🔮 Predictive Analytics

import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.preprocessing import StandardScaler
import joblib

class PredictiveMaintenanceModel:
    def __init__(self):
        self.model = RandomForestRegressor(n_estimators=100)
        self.scaler = StandardScaler()
        
    def prepare_features(self, sensor_data):
        """Extract predictive features from sensor data"""
        features = {}
        
        # Rolling statistics
        features['temp_mean_7d'] = sensor_data['temperature'].rolling(7).mean()
        features['pressure_std_7d'] = sensor_data['pressure'].rolling(7).std()
        features['vibration_max_7d'] = sensor_data['vibration'].rolling(7).max()
        
        # Trend analysis
        features['temp_trend'] = sensor_data['temperature'].diff(7)
        features['pressure_trend'] = sensor_data['pressure'].diff(7)
        
        # Frequency domain features
        from scipy.fft import fft
        features['vibration_fft_peak'] = np.abs(fft(sensor_data['vibration'])).max()
        
        return pd.DataFrame(features)
    
    def predict_failure_probability(self, current_data):
        """Predict failure probability for next 30 days"""
        features = self.prepare_features(current_data)
        features_scaled = self.scaler.transform(features)
        
        failure_prob = self.model.predict_proba(features_scaled)[:, 1]
        
        # Generate 30-day forecast
        forecast = []
        for day in range(1, 31):
            prob = failure_prob[-1] * (1 + 0.02 * day)  # Increasing risk
            forecast.append(min(prob, 1.0))
            
        return forecast
    
    def get_maintenance_recommendation(self, failure_forecast):
        """Generate maintenance recommendations"""
        critical_threshold = 0.8
        warning_threshold = 0.6
        
        critical_days = [i for i, prob in enumerate(failure_forecast) 
                        if prob > critical_threshold]
        
        if critical_days:
            return {
                'action': 'URGENT_MAINTENANCE',
                'days_until_critical': min(critical_days) + 1,
                'recommendation': 'Schedule maintenance within 7 days'
            }
        elif max(failure_forecast) > warning_threshold:
            return {
                'action': 'PLANNED_MAINTENANCE',
                'days_until_critical': 30,
                'recommendation': 'Schedule maintenance within 2 weeks'
            }
        else:
            return {
                'action': 'MONITOR',
                'days_until_critical': None,
                'recommendation': 'Continue monitoring'
            }

# Usage example
model = PredictiveMaintenanceModel()
sensor_data = load_sensor_data()  # Load recent sensor readings
forecast = model.predict_failure_probability(sensor_data)
recommendation = model.get_maintenance_recommendation(forecast)

print(f"Action: {recommendation['action']}")
print(f"Recommendation: {recommendation['recommendation']}")