Real-time Customer Analytics Dashboard

import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
import plotly.graph_objects as go
from datetime import datetime, timedelta

class CustomerAnalytics:
    def __init__(self):
        self.clv_model = RandomForestRegressor(n_estimators=100, random_state=42)
        self.churn_model = None
        self.feature_columns = [
            'total_purchases', 'avg_order_value', 'days_since_last_purchase',
            'purchase_frequency', 'total_sessions', 'support_tickets',
            'referral_count', 'subscription_tier'
        ]

    def calculate_clv_features(self, customer_data: pd.DataFrame) -> pd.DataFrame:
        """Calculate customer lifetime value features"""
        features = customer_data.copy()

        # Recency features
        features['days_since_last_purchase'] = (
            datetime.now() - features['last_purchase_date']
        ).dt.days

        # Frequency features
        features['purchase_frequency'] = (
            features['total_purchases'] / features['customer_age_days']
        )

        # Monetary features
        features['avg_order_value'] = (
            features['total_revenue'] / features['total_purchases']
        )

        # Engagement features
        features['sessions_per_day'] = (
            features['total_sessions'] / features['customer_age_days']
        )

        return features[self.feature_columns]

    def predict_customer_lifetime_value(self, customer_data: pd.DataFrame) -> pd.DataFrame:
        """Predict customer lifetime value for next 12 months"""
        features = self.calculate_clv_features(customer_data)

        # Predict CLV
        predicted_clv = self.clv_model.predict(features)

        # Calculate confidence intervals
        predictions = []
        for estimator in self.clv_model.estimators_:
            predictions.append(estimator.predict(features))

        predictions = np.array(predictions)
        clv_lower = np.percentile(predictions, 10, axis=0)
        clv_upper = np.percentile(predictions, 90, axis=0)

        results = customer_data.copy()
        results['predicted_clv'] = predicted_clv
        results['clv_lower_bound'] = clv_lower
        results['clv_upper_bound'] = clv_upper
        results['prediction_date'] = datetime.now()

        return results

    def generate_customer_segments(self, customers: pd.DataFrame) -> pd.DataFrame:
        """Segment customers based on RFM analysis"""
        # Calculate RFM scores
        customers['recency_score'] = pd.cut(customers['days_since_last_purchase'],
                                          bins=5, labels=[5,4,3,2,1])
        customers['frequency_score'] = pd.cut(customers['purchase_frequency'],
                                            bins=5, labels=[1,2,3,4,5])
        customers['monetary_score'] = pd.cut(customers['avg_order_value'],
                                           bins=5, labels=[1,2,3,4,5])

        # Create segments
        def assign_segment(row):
            if row['recency_score'] >= 4 and row['frequency_score'] >= 4:
                return 'Champions'
            elif row['recency_score'] >= 3 and row['frequency_score'] >= 3:
                return 'Loyal Customers'
            elif row['recency_score'] >= 3 and row['frequency_score'] <= 2:
                return 'Potential Loyalists'
            elif row['recency_score'] <= 2 and row['frequency_score'] >= 3:
                return 'At Risk'
            else:
                return 'New Customers'

        customers['segment'] = customers.apply(assign_segment, axis=1)
        return customers