PRT564 Data Analytics and Visualisation Project

PRT564 DATA ANALYTICS AND VISUALISATION
Project
__________________________________________________________________________________
Due on Fri Week 12
INTRODUCTION
You are provided a list of potential topics to choose from to research and write the final report on,
but each topic will only be provided one reference material to start with. However, for your final
report at least 15 references are required so you can do more of your own research along the line of
the topic. You may also choose your own topic of interest subject to the lecturer’s approval even if it
is not in the suggested list.
A relevant case study must be included in your report that matches the report theme. Case study
should contain text descriptions, plots, code snippets about:
i. framing a problem with hypotheses relevant to the reporting topic;
ii. choosing, collecting, processing, feature engineering a dataset (of size at least several
megabytes) for the problem; and
iii. visualising and analysing the processed dataset through model/algorithm building and
evaluation.
Recall that the data analytics lifecycle includes six phases of Discovery, Data Preparation, Model
Planning, Model Building, Communicate Results, and Operationalize. Whereas in our case study we
only focus on the first five phases.
PROJECT SPECIFICATION
For case study, you can use any public dataset available online. Many datasets are available from
https://archive.ics.uci.edu/ml/index.php. Alternatively, you can google for both datasets and source
codes matching the theme of your report, but make sure you provide references to them (e.g. from
a paper or a Kaggle problem). Your data + code package must also be submitted through Learnline
and accompanied with a brief user manual/running and testing instruction subsection under the
Case Study section of your report for the lecturer to run and test your uploaded package.
For completing this assignment, you are to work in a group of 2 or 3. Groups must be formed by end
of week 2. You should then proceed to choose a topic of interest.

DELIVERABLE TEMPLATE & SUBMISSION
The final report (approximately 3500 words) should contain:
• Abstract - Brief summary of the contents of the report
• Introduction - An explanation of the purpose of the study; a statement of the state-of-art
research question(s) and a brief introduction of the relevant case study to be presented
next.
• Literature review - A critical assessment of the work done so far on this topic. The
assessment should systematically outline, compare, discuss the methods and results of the
previous studies and then point out the shortcomings involved and suggest further
improvements.
• Case Study - State how the case study is related to the report topic/theme/domain.
i. Specific problem scenario and framing with initial hypotheses (e.g. churn prediction
to prevent the loss of customers). Make sure the problem is related to the report
topic.
ii. Data selection, collection and description; initial data processing (into clean training
and testing datasets or dataframes), data analysis and feature engineering
iii. Model and parameter selection via testing various model sketches and
configurations to gain insights of trade-offs, scenarios and sensitivity analyses.
iv. Model evaluation and significant results presentation (to high-level sponsors or
expert-level analysts) through visualisation and discussion. Discuss model limitation,
lessons learnt, suggestions, analytics value and quantifiable value added to the
client.
v. A brief user manual/running and testing instruction for the submitted code and
data.
• Conclusion - state the conclusions, findings and implications of the literature review and the
presented case study; also point to directions for further work in the area.
• References - at least 20 relevant articles need to be covered. For referenced conference
papers, the quality/ranking can be checked through http://portal.core.edu.au/confranks/.
For journal articles, check their impact factors and IEEE/ACM Transactions are
recommended. Other online articles and resources should also be of high quality and
authority. Use Google, Google Scholar and online Digital Libraries for your research.
Note that for case study, you are not required to strictly adhere to the textbook formats and
examples (e.g. the final presentation components). Instead they can be your reference points.
Contents, logical workflow, insightful analysis and presentation quality of the case study will be more
valued. Feel free to exercise and document your findings and thoughts as a data scientist in this
assignment!
Each group is required to submit a report as outlined above together with a zipped data+code
package. Note that at the start or the end of each paragraph/section of the report, please note down
the name(s) of the respective contributor(s) for the purpose of peer assessment.