="margin: 0 0 0.5rem 0; font-size: 0.9rem; opacity: 0.9;">High Risk Customers</h3> <div style="font-size: 2.5rem; font-weight: bold;">{high_risk_customers}</div> </div> </div> <div style="background: #f8fafc; padding: 1.5rem; border-radius: 12px; border-left: 4px solid #6366f1;"> <h4 style="margin: 0 0 1rem 0; color: #374151;">Customer Segments Overview</h4> <div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(150px, 1fr)); gap: 1rem;"> {' '.join([f'<div><strong>{segment}:</strong> {count}</div>' for segment, count in segment_dist.items()])} </div> </div> """ def _prepare_preview_data(self) -> pd.DataFrame: """Prepare data preview""" if self.raw_data is None: return pd.DataFrame() preview = self.raw_data.merge( self.customer_metrics[['customer_id', 'Segment', 'Churn_Risk']], on='customer_id', how='left' ) return preview.head(20) def _format_model_results(self, metrics: Dict) -> str: """Format model training results""" return f""" <div style="background: white; padding: 2rem; border-radius: 1rem; box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1); margin-bottom: 2rem;"> <div style="text-align: center; margin-bottom: 2rem;"> <h3 style="color: #1f2937; font-size: 1.5rem; font-weight: bold; margin-bottom: 0.5rem;"> Model Training Completed Successfully </h3> <p style="color: #6b7280;">{metrics['model_name']} with Advanced Feature Engineering</p> </div> <div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(150px, 1fr)); gap: 1rem; margin-bottom: 2rem;"> <div style="background: linear-gradient(135deg, #6366f1, #4f46e5); padding: 1rem; border-radius: 8px; text-align: center; color: white;"> <div style="font-size: 2rem; font-weight: bold;">{metrics['accuracy']:.1%}</div> <div style="font-size: 0.9rem;">Accuracy</div> </div> <div style="background: linear-gradient(135deg, #10b981, #059669); padding: 1rem; border-radius: 8px; text-align: center; color: white;"> <div style="font-size: 2rem; font-weight: bold;">{metrics['auc_score']:.3f}</div> <div style="font-size: 0.9rem;">AUC Score</div> </div> <div style="background: linear-gradient(135deg, #f59e0b, #d97706); padding: 1rem; border-radius: 8px; text-align: center; color: white;"> <div style="font-size: 2rem; font-weight: bold;">{metrics['n_features']}</div> <div style="font-size: 0.9rem;">Features Used</div> </div> <div style="background: linear-gradient(135deg, #8b5cf6, #7c3aed); padding: 1rem; border-radius: 8px; text-align: center; color: white;"> <div style="font-size: 2rem; font-weight: bold;">{metrics['cv_mean']:.3f}</div> <div style="font-size: 0.9rem;">CV Score</div> </div> </div> </div> """ def _format_customer_profile(self, customer) -> str: """Format individual customer profile""" churn_prob = customer.get('churn_probability', 0.5) recommendations = self._get_customer_recommendations( customer['Segment'], customer['Churn_Risk'], churn_prob, customer['recency_days'] ) return f""" <div style="background: white; padding: 2rem; border-radius: 1rem; box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1); margin-bottom: 1rem;"> <h3 style="text-align: center; color: #1f2937; margin-bottom: 1.5rem;">Customer Profile: {customer['customer_id']}</h3> <div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(200px, 1fr)); gap: 1rem; margin-bottom: 2rem;"> <div style="background: linear-gradient(135deg, #6366f1, #4f46e5); padding: 1rem; border-radius: 8px; color: white; text-align: center;"> <h4 style="margin: 0 0 0.5rem 0; font-size: 0.9rem; opacity: 0.9;">Segment</h4> <div style="font-size: 1.2rem; font-weight: bold;">{customer['Segment']}</div> </div> <div style="background: linear-gradient(135deg, #ef4444, #dc2626); padding: 1rem; border-radius: 8px; color: white; text-align: center;"> <h4 style="margin: 0 0 0.5rem 0; font-size: 0.9rem; opacity: 0.9;">Churn Risk</h4> <div style="font-size: 1.2rem; font-weight: bold;">{customer['Churn_Risk']}</div> </div> <div style="background: linear-gradient(135deg, #8b5cf6, #6d28d9); padding: 1rem; border-radius: 8px; color: white; text-align: center;"> <h4 style="margin: 0 0 0.5rem 0; font-size: 0.9rem; opacity: 0.9;">Churn Probability</h4> <div style="font-size: 1.2rem; font-weight: bold;">{churn_prob:.1%}</div> </div> </div> <div style="background: #f8fafc; padding: 1.5rem; border-radius: 8px; margin-bottom: 1rem;"> <h4 style="color: #374151; margin-bottom: 1rem;">Transaction Analytics</h4> <div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(150px, 1fr)); gap: 1rem;"> <div> <div style="font-size: 0.8rem; color: #6b7280; margin-bottom: 0.2rem;">Purchase Frequency</div> <div style="font-size: 1.5rem; font-weight: bold; color: #1f2937;">{customer['frequency']}</div> </div> <div> <div style="font-size: 0.8rem; color: #6b7280; margin-bottom: 0.2rem;">Total Spent</div> <div style="font-size: 1.5rem; font-weight: bold; color: #1f2937;">${customer['monetary']:,.0f}</div> </div> <div> <div style="font-size: 0.8rem; color: #6b7280; margin-bottom: 0.2rem;">Avg Order Value</div> <div style="font-size: 1.5rem; font-weight: bold; color: #1f2937;">${customer['avg_order_value']:.0f}</div> </div> <div> <div style="font-size: 0.8rem; color: #6b7280; margin-bottom: 0.2rem;">Days Since Last Order</div> <div style="font-size: 1.5rem; font-weight: bold; color: #1f2937;">{customer['recency_days']}</div> </div> </div> </div> <div style="background: linear-gradient(135deg, #f0f9ff, #e0f2fe); border-left: 4px solid #3b82f6; padding: 1rem; border-radius: 4px;"> <h4 style="color: #1e40af; margin-bottom: 0.5rem;">Recommendations</h4> <p style="color: #1f2937; margin: 0;">{recommendations}</p> </div> </div> """ def _get_customer_recommendations(self, segment: str, risk_level: str, churn_prob: float, recency: int) -> str: """Generate personalized recommendations""" recommendations = [] if risk_level == 'High' or churn_prob > BUSINESS_CONFIG['high_risk_probability']: recommendations.append("URGENT: Personal outreach required within 24 hours") recommendations.append("Offer retention incentive or loyalty program") elif risk_level == 'Medium': recommendations.append("Send personalized re-engagement campaign") if segment == 'Champions': recommendations.append("Invite to VIP program or advisory board") elif segment == 'At Risk': recommendations.append("Proactive customer success intervention needed") elif segment == 'New Customers': recommendations.append("Deploy onboarding campaign sequence") elif segment == 'Lost Customers': recommendations.append("Win-back campaign with deep discount offer") if recency > 60: recommendations.append("Re-engagement campaign with special offer recommended") return " • ".join(recommendations) if recommendations else "Continue monitoring customer engagement patterns." def update_revenue_chart(analytics_instance, time_gran, customer_id): """Update revenue chart based on filters""" try: chart = analytics_instance.get_revenue_chart_with_filters(time_gran, customer_id) return chart except Exception as e: return None def update_customer_dropdown(analytics_instance): """Update customer dropdown options""" try: customers = analytics_instance.get_customer_list() return gr.Dropdown(choices=customers, value='all') except: return gr.Dropdown(choices=['all'], value='all') def create_gradio_interface(): """Create the enhanced Gradio interface""" # Custom CSS for modern styling custom_css = """ .gradio-container { font-family: 'Inter', system-ui, sans-serif !important; max-width: 1200px !important; } .tab-nav { background: #f8fafc !important; border-radius: 8px !important; } """ with gr.Blocks(theme=gr.themes.Soft(), title="B2B Customer Analytics", css=custom_css) as demo: # Initialize analytics instance per session analytics = gr.State(B2BCustomerAnalytics()) gr.HTML(""" <div style="background: linear-gradient(135deg, #6366f1 0%, #8b5cf6 100%); padding: 2rem; border-radius: 1rem; color: white; text-align: center; margin-bottom: 2rem;"> <h1 style="font-size: 2.5rem; font-weight: bold; margin-bottom: 0.5rem;">B2B Customer Analytics Platform</h1> <p style="font-size: 1.1rem; opacity: 0.9;">Advanced Customer Segmentation & Churn Prediction</p> <div style="font-size: 0.9rem; opacity: 0.8; margin-top: 1rem;"> Upload your customer data CSV with columns: customer_id, order_date, amount (or similar) </div> </div> """) with gr.Tabs(): with gr.Tab("Data Upload & Dashboard"): with gr.Row(): with gr.Column(): file_input = gr.File( label="Upload Customer Data CSV", file_types=[".csv"], type="filepath" ) load_btn = gr.Button( "Load & Process Data", variant="primary", size="lg" ) load_status = gr.Textbox( label="Status", interactive=False, max_lines=2 ) summary_display = gr.HTML() data_preview = gr.DataFrame(label="Data Preview (First 20 Rows)") with gr.Tab("Customer Segmentation"): with gr.Row(): with gr.Column(): segment_chart = gr.Plot(label="Customer Segments Distribution") with gr.Column(): rfm_chart = gr.Plot(label="RFM Behavior Analysis") customer_table = gr.DataFrame(label="Customer Segmentation Details") gr.HTML(""" <div style="background: #f0f9ff; padding: 1rem; border-radius: 8px; border-left: 4px solid #3b82f6; margin-top: 1rem;"> <h4 style="color: #1e40af; margin: 0 0 0.5rem 0;">Segment Definitions</h4> <p style="margin: 0; color: #1f2937; font-size: 0.9rem;"> <strong>Champions:</strong> High value, frequent customers • <strong>Loyal Customers:</strong> Regular, valuable customers • <strong>At Risk:</strong> Previously valuable but declining activity • <strong>Lost Customers:</strong> Haven't purchased recently </p> </div> """) with gr.Tab("Churn Prediction"): train_btn = gr.Button( "Train Churn Prediction Model", variant="primary", size="lg" ) model_results = gr.HTML() with gr.Row(): with gr.Column(): feature_importance_chart = gr.Plot(label="Feature Importance Analysis") with gr.Column(): churn_distribution_chart = gr.Plot(label="Churn Risk Distribution") gr.HTML(""" <div style="background: #fef3c7; padding: 1rem; border-radius: 8px; border-left: 4px solid #f59e0b; margin-top: 1rem;"> <h4 style="color: #92400e; margin: 0 0 0.5rem 0;">Model Information</h4> <p style="margin: 0; color: #1f2937; font-size: 0.9rem;"> The model uses advanced features including customer lifetime, purchase patterns, and RFM metrics. Customers with >90 days since last purchase are considered churned for training purposes. </p> </div> """) with gr.Tab("Revenue Analytics"): with gr.Row(): with gr.Column(scale=1): time_granularity = gr.Radio( choices=['day', 'week', 'month', 'year'], value='month', label="Time Granularity" ) customer_filter = gr.Dropdown( choices=[], # Will be populated dynamically value='all', label="Customer Filter" ) update_chart_btn = gr.Button("Update Chart", variant="primary") with gr.Column(scale=3): revenue_chart = gr.Plot(label="Revenue Trends") # Add the info box gr.HTML(""" <div style="background: #ecfdf5; padding: 1rem; border-radius: 8px; border-left: 4px solid #10b981; margin-top: 1rem;"> <h4 style="color: #065f46; margin: 0 0 0.5rem 0;">Interactive Revenue Analysis</h4> <p style="margin: 0; color: #1f2937; font-size: 0.9rem;"> Select time granularity (day/week/month/year) and specific customers to analyze revenue patterns. Use "all" to view aggregate trends across all customers. </p> </div> """) with gr.Tab("Customer Insights"): with gr.Row(): customer_id_input = gr.Textbox( label="Customer ID", placeholder="Enter customer ID for detailed analysis", scale=3 ) insights_btn = gr.Button( "Get Customer Profile", variant="primary", scale=1 ) customer_insights = gr.HTML() with gr.Tab("Reports"): with gr.Row(): with gr.Column(): gr.HTML(""" <div style="background: white; padding: 2rem; border-radius: 1rem; box-shadow: 0 2px 4px rgba(0,0,0,0.1);"> <h3 style="color: #1f2937; margin-bottom: 1rem;">Generate Comprehensive Report</h3> <p style="color: #6b7280; margin-bottom: 1.5rem;"> Create a detailed PDF report including customer segmentation analysis, churn predictions, and actionable business insights. </p> </div> """) report_btn = gr.Button( "Generate PDF Report", variant="primary", size="lg" ) with gr.Column(): report_file = gr.File( label="Download Report", interactive=False ) # Event handlers with proper error handling def safe_load_data(analytics_instance, file): try: if file is None: return analytics_instance, "Please upload a CSV file", "", None, None, None, None, None, None, gr.Dropdown(choices=['all'], value='all') status, dashboard, preview = analytics_instance.load_data(file) if "successfully" in status: charts = analytics_instance.get_visualizations() table = analytics_instance.get_customer_table() # Update customer dropdown customers = analytics_instance.get_customer_list() customer_dropdown = gr.Dropdown(choices=customers, value='all') return analytics_instance, status, dashboard, preview, *charts, table, customer_dropdown else: return analytics_instance, status, "", None, None, None, None, None, None, gr.Dropdown(choices=['all'], value='all') except Exception as e: error_msg = f"Error loading data: {str(e)}" return analytics_instance, error_msg, "", None, None, None, None, None, None, gr.Dropdown(choices=['all'], value='all') def safe_train_model(analytics_instance): try: result_html, chart = analytics_instance.train_churn_model() # Update churn chart after training updated_charts = analytics_instance.get_visualizations() return analytics_instance, result_html, chart, updated_charts[2] except Exception as e: error_msg = f"Error training model: {str(e)}" return analytics_instance, error_msg, None, None def safe_get_insights(analytics_instance, customer_id): try: return analytics_instance.get_customer_insights(customer_id) except Exception as e: return f"Error getting insights: {str(e)}" def safe_generate_report(analytics_instance): try: if analytics_instance.customer_metrics is None: return None pdf_bytes = analytics_instance.generate_report() # Save to temporary file import tempfile with tempfile.NamedTemporaryFile(delete=False, suffix='.pdf') as tmp: tmp.write(pdf_bytes) return tmp.name except Exception as e: gr.Warning(f"Error generating report: {str(e)}") return None # Wire up events load_btn.click( fn=safe_load_data, inputs=[analytics, file_input], outputs=[analytics, load_status, summary_display, data_preview, segment_chart, rfm_chart, churn_distribution_chart, revenue_chart, customer_table, customer_filter] ) # Update chart when filters change update_chart_btn.click( fn=update_revenue_chart, inputs=[analytics, time_granularity, customer_filter], outputs=[revenue_chart] ) # Auto-update on filter change time_granularity.change( fn=update_revenue_chart, inputs=[analytics, time_granularity, customer_filter], outputs=[revenue_chart] ) customer_filter.change( fn=update_revenue_chart, inputs=[analytics, time_granularity, customer_filter], outputs=[revenue_chart] ) train_btn.click( fn=safe_train_model, inputs=[analytics], outputs=[analytics, model_results, feature_importance_chart, churn_distribution_chart] ) insights_btn.click( fn=safe_get_insights, inputs=[analytics, customer_id_input], outputs=[customer_insights] ) report_btn.click( fn=safe_generate_report, inputs=[analytics], outputs=[report_file] ) # Auto-update customer insights on Enter key customer_id_input.submit( fn=safe_get_insights, inputs=[analytics, customer_id_input], outputs=[customer_insights] ) return demo if __name__ == "__main__": demo = create_gradio_interface() demo.launch( server_name="0.0.0.0", server_port=7860, share=True, show_error=True )
# -*- coding: utf-8 -*-
import gradio as gr
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import train_test_split, cross_val_score
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report, confusion_matrix, accuracy_score, roc_auc_score, precision_recall_curve
import plotly.express as px
import plotly.graph_objects as go
from plotly.subplots import make_subplots
import plotly.io as pio
from datetime import datetime, timedelta
import io
import base64
import warnings
from typing import Optional, Tuple, Dict, Any
warnings.filterwarnings('ignore')
# Try importing optional dependencies
try:
import xgboost as xgb
XGBOOST_AVAILABLE = True
except ImportError:
XGBOOST_AVAILABLE = False
try:
from reportlab.lib.pagesizes import letter, A4
from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer, Table, TableStyle
from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle
from reportlab.lib.units import inch
from reportlab.lib import colors
REPORTLAB_AVAILABLE = True
except ImportError:
REPORTLAB_AVAILABLE = False
# Business configuration
BUSINESS_CONFIG = {
'churn_threshold_days': 90,
'high_risk_probability': 0.7,
'rfm_quantiles': 5,
'min_customers_for_model': 10
}
# UI color scheme
COLORS = {
'primary': '#6366f1',
'success': '#10b981',
'warning': '#f59e0b',
'danger': '#ef4444',
'purple': '#8b5cf6',
'pink': '#ec4899',
'blue': '#3b82f6',
'indigo': '#6366f1'
}
class DataProcessor:
"""Handles data loading, validation, and preprocessing"""
@staticmethod
def load_and_validate(file) -> Tuple[Optional[pd.DataFrame], str]:
"""Load and validate CSV file"""
if file is None:
return None, "Please upload a CSV file"
try:
df = pd.read_csv(file.name)
# Flexible column mapping
column_mapping = DataProcessor._map_columns(df.columns)
if not column_mapping:
return None, f"Required columns not found. Available: {list(df.columns)}"
df = df.rename(columns=column_mapping)
# Clean and validate data
initial_rows = len(df)
df = DataProcessor._clean_data(df)
final_rows = len(df)
if final_rows == 0:
return None, "No valid data after cleaning"
status = f"Data loaded successfully! {final_rows} records from {df['customer_id'].nunique()} customers"
if initial_rows != final_rows:
status += f" ({initial_rows - final_rows} invalid rows removed)"
return df, status
except Exception as e:
return None, f"Error loading data: {str(e)}"
@staticmethod
def _map_columns(columns) -> Dict[str, str]:
"""Map CSV columns to standard names"""
required = ['customer_id', 'order_date', 'amount']
mapping = {}
column_variations = {