Effects Of Big Data Policing On Crime Prevention: A Study Of Burglary In England Essay.

Big Data Policing and Crime Prevention

Big Data and Crime Prevention

Does big data policing work? And if so, what are the metrics that are used to measure the extent of the effectiveness of the policing products associated with big data policing? And even better which algorithm is perfect for predicting human behavior? While all these questions are controversial ethically, the role of big data in developing decision support tools follows a consensus (Poleto, et al., 2015; Grander, et al., 2021). As observed by (Grander, et al., 2021), techniques related to big data, analytics, machine learning algorithms, as well as technologies in computer science and cloud computing are commonly being used in decision support tools.

Following the growing popularity of big data concepts, it has been adopted in several areas including applications in crime prevention (Vassilakos, et al., 2019). (Vassilakos, et al., 2019) argue that some areas tend to be prone to crime relative to other areas e.g., crime rates in some areas of metropolitan cities tend to be high. Studies such as (Vassilakos, et al., 2019; Danziger, 2020)  propose that big data can be used to underscore best practices and augment solutions to enable the prevention of crime before it happens.

Ethical Concerns

However, just like in the judicial system, (Danziger, 2020) notes that the adoption of big data by police units and law groups to help them address cases comes with underlying legal ramifications or implications that are not always clear. This implies that the role of big data in developing decision tools to combat crime has good and bad effects. Through broad and extensive data analyses, big data can be a useful tool for law enforcement, on the other hand, the use of big data analysis raises several ethical questions (Magalhaes & Fontana, 2020; Weinhardt, 2021). But that is not to say big data cannot be used to learn different aspects of crime such as distribution, trends, among other crime demographics albeit with some marginal errors due to assumptions that occur as a result of generalizations. Studies such as (Završnik, 2019; Heaven, 2020) argue that tools that draw data on individuals including “…age, gender, marital status, history of substance abuse, and criminal record” to learn and predict the probability of committing a crime using algorithms that are fed by such data are easily skewed by arrest rates, gender, and race.

Following the hypothesis that individuals who live in areas that are frequented by burglary cases are more likely to seek insurance covers, it is important to learn the characteristics of such areas that make them prone to burglaries. To this end, the current study seeks to understand the demographics of burglary crimes in England including whether or not the associated areas are areas of affluence, where a premium policy with high benefits could be sold, or one of relative deprivation where a Low-Cost Economic Policy with proportionately lower pay-outs would be more appropriate. Besides, the study seeks to understand the trend in the number of reports related to burglary crimes. To address the research objective, we will answer questions related to whether:

1. There are more burglaries in more affluent areas
2. Burglaries are increasing, decreasing, or are stable 

Measuring the Effectiveness of Big Data Policing

Approaches and Methodologies

Several approaches have been proposed in the analysis of crime data to extract insights regarding a specified aspect of crime. In the current work, three analysis approaches were adopted. Using statistical aggregation i.e., through generating counts of burglary cases, exploratory visual analysis regarding the distribution of the different demographics of crime including the count of crimes in areas reported to either be affluent or non-affluent as well as areas with low/no deprivation and those that are highly deprived. Besides, machine learning methodologies such as linear regression and association rule mining were adopted.

Association Rule Mining

We adopted an Apriori Algorithm to generate frequent itemsets between crime type, affluence, and deprivation attributes. Our objective was to determine the top 5 items that occur together to determine whether burglar crimes are closely related with affluent/non-affluent neighborhoods or low/no/high derivation.

Linear Regression

Since the company objective was to target areas where burglary cases are increasing or put another way, areas where the rate of burglary increase is high or decrease is low, a linear regression and trend analysis approach was adopted. Using linear regression, the company can be able to determine the rate of increase or decrease depending on the underlying objective.

Descriptive Analytics and Visualization

As observed above, descriptive and visual exploratory analyses were used to determine the distribution of the demographics related to burglary crimes. That is, what is the distribution of the reports of crime related to burglaries in terms of affluence and deprivation.

Are there are More Burglaries in More Affluent Areas?

There exist several studies that argue that burglars tend to target affluent areas more which are however linked to increased demand for drugs in affluent areas (Melling, 2021). On the other hand, some researchers observe that compared to affluent areas, disadvantaged neighborhoods are more attractive for burgling (Chamberlain & Boggess, 2016). (Cuthbertson, 2018) observes that individuals who earn below £10,000 are twice as likely to be burgled compared to individuals with an income of above £50,000. That is, more deprived individuals are likely to be burgled compared to the less deprived. 1

Figure 1 below shows the percent distribution of crimes that happened in affluent and non-affluent postcodes between 2010 and 2021.

Figure 1: Proportion of burglary by affluence

Overall, there are more burglar cases in non-affluent areas (see figure 1) which supports the arguments by (Chamberlain & Boggess, 2016) that disadvantaged neighborhoods are more attractive for crimes. However, examining the rate of increase/decrease in the reports of burglary, we note as shown in figures 2 and 3 respectively that, the rate of decrease in affluent areas is lower relative to the rate of decrease in non-affluent areas at which burglary cases are being reported.

Figure 3: Burglary in non-affluent areas

Interestingly, as shown in figure 4 below, most of the crimes were related to areas where individuals face no deprivation which contradicts the argument by (Cuthbertson, 2018) that individuals who earn less are more likely to be burgled.

Figure 4

Using confidence as the evaluation measure, table 1 below highlights the associations between burglary, affluence, and deprivation from where we note that burglary is more likely to occur in affluent areas.

Algorithm for Predicting Human Behavior

Table 1: Top 5 associations

We also note that high deprivation is associated with non-affluence with high deprivation having the second-highest association with burglary.

The Trend of Burglary by Deprivation or Lack of it

From figures 5 and 6 below we note that the rate of decline in the number of burglar cases is high compared to that of areas with high deprivation.

Figure 6

Are Burglaries Are Increasing, Decreasing, Or Are Stable

Figures 5 and 6 below show the trend of the number of burglaries reported over time as well as the rate of the trend respectively.

Trend

Figure 7: Overall distribution of the rate of burglary crimes

From figure 5 above we establish that there is a decreasing tendency of the number of burglar cases reported in England over time. Particularly, when examining the overall rates at which burglar cases are declining in the UK relative to affluent and deprivation, the overall rate is a bit lower compared to that of affluence which is relatively lower compared to those of deprivation.

Evaluation and Conclusion

While other factors could have contributed to burglar cases including instances of unemployment, ease of access to drugs and hence the use of it, etcetera, the current study was mainly inclined towards understanding the burglar phenomenon in relation to affluence and deprivation. In the current work, we have established that affluence and burglary, as well as high deprivation and burglary, are the most likely items to occur together given that they have the highest confidence (a 1.0 confidence on the antecedents and consequents) measure which indicates that affluence high levels of deprivation are more attractive to burglaries. We have also established that areas that have high levels of deprivation and affluent areas have a lower decline in crime rates i.e., -19.6 and -13.4 respectively.

Essentially, the findings of the current study are in line with the original hypothesis that there are more burglaries in affluent areas. Our findings also support the argument that areas with deprivations are more prone to burglaries and that the level of burglaries is decreasing. As such, the company needs to target affluent areas with a premium policy with high benefits and areas that are highly deprived with a low-cost economic policy with proportionately lower pay-outs. This follows the assumption that individuals in areas that are prone to burglaries are more likely to be interested in the new insurance policy for their property.

The current study uses a sample of the data which might fail to present the complete picture of the distribution of burglary crimes in England. As such, the findings in this study should be considered indicative but not absolute. For future study, we propose using a more comprehensive sample to understand the behavior of burglary crime in the UK. 

References

Chamberlain, A. W. & Boggess, L. N., 2016. Why disadvantaged neighborhoods are more attractive targets for burgling than wealthy ones.. [Online]
Available at: https://blogs.lse.ac.uk/usappblog/2016/09/26/why-disadvantaged-neighborhoods-are-more-attractive-targets-for-burgling-than-wealthy-ones/
[Accessed 22 December 2021].

Cuthbertson, P., 2018. Poverty and Crime: Why a new war on criminals would help the poor most, s.l.: CIVITAS.

Danziger, C., 2020. The Good the Bad of Big Data in the Criminal Justice System. s.l., inside big data.

Grander, G., Silva, L. F. d. & Gonzalez, E. D. R. S., 2021. Big data as a value generator in decision support systems: a literature review. Revista de Gestão, 28(3).

Heaven, W. D., 2020. Predictive policing algorithms are racist. They need to be dismantled. MIT Technology Review, 17 July.

Magalhaes, M. & Fontana, A., 2020. A Perspective on Ethics & Law in AI and Big Data Analytics. The Harvard Law Record.

Melling, J., 2021. Why Is Crime Rising in Affluent Areas?. [Online]
Available at: https://www.churchillgroup.com/resources/news-insights/why-is-crime-rising-in-affluent-areas/
[Accessed 22 December 2021].

Poleto, T., Carvalho, V. D. H. d. & Costa, A. P. C. S., 2015. The Roles of Big Data in the Decision-Support Process: An Empirical Investigation. In: Decision Support Systems V – Big Data Analytics for Decision Making. s.l.: Springer International, pp. 10-21.

Vassilakos, A., Gottlieb, M. & Dawson, M., 2019. Towards Crime Prevention Using Big Data Analytics: A Literature Review with an Explorative Case Study. s.l., Research Gate.

Weinhardt, M., 2021. Big Data: Some Ethical Concerns for the Social Sciences. Social Sciences, 10(2), p. 36.

Završnik, A., 2019. Algorithmic justice: Algorithms and big data in criminal justice settings. European Journal of Criminology, 18(5), pp. 623-642.