Correlations Between Population Demographics and Crime

Table of Contents Executive Summary: 3 Project Goal: 3 Data Description: 3 Techniques Used: 4 Clustering: 4 Segment Profile: 5 Regression: 7 Decision Tree: 8 Model Comparison: 9 Knowledge, interpretation and conclusion 9

Executive Summary:
This summary presents findings of the data mining techniques used in the Crime Statistics by City dataset. The aim of the project is to enhance the knowledge and hands on experience on EM tool and to make appropriate projections about crime rates in cities.
Project Goal:
Our goal is to determine the similarities between cities and their demographics.
Data Description:
Our dataset contains demographic data based on race, income, and other factors by city and includes different types of crime. The types include gun owners, convicted felons, race, and income. Examples of cities include Detroit, Chicago, New York, Houston, and Spring Field.

Techniques Used:

Clustering:

As seen above, a large number of cities have similar demographics and were clustered together. Cluster 19 and cluster 7 hold the majority of cities.

Population is the most important variable in determining cluster association.
Segment Profile:

The segment profile suggests that segment 19 holds a higher mean income than that of the overall population. Segment 7 is the opposite, it holds a lower mean income than the overall population. This suggests that ….

Segment 19 and Segment 7 are focused mostly on mean income, whereas other segments are focused on other variables. These include Population and Race. Examples are shown below:

Regression:

The regression technique allows us to see that the demographic data accurately predicts the crime rates. The closer the mean predicted and mean target lines are, the better the data is at predicting the target variables.

This information is further evidence that the variables accurately predict the crime rates.
Decision Tree:

The first branch of the decision tree split based on population greater than 658857 had a higher average crime than lower population rates. The second split is based on the number of whites in the population and it shows a higher amount of crime in cities with less whites. Furthermore, branches of lower populated cities showed an increase in crime with high populations of gun owners.
Model Comparison:

Both techniques of predictive modeling show an effective correlation between the demographics and the crime rates.
Knowledge, interpretation and conclusion
What we learned is that areas of denser populations tend to have higher rates of crime consistently. In areas of lower population primary factors of crime rates depend on mean income, racial diversity, and gun ownership.