Loading...

​​EuroVis 2019 Plan

​​Writing-up/reading-up:

• ​​[5/11] Feature Engineering through Visual Analytics/Visualisation — VAST papers [Mit]

• ​​[12/11] Write-up the skeleton/abstract/points for the Introduction [Cagatay] ​+EuroVis 2019 - Paper Skeleton​ 

• ​​[12/11] Related work section

• ​​Email visualisation

• ​​Feature engineering

• ​​Active learning

​​

​​Next meeting: Monday 12/11

​​

​​Implementation:

• ​​Generate initial labels for a sample

• ​​T-sne projection of thread features

• ​​assign labels to strong clusters

• ​​examine ‘threads of interest’, possibly identify interesting patterns (labels) as well

• ​​Look at the thread features view

• ​​labelling interface

• ​​Find active learning library to use

• ​​Loop

• ​​Build active learning model ( https://github.com/modAL-python/modAL)

• ​​Get sample to label (export to local file)

• ​​Show the sample in the visualisation and support labelling

• ​​Export labelled sample

• ​​Raw data features [Cagatay]

• ​​# messages, # unique senders (list here ​+Threadlets: Building a Vocabulary of Communication Patterns through Interactive Visual Analysis​ )

​​

• ​​Individual perspective [If time allows]

• ​​Individual-based representation (Adjacency matrix) [Phong]

• ​​Individual-based features [Cagatay] 

​​

​​Evaluation:

• ​​Methodology: 

• ​​Case Study Based [Mit]

• ​​Quantitative Evaluation [CT] [To be considered if feasible]

• ​​Build a classifier with the solution

• ​​Run the classifier on unseen data (what would be cool is to have the Redsift email data)

• ​​Get people to evaluate the the classification results (sanity check!)

​​