Abstract

Criminal activity is on the rise everywhere and it refers to an illegal act that is subject to legal penalties. Crimes can range in severity from less serious offenses like traffic infractions and smalltime larceny to more serious ones like murder, robbery, and fraud. While those who have committed such crimes and been adjudged guilty by a court of law are known as criminals. They may lack education, be underprivileged, employed, or unemployed, or they may be wealthy. Add to this, crime may be committed differently depending on a number of variables, including family history, religion, gender, and so on. However, most people who participate in unlawful or violent activity often belong to specific social groupings, have less education, and work in industries where minorities predominate. As opposed to that, Criminal people who focused on improving their financial and living situations frequently engage in criminal activity including stealing, robbery, and theft. Although criminals are typically ostracized and held in low regard in most societies. However, it's crucial to keep in mind that criminals are not all the same, and some may be affected by external factors or societal problems like discrimination or poverty. It's also crucial to understand that not everyone who commits a crime is apprehended or punished and that occasionally the criminal justice system exhibits bias or flaws. Therefore, the main goal of this thesis is to develop strategies with the primary goal of reducing or predicting the occurrence of criminal acts. To apply this, three machine learning technologies which are Decision Tree, Logistic regression, and K-Nearest Neighbor are used to predict future criminal acts based on historical criminal data gathered from the Cincinnati, USA website between 2009 and 2022 which has 485,469 observations with 40 attributes. Cincinnati data is selected to be observed and tested by ML algorithms as UAE data are very confidential and Cincinnati has a high crime rate of 6.4/10 based on US news and world reports (Cincinnati, OH, Crime Rate & Safety | U.S. News Best Places, 2023). Further based on crime pattern theory and environmental criminology that are covered in the literature review section of this thesis, crime is not random; it is either premeditated or motivated by opportunity. This was confirmed by the thesis’s historical crime data analysis which showed that attributes such as VICTIM_AGE, LOCATION, and WEAPONS are important and could affect on crime prediction. As a result, several actions can be taken, such as law enforcement departments can mitigate crime by assigning the right type of forces in the predicted area with proper resources for quick response. Additionally, the thesis results showed that the Decision tree algorithm is one of the effective machine learning algorithms for predicting crime due to its simplicity to understand, versatility to be used for regression and classification problems, scalability, and robustness in handling large datasets with high dimensionality while the collected results were with high accuracy and moderate execution time.

Sally Samir Matter Isbaih, Zayed University

Name of Guide/supervisor

Feras Al-Obeidat

Year of awarding

2023

Download

Share

COinS