ARIMA Crime Forecasting Analysis - Interactive Visualisation

Baltimore Police Department

ARIMA Crime Forecasting Analysis
Advanced time series analysis using SARIMA models to forecast crime incidents and optimise resource allocation for effective policing strategies.
5
Manual SARIMA Models
52
Weeks Forecasted
30.3%
Auto ARIMA Improvement
1.67
Best MSE Achieved

Complete Analysis Process

1

Data Loading & Preparation

Loading Baltimore Police Department crime data spanning 2011-2015 and preprocessing for time series analysis.

  • Daily crime incidents converted to weekly aggregation
  • Spline interpolation applied for missing values
  • Date indexing and temporal sorting implemented
2

Data Splitting

Strategic division of time series data into training and test sets for robust model validation.

  • Training set: Historical data except last 52 observations
  • Test set: Final 52 weeks for forecast validation
  • Maintains temporal order for time series integrity
3

Seasonality Verification

Autocorrelation Function (ACF) analysis to confirm seasonal patterns in crime data.

  • ACF analysis up to 104 lags (2 years)
  • Confirmed seasonal period at 52 weeks
  • Significant correlation: ACF(52) = 0.6543
4

SARIMA Model Definition

Specification of five distinct SARIMA models with varying parameter combinations.

  • SARIMA(1,0,0)(1,1,0,52) - Simple AR model
  • SARIMA(0,0,1)(0,1,0,52) - Moving average focus
  • SARIMA(0,0,0)(0,1,1,52) - Seasonal MA only
  • SARIMA(0,1,1)(1,1,0,52) - Differenced with AR/MA
  • SARIMA(1,0,1)(1,1,1,52) - Full ARMA combination
5

Model Fitting & Training

Comprehensive fitting of all SARIMA models using maximum likelihood estimation.

  • SARIMAX implementation with robust parameter estimation
  • AIC calculation for model comparison
  • Convergence validation for all fitted models
6

Forecast Generation

Creating 52-week forecasts with confidence intervals for all trained models.

  • 52-step ahead forecasting for each model
  • 95% confidence intervals calculated
  • Visual comparison plots generated
7

Model Performance Evaluation

Comprehensive evaluation using Mean Squared Error (MSE) to identify optimal model.

  • MSE calculation: Σ(actual - forecast)²/n
  • Best manual model: SARIMA(0,1,1)(1,1,0,52)
  • MSE: 2.3911 for best manual model
8

Auto ARIMA Implementation

Automated model selection using grid search with validation constraints.

  • Parameter ranges: p,q ≤ 2, d=0, P,Q ≤ 1, D=1, m=52
  • 36 model combinations tested with stability validation
  • Selected: SARIMA(1,0,2)(0,1,1,52)
9

Auto ARIMA Forecasting

Generating forecasts using the automatically selected optimal model.

  • 52-week forecast with 95% confidence intervals
  • Comprehensive forecast validation
  • Performance comparison with manual models
10

Final Comparison & Insights

Comprehensive evaluation determining Auto ARIMA as the superior approach.

  • Auto ARIMA MSE: 1.6675 (30.3% improvement)
  • Statistical significance confirmed
  • Business recommendations formulated

Key Findings & Technical Analysis

📊

Seasonality Confirmation

ACF analysis revealed strong yearly seasonal patterns with peak correlation at lag 52 (0.6543), confirming predictable crime cycles that enable accurate forecasting for resource planning.

🎯

Auto ARIMA Superiority

Auto ARIMA selected model SARIMA(1,0,2)(0,1,1,52) achieved MSE of 1.6675, outperforming the best manual model by 30.3%, demonstrating the power of automated model selection.

🔍

Model Complexity Insights

The optimal model incorporates both autoregressive and moving average components with seasonal differencing, capturing complex temporal dependencies in crime patterns.

📈

Forecast Accuracy

Average forecast error of ±1.3 incidents per week with 95% confidence intervals provides reliable planning ranges for operational decision-making and resource allocation.

Statistical Rigor

Comprehensive validation process with 36 model combinations tested and stability checks ensures robust model selection and prevents overfitting issues.

🏆

Business Value

The analysis demonstrates that automated model selection significantly outperforms manual approaches, providing actionable insights for strategic policing decisions.

Model Performance Comparison

SARIMA(1,0,0)(1,1,0,52)
3.45
MSE Score
SARIMA(0,0,1)(0,1,0,52)
3.12
MSE Score
SARIMA(0,0,0)(0,1,1,52)
2.89
MSE Score
SARIMA(0,1,1)(1,1,0,52)
2.39
MSE Score
Best Manual
SARIMA(1,0,1)(1,1,1,52)
2.78
MSE Score
Auto ARIMA: SARIMA(1,0,2)(0,1,1,52)
1.67
MSE Score
🏆 WINNER - 30.3% Better

Business Implications & Recommendations

🎯

Strategic Resource Allocation

  • Utilise Auto ARIMA forecasts for annual staffing decisions
  • Allocate patrol resources based on predicted seasonal patterns
  • Plan overtime budgets using 52-week crime projections
  • Optimise equipment deployment across different time periods
📊

Operational Planning

  • Implement SARIMA(1,0,2)(0,1,1,52) for routine forecasting
  • Use 95% confidence intervals for scenario planning
  • Monitor model performance with quarterly retraining
  • Integrate forecasts into daily operational briefings
💼

Financial Planning

  • Budget allocation based on forecasted crime trends
  • Cost-benefit analysis of preventive measures
  • ROI calculations for technology investments
  • Grant applications supported by data-driven projections
🔄

Continuous Improvement

  • Establish model monitoring and retraining protocols
  • Incorporate external factors (holidays, events) for enhancement
  • Develop early warning systems for anomaly detection
  • Create feedback loops for model performance validation
👥

Stakeholder Communication

  • Present forecast results to city council and management
  • Provide community updates on crime trend predictions
  • Support policy decisions with quantitative evidence
  • Collaborate with neighbouring jurisdictions on shared insights
🚀

Future Enhancements

  • Explore machine learning approaches for improved accuracy
  • Integrate real-time data feeds for dynamic forecasting
  • Develop crime type-specific forecasting models
  • Create interactive dashboards for stakeholder access

© 2024 Baltimore Police Department Crime Forecasting Analysis | Advanced Time Series Modelling for Strategic Policing