MSc Data Science (with Distinction)
School of Mathematical and Computer Science,
Heriot-Watt University, Edinburgh, EH14 4AS
ag72 at hw dot ac dot uk
+44 (0)131 451 4166
My main interest is to study and experiment possible ways of making process and results of unsupervised machine learning algorithms especially topic modelling more interpret-able and intuitive, primarily focusing on visualisation techniques I like to present explanations of why and how topic modelling produced these results. Explanatory visualisations can be used by high profile stakeholders for strategic planning as well as addressing the “right to explanation” issue of European Union’s new General Data Protection Regulation. As a means of increasing stakeholder trust and helping them defend their decisions to third parties, for instance in situations where decision makers use the results for strategic planning involving significant budget or resource that affect other people.
Visualising results of unsupervised algorithms
Each run of unsupervised algorithms produce different output on the same data. End-users often are not aware of this which I believe is an ethical issue but on the other hand knowing this will have a negative effect on their trust and confidence on such algorithms. I currently work on giving user option to select the BEST output for their specific needs with data-driven explanatory visualisation which they can use to defend their decisions.