Presentation is loading. Please wait.

Presentation is loading. Please wait.

How people visually explore geospatial data Urška Demšar Geoinformatics, Dept of Urban Planning and Environment Royal Institute of Technology (KTH), Stockholm,

Similar presentations


Presentation on theme: "How people visually explore geospatial data Urška Demšar Geoinformatics, Dept of Urban Planning and Environment Royal Institute of Technology (KTH), Stockholm,"— Presentation transcript:

1 How people visually explore geospatial data Urška Demšar Geoinformatics, Dept of Urban Planning and Environment Royal Institute of Technology (KTH), Stockholm, Sweden ICA WS on Geospatial Analysis and Modeling 8 th July 2006 Vienna

2 Developing geovisualisation tools Developing a usable and useful information system User-centred design Human-Computer Interaction (HCI) Knowledge about users and how they use the system Geovisualisation tools and systems visual exploration analysis presentation of geospatial data For a long time: technology-driven development A recent shift in attitude: user-centred development

3 Usefulness Utility Usability Can the functionality of the system do what is needed? How well can typical users use the system? Usability evaluation Process of systematically collecting & analysing data on how users use the system for a particular task in a particular environment. User-centred design Usability of an information system is the extent to which the system supports users to achieve specific goals in a given context of use and to do so effectively, efficiently and in a satisfactory way. Nielsen 1993 of a computer system Evaluate systems functionality Assess users experience Identify specific problems

4 Usability testing Formal evaluation Exploratory usability User testing Observing users Measuring the accuracy and efficiency of users performance on typical tasks Assessing how the users work with the system performing predefined tasks questionnaires thinking-aloud methodology observation, video controlled measurements: errors, time descriptive data: verbal protocols Qualitative evaluation Quantitative evaluation Methods complement each other! evaluation through user participation

5 Exploratory usability experiment GeoVISTA - based visual data mining system Dataset with clearly observable spatial and other patterns Exploratory usability experiment How people visually explore geospatial data? Which exploration strategies they adopt? Which visualisations they prefer to use? Formal usability issues: Edsall 2003, Robinson et al. 2005

6 Data Iris setosaIris versicolorIris virginica Iris dataset - famous from pattern recognition Fischer plants, 50 in each class, 4 attributes Linear separability in attribute space Original dataset new attributes plant measurements bedrock soil landuse put in a spatial context Linear separability in geographic space

7 Visual data mining: Data mining method which uses visualisation as a communication channel between the user and the computer to discover new patterns. Data exploration by visual data mining Data mining = a form of pattern recognition the human brain The best pattern recognition apparatus How to use it in data mining? Computers communicate with humans visually. Computerised data visualisation

8 Visualisations geoMap Multiform bivariate matrix Parallel Coordinates Plot (PCP) Brushing & linking + interactive selection Exploration system Gahegan et al. 2002, Takatsuka and Gahegan 2002 GeoVISTA Studio

9 Participants Small number of participants: 6Discount usability engineering Nielsen 1994, Tobon 2002 The majority of the usability issues are detected with 3-5 participants. cost & staff limitations Students of the International Master Programme in Geodesy and Geoinformatics at KTH non-native English speakers, fluent in English nationality/ mother tongue Ghanian Russian Slovenian Spanish Swedish gender 50/50 engineering background familiar with GIS voluntary participation Not colour-blind

10 Experiment design 1. Introduction: - what the test was about, consent for using the data, etc. Usability test in English performed individually under observation h per participant 5 steps 2. Background questionnaire: - gathering information on gender, mother tongue, background, etc. 3. Training: (unlimited time: ca min per participant) - introduction to data and visual data mining system - independent work though a script - questions allowed

11 4. Free exploration: (limited time: 15 min per participant) - whatever exploration in whatever way the participant wanted - no questions allowed - Verbal Protocol analysis – thinking-aloud - cooperative evaluation: if the participant stops talking, the observer can ask questions (What are you trying to do?, What are you thinking now?) 5. Rating questionnaire: - gathering information on participants opinion about the system - measuring perceived usefulness & learnability The main part of the test

12 Results 1. Perceived usefulness & learnability The bivariate matrix the easiest to use. The map the easiest to understand. The PCP the most difficult to understand and use. 2. Exploratory usability Analysis of the thinking-aloud protocols Hypotheses extraction classification acc. to source background knowledge prompted by a visual pattern refinement of a previous hypothesis Counting visualisations total frequency relative frequency

13 Hypotheses classification background knowledge prompted by a visual pattern Refinement of a previous hypothesis Higher flowers probably have longer leaves. Are sepal length and sepal width correlated? There seem to be two clusters in each of these scatterplots. Not only are there two clusters, but the big cluster consists of two subclusters according to petal length. assign colour acc. to petal length. Flowers of the same species probably grow in the same area.

14 Visualisation frequencies Hypotheses generated f R (i,j)=f T (i,j)/N j i – visualisation j – participant Relative frequency:

15 Browse Form ideas or hypotheses Manipulate graphicsInterpret data Amend initial idea according to new information Look for content Adjust browsing/ decide where to look Gather evidence Get new/more information Evaluate initial idea Adjust browsing/ decide where to look Tobon Exploration strategies Model of the visual investigation of data 3 groups mapping the strategies as paths

16 Browse Form ideas or hypotheses Manipulate graphicsInterpret data Amend initial idea according to new information Look for content Adjust browsing/ decide where to look Gather evidence Get new/more information Evaluate initial idea Adjust browsing/ decide where to look Strategy no. 1: Confirm/reject a hypothesis based on background knowledge and then discard it. Repeat from the start. Confirming a priori hypothesis

17 Browse Form ideas or hypotheses Manipulate graphicsInterpret data Amend initial idea according to new information Look for content Adjust browsing/ decide where to look Gather evidence Get new/more information Evaluate initial idea Adjust browsing/ decide where to look Strategy no. 2: Form a hypothesis based on what you see, interpret and adapt it, confirm/ reject it and discard it. Repeat from the start. Confirming a hypothesis based on a visual pattern

18 Browse Form ideas or hypotheses Manipulate graphicsInterpret data Amend initial idea according to new information Look for content Adjust browsing/ decide where to look Gather evidence Get new/more information Evaluate initial idea Adjust browsing/ decide where to look Strategy of group no. 3: Form a hypothesis based on what you see, explore further and adapt/refine it, according to what you see in other visualisations, confirm the refined version or adapt again and continue. Seamless exploration

19 Small study size: - conclusions can not be too general, observations only Conclusions Training necessary: - new concepts visual data mining unusual visualisations interactivity of geoVISTA-based tools Cooperative evaluation vs. strict thinking-aloud: - cooperative evaluation better (compared to a previous experiment) - no silent participants - easier to keep protocols Discrepancy in perceived vs. actual learnability: - PCP very difficult to understand - PCP used most frequently of all visualisations - spaceFills almost never used

20 Exploration strategies: - three different exploration strategies not related togender academic background nationality/mother tongue GIS experience Investigating spatial data visually is not so simple! Substantial interpersonal differences in forming exploration strategies Why?Question for the future

21 Thank you!


Download ppt "How people visually explore geospatial data Urška Demšar Geoinformatics, Dept of Urban Planning and Environment Royal Institute of Technology (KTH), Stockholm,"

Similar presentations


Ads by Google