Renmin University of China Ag

This individual coursework assesses your ability to use a Visual Analytics approach to investigate some research questions using a suitable dataset of your choice and your ability to critically reflect on this. Data

You  can choose whichever datasets you like for your project as long they  are complex enough (and that you have any required permissions to use  them). As a guide, if you consider your data as one big table, we’re  looking for something like 150 row/cols x 10 cols/rows as an absolute  minimum. This is simply to help increase the chances that there is  something worth investigating in there. Combining more than one dataset  is not necessary, but may improve the scope of what to investigate.

Tools

You can use whichever software tools you like for  your project, as long as you use them using a visual analytics  approach. This means that your interpretation of visual outputs must  inform the computational analysis and computational outputs must inform  the way you visualise and interpret the outputs. You may one or multiple  tools. You may do it all in Python. You may use tightly-coupled visual  analytic approaches (e.g. Orange, V-Analytics) or loose-coupled  approaches such as file imports/exports from/to R, Python and Excel. You  may even write bits of your own software, though marks will only be for  the way the tools are applied.

Format and marking criteria

You must use this template.

The report length must be  less than 4000 words and up to 10 figures in total, obeying the  specific limitations given for each section. Exceeding the limit for any  section by more than 5% will decrease your mark,  because you are supposed to be able to express your ideas and describe  your work succinctly. The figures must have appropriate size to be  readable.

1. Problem statement (5%; <250 words). Describe  the analysis problem you aim to solve. Specify the phenomenon you are  studying (e.g., bank transactions, city traffic, social media, sport,  climate, etc) and the question(s) you strive to answer. Briefly present  the content and structure of the data you are going to analyse and  explain why these data are suitable for answering your questions.
   You’ll be marked on the (A) choice of the questions (how interesting and non-trivial they are) and (B) suitability of the data.
2. State of the art (15%; <500 words). Find  2-3 papers published in respected journals or conferences that deal  with similar data and problems and apply visualisations for solving the  problems (see the groupwork description and below for ideas). Briefly  present the specifics of the data they analysed, questions they  addressed, and the approaches they applied. Discuss to what extent these  approaches are applicable to your data and problem. Substantiate your  judgements by comparing your problem setting with those from the  surveyed papers, paying particular attention to any assumptions taken.  Specify what you have learned from these papers that is useful for your  study.
   You’ll be marked on the (A) choice of the papers to survey  (relevance to your problem, use of visual analytics approaches), (B)  quality of the summarised description of the problems and approaches,  and (C) quality of discussing the applicability of the approaches and  deriving useful lessons.
3. Properties of the data (15%; <500 words; <=2 images). Describe  in more detail the dataset(s) you are using. How were the data  collected? Describe the structure, i.e., the fields, their meanings,  types of the values. In case of two or more datasets used, describe how  they are linked, e.g., by references in one dataset to data items in  another dataset. Specify the amount of the data and the coverage  (territory in space, period in time, number of distinct objects, etc.)  In case of spatial and/or temporal data, specify the precision and  resolution.
   Describe your process of investigating the quality of the  data and the problems detected, such as erroneous values, missing  values, outliers, peculiar distributions, gaps or non-uniformity in  spatial and/or temporal coverage, noisiness, and biases. Include 1-2  figures with visualisations that helped you to explore the data quality.  Tell how you are going to account for the data properties and problems  in your analysis.
   You’ll be marked on the quality of the (A)  investigation of the data properties and (B) deriving conclusions for  conducting the further analysis.
4. Analysis (45%). The  analysis section includes 3 subsections: approach, process, and  results. The mark will be based on their joint content, but each  subsection must be present.
   • Analysis Approach (<500 words, 1 diagram). Present  your approach to solving the problem: plan of the analysis workflow  (steps of the analysis and their purposes) and methods applied. It is  advisable to represent the workflow schematically by a diagram.  Particularly specify where in the workflow human reasoning or judgement  is required in order to proceed with the analysis. What information  needs to be presented to the human analyst for this purpose? What  visualisation method is suitable for presenting this information?
     It  is expected that your problem is sufficiently complex and cannot be  solved simply by applying some algorithm and interpreting its outcomes.  It also should not be solvable simply by considering some visual  representation. Essential is involvement of human reasoning in the  analysis process, not only in making final interpretation and  conclusions. The human reasoning needs to be supported by appropriate  visual displays. However, the involvement of human reasoning needs to be  justified, so that the human does not do the work that can be done by  computers. Make sure that the human is doing what humans do best and the  computational methods are doing what computational methods do best.
     Apart  from visual analysis, the analytical workflow may include data  transformations and/or application of computational analysis methods.  The use of computational methods is not strictly required. It is  essential that results of some of the steps in the workflow are used in  the further analysis or affect the further analysis. The possible types  of results may be derived data, decisions of what method or  transformation to apply next, or choice of appropriate parameter  settings.
     You’ll be marked on the (A) appropriateness of the  approach, (B) appropriateness of the activity assigned to the human  analyst, and (C) interrelations between human cognition and computer  operations. The mark does not depend on the complexity of the chosen  methods.
   • Analysis Process (<1500 words, <=7 images). Describe  the process of your analysis, in which you apply the approach presented  in the previous subsection. Give particular attention to the human side  of the process, including interpretation of information displays,  reasoning about what is seen, making decision concerning the ways to  proceed with the analysis, and drawing conclusions that answer the  questions of your study. The key thing is to show how the repeated use  of visual and computational methods and/or data transformations helped  refine your answers and informed subsequent analytical steps and how –  ultimately – they generated the findings reported in the next section.  Show the stopping condition – i.e. how do you know when you’ve reached a  good solution? Justify your use of parameters and variables with your  investigations or with insights from other studies. Illustrate the  description of your analysis with images of the information displays  that were essential for your reasoning, decisions, and conclusions.
     You’ll  be marked on the (A) correctness of the use of the visual and  computational techniques, including justified selection of variables,  parameters, and transformations, (B) logics of your reasoning and  justification of the steps taken.
   • Analysis Results (<200 words, <=2 images). Summarise  the most important findings and use these to answer your research  questions. Illustrate the answers with appropriate visuals. Discuss the  implications of the findings for your chosen application/domain. If you  cannot answer your research questions fully, say so and expand in the  next section.
     You’ll be marked on how convincing your findings are and how well you apply them to your research questions.
5. Critical reflection (20%; <500 words). Reflect  on the chosen approach and the course of the analysis process. Discuss  where and how your thinking was important for taking the further steps  or for obtaining components of your answer. Discuss the role of the  visual representations for your reasoning. If you did not find complete  answers to your research questions, discuss the reasons, such as  properties of the data or weaknesses of the techniques you used.  Irrespectively of the results, try to imagine what could be done  differently, e.g., what software functionality could make your analysis  more efficient. Discuss the limitations of your approach (e.g.,  assumptions, particular requirements to data, disregard of some aspects,  etc.) and its potential applicability to other data and problems.  Present lessons you have learnt and give recommendations to other  analysts having to deal with similar data.
   You’ll be marked on  your ability to (A) think critically of what you have done and draw  useful lessons and (B) understand the limitations and generalisability  of your data, research questions and methods.
6. References. Your  report must include full references to the papers you reviewed in  Section 2 (State of the art). It may also include references to other  relevant materials, such as the source of the data, papers describing  the algorithms you used, web sites of the software tools you used, and  your own code (e.g., Python notebook). Put the list of references at the  end of the report. The references are not included in the word count.
7. Table of word counts.  At the end of the report, include a table with the word number in each  section. To count the words, you may use the functions of MS Word or  another text editor. The table content is not included in the word  count. Also, references, numerical content of tables and figure captions  that just describe the figure are not included in the word counts.

See below for an indicative descriptions what we expect at different grade levels, but please use the marking criteria listed above.

distinction

85-100AA+OutstandingThis  impressive work is an excellent piece of visual analytics. It is  extremely well-motivated, identifying excellent research questions that  are suitable for being answered by visual analytics approaches using the  datasets chosen. The analytical tasks and approaches are perfectly  suited for answering the research questions in which visual and  computation techniques fully inform each other. The analytical steps  perfectly demonstrate how visual analytics techniques were used to  arrive at the answer, including how the parameters used were chosen. The  findings clearly follow from the analytical steps. The reflection shows  a high degree of independent thought in its construction and a high  degree of clarity and creativity. This report is of publishable quality.80-84AExcellent75-79Very GoodThis  is a very good piece of visual analytics. It is well-motivated,  identifying very good research questions that are suitable for being  answered by visual analytics approaches using the datasets chosen. The  analytical tasks and approaches are very well suited for answering the  research questions in which visual and computation techniques inform  each other. The analytical steps demonstrate how visual analytics  techniques were used to arrive at the answer, including how the  parameters used were chosen. The findings follow from the analytical  steps. The reflection has attention to detail and clarity, but a more  creative synthesis could have improved this report.70-74A-

merit

67-69BB+GoodThis  is a good piece of visual analytics. It is well-motivated, identifying  good research questions that are suitable for being answered by visual  analytics approaches using the datasets chosen, though there may be some  limitations. The analytical tasks and approaches are able to help  answer the research questions in which visual and computation techniques  inform each other to some degree, but there may be more suitable  methods. The analytical steps demonstrate how visual analytics  techniques were used to arrive at the answer those of of the details may  be missing. The findings generally follow from the analytical steps,  though not completely. The reflection and discussion may be limited in  places and there may be some omissions.64-66B60-63B-

pass

57-59CC+SatisfactoryThis  is an adequate piece of visual analytics. It identifies research  questions that can answered with visual analytics with the datasets  chosen, though there will be limitations. The analytical tasks and  approaches can help answer the research questions the visual and  computational methods may not inform each other very well and may not be  entirely appropriate. The analytical steps indicate how the visual  analytics techniques helped answer the research questions but some  important details will be missing. The findings will follow from the  analytical steps to some degree. The reflection and discussion will be  limited and lack depth, comprehensiveness and clearness.54-56C50-53C-

fail

47-49DD+PoorThis  poor piece of visual analytics works to some degree but contains many  omissions. The report is poorly structured and papers may not illustrate  the approaches well. The report lacks clarity and critical reflection.44-46D40-43D-Very PoorThis  is a problematic submission in general. Although some knowledge of  basic concepts have been demonstrated, there is no clear structure. Or  it is not a valid submission.20-40EE0-20

Some tips

• Follow the structure provided – it’s in your interest to make it as easy for us to mark as possible!
• Don’t  forget that both positive and negative results are equally valid, as  long as your data and research methods are appropriate and you  understand the limitation of the methods.
• You’re marked on your  application of visual analytics to answering research questions. Marks  will only be awarded in the context of data analysis and its  interpretation. Credit will not be given for sophistication of coding,  complexity of techniques, quantity of statistical methods or similar,  unless it has direct analytical benefit.
• If you use a method like clustering, make sure you justify its use and the features you use to cluster.
• Please do use graphics, but only if directly referred to in the text. Don’t expect the marker to interpret them for you.