Location

Comstock Memorial Union, MSUM

Document Type

Poster

Event Website

https://www.mnstate.edu/sac/

Start Date

15-4-2025 12:00 AM

End Date

15-4-2025 12:00 AM

Publication Date

4-15-2025

Description

This senior project investigates methods for forecasting crime statistics using historical data from the State of North Dakota compiled in a dataset. Multiple forecasting techniques, including classical time series models (ARIMA and Prophet), ensemble machine learning (Random Forest), and deep learning via Long Short-Term Memory (LSTM) networks, were implemented and compared to determine their effectiveness in predicting crime rates. The study emphasizes a rigorous experimental approach where models are evaluated based on their predictive accuracy using established error metrics. While each method has its strengths, the deep learning approach demonstrated superior performance, suggesting that LSTM networks are particularly well-suited to capturing the complex temporal patterns inherent in crime data. These findings provide a foundation for developing robust predictive tools that can support proactive law enforcement and enhance public safety.

Share

COinS
 
Apr 15th, 12:00 AM Apr 15th, 12:00 AM

Forecasting Crime Trends in North Dakota: A Comparative Machine Learning Approach

Comstock Memorial Union, MSUM

This senior project investigates methods for forecasting crime statistics using historical data from the State of North Dakota compiled in a dataset. Multiple forecasting techniques, including classical time series models (ARIMA and Prophet), ensemble machine learning (Random Forest), and deep learning via Long Short-Term Memory (LSTM) networks, were implemented and compared to determine their effectiveness in predicting crime rates. The study emphasizes a rigorous experimental approach where models are evaluated based on their predictive accuracy using established error metrics. While each method has its strengths, the deep learning approach demonstrated superior performance, suggesting that LSTM networks are particularly well-suited to capturing the complex temporal patterns inherent in crime data. These findings provide a foundation for developing robust predictive tools that can support proactive law enforcement and enhance public safety.

https://red.mnstate.edu/sac/2025/cbac/5