INF6027 Introduction to Data Science Analysis of the UK Police Dataset-白红宇

INF6027 Introduction to Data Science Analysis of the UK Police Dataset

阅读量：5154 次

发布时间：2019-06-13

本文共 19507 字，大约阅读时间需要 65 分钟。

INF6027作业代写、代写R/RStudio留学生作业、代做R程序设计作业、代写R实验作业

Postgraduate coursework, Information School

INF6027 Introduction to Data Science (2018-19)

Analysis of the UK Police Dataset

1. Introduction

This part of the assessment for INF6027 Introduction to Data Science comprises a piece of individual coursework to assess your

ability to analyse data using R/RStudio and to then communicate your findings. Given a specific topic and dataset (see Section 2),

you should identify a specific problem or topic you would like to investigate (e.g., where or when particular types of crime occur, or

co-occur). You will then need to pre-process and analyse the dataset to identify patterns and relationships that address your selected

problem/topic. This should involve using techniques learned throughout the practical sessions that will help you to demonstrate your

R skills, such as summarising datasets, statistical modelling or data visualisation, to highlight and illustrate particular aspects of the

data you want to communicate (e.g., particular patterns or trends).

This coursework aims to follow the stages involved in a ‘typical’ data science process: (i) define the question(s) to address (note,

sometimes this does not come at the start of the process, but after initial exploration of the data); (ii) gather data; (iii) transform, clean

and structure the data; (iv) explore and analyse the data; and (v) communicate the findings of the data analysis. This often occurs in

an iterative manner and centred on one or multiple questions you are seeking to address. For example, the data discovery process

in Figure 1 presents an example of the stages involved in data discovery as an iterative process1

and you can find more details in

Section 3. This is also similar to the data science process we have been using in class from the “Doing Data Science” book (O’Neil &

Schutt, 2013).

Fig. 1 Example data discovery process (Jones, 2014: p.2)

You should write a 3,000 word structured report (see Section 4) that describes the approach you have taken to explore and

analyse the data for the selected problem/topic. You report should clearly communicate the results of your data analysis and be

written in a way that helps the reader interpret your findings. Note: charts, tables, and appendices are not included in the word count.

This assessment is worth 100% of the overall module mark for INF6027. A pass mark of 50 is required to pass the module as a

whole. Submission deadline: 10am Monday 21st January 2019 (Week 14) via Turnitin. See Section 5 for more general

information about Coursework Submission Requirements within the Information School.

2. The UK Police Dataset

The dataset to be used in this assessment is the UK Police Dataset, which has made public crime data since 2011 (this is an

example of Open Data). There has been a lot of recent interest in analysing publicly available datasets to identify patterns of crime

and gain insights into criminal activity, see for example the crime activity browser by IBM2

. If interested in the topic you can also find

further crime-related datasets produced by the UK Data Service (https://www.ukdataservice.ac.uk/get-data/themes/crime). There is

also an increasing use of crime Open Data used in the media to highlight aspects of policing and criminal activities (see, e.g.,

https://www.bbc.co.uk/news/uk-44044537).

You can find out more about this process in (Jones, 2014: p.2): https://tanthiamhuat.files.wordpress.com/2015/07/communicating-data-with-tableau.pdf

Open Crime Data, Free for All: https://developer.ibm.com/clouddataservices/2016/11/03/open-crime-data/

A description of the data is available here: http://data.police.uk/about/ also including an explanation on how to download the data3

The data are provided as CSV files (note that there is also an API available if you prefer) and provide street-level crime, outcome

and stop and search information broken down by police force4

(in the UK there are 45 territorial police forces and 3 Special Forces)

and 2011 Lower Layer Super Output Areas (LSOA).

The dataset describes crimes reported to UK police during each month in different areas of the UK. Information in the dataset

includes the following: geographical location (longitude and latitude), date (month, year), LSOA code (i.e., the census area), and type

of crime (e.g., vehicle crime, burglary, robbery, etc.). You can select any data from the UK Police Dataset. (This may require multiple

downloads.) You can also aggregate the dataset with other data sources if you want (e.g., census data), which would demonstrate

your ability to join datasets (although you don’t have to do this to pass the coursework as the emphasis of the coursework is on how

you carry out your analysis in R/RStudio and communicate your findings on the UK Police Dataset).

3. What you need to do

The following sections describe what you need to do in order to carry out the coursework. This roughly follows the steps shown in

Fig. 1, but you don’t have to be constrained by this or follow them in this particular order; it is just a suggestion. Also, all the R we

have done in the practical sessions (and the final sessions) should be enough to conduct the coursework, although you may need to

investigate certain areas further that relate specifically to the problem you tackle in your investigation.

3.1. Review the literature and identify research question(s)

As mentioned previously, you should select a specific problem/topic related to the data (the ‘question’ stage in Fig. 1). To decide

what area to focus on you could start by undertaking a brief review of the relevant literature around areas, such as analysis of crime

data, geographical analysis of crime, predictive policing, crime sensing, analysis of crime statistics, etc. For example, these articles

may be a useful starting point:

Vandeviver, C., and Bernasco, W. (2017) The geography of crime and crime control, Applied Geography, Volume 86, pp.

220-225. (Available online: http://www.sciencedirect.com/science/article/pii/S014362281730838X)

Field, S. (1992) The Effect of Temperature on Crime, The British Journal of Criminology, Volume 32, Issue 3, pp. 340–351.

(Available online: https://doi.org/10.1093/oxfordjournals.bjc.a048222)

Reviewing past literature will help you understand what kinds of analyses are typically undertaken using crime data and provide a

possible source of ideas for what you could do with the UK Police Dataset. Examples of possible topics include, but are not restricted

to, the following:

Evolution of crimes in an area over time;

Trends and predictions of crimes and crime rates;

Analysis of certain types of crime (e.g., vehicle crimes);

Comparisons of crime types in a region;

Clustering and classification of data, e.g., by type of crime;

Normalisation and integration with other datasets (e.g., LSOA census statistics);

Focus on a certain census dimension (e.g., age of residents in the area);

Visualisation of the data (e.g., on maps).

3.2. Download, pre-process and explore the data

As well as reviewing relevant academic literature you should also download some data from http://data.police.uk/ and perform an

exploratory analysis (i.e. ‘play’ with the data), to better understand the dataset and also help you to identify a particular problem or

topic you might want to focus on.

This part of your investigation will include steps to pre-process and transform the data, such as cleaning up the data, dealing with

missing values, standardising numeric values, etc. This may also include combining or joining the data with further datasets, e.g.

census or deprivation data. This reflects the ‘gather’ and ‘structure’ stages in Fig. 1. (Note: this part of the analysis could take a lot of

time so don’t underestimate how much time you will need to spend on this part of the coursework.)

You can also find an article describing the accuracy of the data here: https://www.tandfonline.com/doi/full/10.1080/15230406.2014.972456

https://en.wikipedia.org/wiki/List_of_police_forces_of_the_United_Kingdom

3.3. Analyse and explore the data

As you identify a topic of interest for your analysis then you should identify the most appropriate techniques (using R and associated

packages) for carrying out your analysis and exploring the data, e.g. you might want to predict crime rates using regression or

compare levels of crime types using statistical tests. This might also be an iterative process whereby you perform some analysis and

then gather (or remove) more data. Where possible relate you analysis to the relevant literature. This relates to the ‘exploring data’

stage in Fig. 3.

Note that this is often an iterative process: as you explore the data you may end up re-designing your research questions, having to

gather more data or having to perform further cleaning as more data quality issues arise. Again, this is all a part of the data discovery

process.

3.4. Write up your findings

Once you have performed analysis on the data and have some results then you need to write up your investigation into a report (this

is the ‘communicate’ stage of Fig. 1). The report should be structured as outlined in Section 4. You will be evaluated on your ability to

plan and undertake data analysis and exploration of crime based on the UK Police Dataset, your ability to engage with the relevant

literature, your use of R (and appropriate packages) and RStudio to process and analyse the data, and the way in which you

communicate your findings within the report for your given problem/topic.

You should also provide your R code as an appendix and marks will be awarded for your clarity, consistency and way in which you

comment your R code (see, e.g. http://stat405.had.co.nz/r-style.html). The specific style you use is not as important as how well you

comment your code so that someone else can follow what you have done and being consistent in whichever style you adopt.

The minimum requirement to pass is to perform at least one type of data analysis (e.g., clustering, prediction, time-series analysis,

etc.) and include at least two visualisations (e.g., charts, maps, etc.) in the report. To obtain a higher mark and more effectively

communicate your findings, you may decide to use more than one dataset or present more than one type of data analysis and/or use

multiple visualisations. Again, you should also engage as much as possible with the appropriate literature.

4. Report structure

You are required to produce a structured report that includes the sections detailed in Table 1. Overall, 90 marks will be awarded

based on the content of be awarded based on the presentation of the report and how well you communicate your findings. You must

state the word count somewhere in the report (or the coversheet). As there is a word count limit (3,000 words) you should aim to

make your writing as concise and informative as possible. Also note that your work will be assessed taking into account the word

limit; therefore, we are not expecting detailed multiple analyses in the report; rather the emphasis should be on the clarity, accuracy

and quality in communicating your findings. Note that words within tables and appendices are not included in the word count.

Table 1: Required content of the structured report.

Section Description Examples of what we will be

looking for and mark allocation Maximum allocated

marks

Structured

abstract

This should provide a summary of your report in a

structured manner, e.g. objective, methods, results,

conclusions. This is not included in the word count.

Brief but informative abstract that is

clearly structured.

Required, but 0 marks

Table of

contents

This should include section titles and page numbers.

This is not included in the word count.

Clearly structured Table of Contents

with use of numbering for sections.

Required, but 0 marks

Introduction and

aim(s)

This section should describe your selected problem or

topic addressed in the report and that forms the focus

for your data analysis. This should include a (brief)

summary of the literature around analysis of crime data

relevant to your selected topic that helps to provide the

background to your chosen topic. You should also state

why you chose this problem/ topic and why you think it

is an important topic to consider in this dataset (ideally

support by the relevant literature)

Clear statement regarding the overall

goal of your investigation.

Brief literature review of data and

crime analysis.

More marks for engagement with the

relevant literature.

10 marks

Methodology This section should describe the process you have

used to gather the data, pre-process and clean the

data, conduct your analyses and visualise the data

(note, you could follow the stages in Fig. 1). This will

Expect to see a clear description of

methodology used in your analyses.

Clear list of the datasets used (and

links to sources) and variables in the

20 marks

include ways in which you gathered, pre-processed,

transformed, and sampled/ filtered the data.

You should try to justify your choices and include

references to relevant literature where appropriate. This

should also include details of the experimental setup,

e.g. which R packages you have used etc. Think of it

like this, if someone else had to replicate your

methodology have you provided enough details (and

clearly enough) for them to reproduce your results.

As well as describing the methodology used to generate

your results, you should list all the UK Police datasets

used (e.g., data covering different regions or time

periods). You should also list any additional external

datasets used (e.g., shape files or census statistics for

LSOA areas). Describe all datasets used, any preprocessing

and how they were joined together (e.g.,

over LSOA area identifiers).

dataset(s).

Clear discussion of methods for preprocessing

data (and appropriate use

of R packages).

More marks for examples of the data.

More marks for multiple data sources

used.

More marks for the range of

techniques used, appropriateness,

links to supporting literature etc. (e.g.,

methods for trend prediction, spatial

data analysis etc.). Techniques can

include types of visualisation and

references to which R libraries have

been used

More marks for the detail of the

description provided, e.g., could

include use of group_by(),

aggregate() etc.

More marks for use of methods to

deal with data quality issues, such as

missing values.

More marks for discussing use of

appropriate techniques for different

types of data, e.g. categorical data.

Results and

discussion

In this section you should present the results of your

data analysis and exploration (e.g., statistics, maps,

trends, predictions). You should use the results to

address the selected problem by presenting and

discussing tables and charts as appropriate.

You should present your findings in a way that helps the

reader interpret the results. You should focus on

effectively communicating the results of the analysis to

the reader by highlighting the trends or patterns you

have observed during your data analysis.

More marks for correct use of

statistics and visualisations.

More marks for packaging results etc.

into tables rather than simply using R

output or command line code.

More marks for a clear narrative and

structure (e.g., adding sections and

sub-sections and guiding the reader

through the analysis).

More marks for clearly explaining the

results and graphics used (e.g., use

of legends etc.).

More marks for using graphics that

convey information (e.g., combine

results) and help identify insights

(e.g., use of log scales to dampen

effects of high values etc.).

More marks for bringing out insights

rather than leaving the reader to

interpret the findings.

More marks for not over-interpreting

the results and recognising biases.

More marks for re-labelling the

variable names in graphs and tables

(rather than using default names).

More marks for how well the data is

summarised and made accessible for

comparison.

50 marks

Conclusion In this section you should summarise the main findings

of your analysis and lessons learned. You should state

the main message the reader should come away with

from your analysis.

You should also highlight any weaknesses of your

analysis and state what you would do to improve your

analysis if you had more time.

Summary of the main findings of the

analysis with respect to the original

aim(s) of the investigation.

More marks for highlighting

limitations/ weaknesses of your

methodology and analysis.

More marks for a clear set of takeaway

messages.

10 marks

R code You should include the full R code as an appendix. ? More marks for well-commented

code.