Machine learning (ML) methods have been used to predict and detect criminal activities with promising results. Vancouver criminal data from the past 15 years was analyzed using boosted decision trees and K-nearest neighbor models, predicting crimes with 39-44% accuracy. Chicago crime data was also examined using various ML models, with KNN classification achieving the highest accuracy of 78.7%. A deep neural network approach combining data from multiple domains achieved 84.25% accuracy in predicting crime occurrences in Chicago, outperforming other models like SVM and kernel density estimation. The studies demonstrate that ML has potential for helping law enforcement agencies predict, detect, and reduce criminal activities.
Machine learning (ML) methods have been used to predict and detect criminal activities with promising results. Vancouver criminal data from the past 15 years was analyzed using boosted decision trees and K-nearest neighbor models, predicting crimes with 39-44% accuracy. Chicago crime data was also examined using various ML models, with KNN classification achieving the highest accuracy of 78.7%. A deep neural network approach combining data from multiple domains achieved 84.25% accuracy in predicting crime occurrences in Chicago, outperforming other models like SVM and kernel density estimation. The studies demonstrate that ML has potential for helping law enforcement agencies predict, detect, and reduce criminal activities.
Machine learning (ML) methods have been used to predict and detect criminal activities with promising results. Vancouver criminal data from the past 15 years was analyzed using boosted decision trees and K-nearest neighbor models, predicting crimes with 39-44% accuracy. Chicago crime data was also examined using various ML models, with KNN classification achieving the highest accuracy of 78.7%. A deep neural network approach combining data from multiple domains achieved 84.25% accuracy in predicting crime occurrences in Chicago, outperforming other models like SVM and kernel density estimation. The studies demonstrate that ML has potential for helping law enforcement agencies predict, detect, and reduce criminal activities.
Since, the goal of a crime detection method is to predict and prevent
the criminal activities; therefore, the significance is very high. Although, the conventional (non-machine learning) methods are useful but they all operate independently. Therefore, a machine which is capable to integrate the encouraging aspects of conventional methods would be extremely advantageous.
ML-based criminal activities were predicted, detected, and examined
by using criminal data of Vancouver for the last 15 years. The ML- based criminal activities analyses were involved in collecting the data, classification, identification of the criminal patterns, prediction, and visualization. The boosted decision tree and K-nearest neighbor (KNN) models were used to examine the criminal dataset. Between 2003 and 2018, a total of 560,000 criminal datasets were investigated, and predicted the criminal activities with accuracy between 39% and 44%.
ML and data science-based models were implemented to predict and
detect the criminal activities from the Chicago criminal dataset [4]. Various combinations of the ML-models, including logistic regression, SVM/KNN classification, decision trees, random forest, and Bayesian models were examined; thereby selected the most accurate model for training. The KNN classification achieved the best accuracy of 78.7%. The main goal of the study was to give a thought to the law enforcement agencies to implement the ML-based approaches for predicting, detecting, and solving the criminal activities at higher rates to reduce the crimes in society. A feature-level data fusion method using a DNN was proposed in [5] to predict crime occurrence with high accuracy by combining multi- model data from different domains with environmental context knowledge. The database to predict the crime was composed of the data taken from an online crime statistics database (Chicago), meteorological and demographic data/images. Different ML models were used for crime prediction, such as SVM, regression analysis, kernel density estimation (KDE). The SVM and KDE models achieved 67.01% and 66.33% accuracies, respectively whereas the proposed ML model achieved 84.25% accuracy.