US Fatal Police Shooting Analysis and Prediction: WP fatal police shooting dataset insight

Table of Links

Abstract and Introduction

Related work

Methodology

Media reporting analysis

WP fatal police shooting dataset insight

Fatal police shooting rate and victims race prediction

Conclusion and References

5. WP fatal police shooting dataset insight

In this part, we use FP-growth and word cloud to reveal the frequent pattern behind the WP dataset. We use location data from the WP dataset to cluster police shooting incidents and find shooting hotspots. We also tried multiattributes such as social economics, demographics, political tendency, education, gun ownership rate, police training hours, etc., to verify the possible reason for the police shooting.

5.1. Frequent Pattern Mining

From the frequent pattern mining, we can conclude a typical victim shot by police: a “man” (96%) “without mental illness” (77%) uses “gun” (57%) “attack” (65%) police then get “shot” (95%) by police who does not wear “body camera” (88%). see below Figure-8 and Figure-9. “California,” “Texas,” “Florida” are the top three states were happened more frequently in total number, see Figure-10.

Therefore, our subsequent analysis considers gun ownership rate, crime rate, Marijuana legality, and governor’s party by state level. The frequent pattern uses FP-growth [HPY00], and the threshold of minimum support is 50% of the total transactions of the WP dataset.

We also apply DBSCAN [5] to the longitude and latitude of fatal police shooting locations to identify hotspot clusters. Set parameters eps=0.5 and min sample=50, we find the dense areas of fatal police shootings, see below Figure11. We discover that Los Angles and Atlanta metropolitan areas have two of the largest hotspots. Generally, all the fatal police shooting hotspots are in the top population cities in the country.

5.2. Correlated variables analysis

5.2.1 Quantitative Variable analysis

To avoid the population distorting the analysis, we normalized the number to the yearly average fatal police shooting per one million people (fatal police shooting rate). We use this density-kind value for the analysis afterwards. Figure12 shows that every year on average, how many people were shot by police. New Mexico and Alaska where have relatively less population, become the top state. The color is getting darker from east to west except for large population states such as California, Washington. Doesn’t it look like the U.S. history of territory expansion?

It looks the longer the state joined the U.S., the lower the fatal police shooting rate in that state. The correlation coefficient is 68% between the fatal police shooting rate and the U.S. history of territory expansion. Our interpretation is: the reason that U.S. police use excessive violence may root from the westward expansion when handling the violent criminals, see Figure-13.

The correlation coefficient is 64% between gun ownership rate [17] and the fatal police shooting rate. 57% of victims hold guns (not including other weapons), and 65% of victims chose to attack police. This hold gun rate doubles than the average gun ownership rate among the country, which is 30% according to Pew’s report [7], see Figure-14.

The third high correlation variable is the land area [16], 59%, followed by violence rate [3], 48%, poverty rate 37%, unemployment rate 29%, see Figure-16. Surprisingly, police basic training hours negatively correlate with the fatal police shooting rate, see Figure-15. Although TrainingReform [10] appeals appeal to increase police training hours, the current data shows the opposite result. It may suggest reviewing and improving the training itself rather than a single slogan for more hours.

We also tested the correlation coefficient’s significance to guarantee the association, which are all proved with relatively small p value, see Table-1.

5.2.2 Categorical variables analysis

In this part, we tested the significance of the fatal police shooting rate by state-level Governor’s party [15] and Marijuana Legality [2]. We failed to reject the null hypothesis, and we can conclude that there is no difference between those states on the fatal police shooting.

T-test results:

H0: µGOP = µDems

HA: the average fatal police shooting rate are not equal between Republican and Democrat governor states

Test result: since value = 0.3254 ¿ 0.05, we failed to reject the null hypothesis. The average fatal police shooting rate are equal between Republican and Democrat governor states

One-way ANOVA:

H0: µF L = µMML = µF I

HA: at least one of the average rates differs from one of the others

Test result: F = 0.6492, Pvalue = 0.527, fail to reject the null hypothesis. The average fatal police shooting rate are equal among different marijuana legality states