Publication of the week: Dongxu Han, Hongbo Du & Sabah Jassim
19 December 2016
“Towards a Confidence-Centric Classification Based on Gaussian Models and Bayesian Principles”, Proceedings of 9th York Doctoral Symposium on Computer Science and Electronics (University of York, November 2016), 46-56.
Classification is an essential form of machine learning. It has been widely used in various application areas. Conventional classification schemes are mostly interested in the class label outcome rather than the strength of the class predictions. This paper introduced a new interpretation of “classification confidence” to complement the predicted class label. The level of confidence is formulated based on Gaussian probability models and the Bayesian classification principles. The paper shows the soundness of the concept and argues the very essence of embedding it in the centre of a classification process. The paper applies the proposed concept over a data set of miscarriage cases in early pregnancy and demonstrates that the concept works well and indeed reflects the degree of certainty regarding the data set.
The paper is based on Dongxu’s initial work in his first year of PhD research. The paper was peer-reviewed and accepted by the 9th York Doctoral Symposium on Computer Science and Electronics, an annual international forum for PhD researchers to share their research ideas and experiences, held at the University of York. Dongxu won the Best Presentation Prize at the conference. Dongxu’s attendance at the conference was funded by the Postgraduate Research Student Conference Fund of the University.
The paper can be downloaded from the conference website.
After graduating with a first class honours degree in 2015, Dongxu is currently an MPhil/DPhil research student in Applied Computing, conducting research in machine learning based decision support systems (DSS) for supporting medical diagnostic decisions and most effective treatments of patients under the joint supervision by Prof Sabah Jassim and Mr Hongbo Du. The research is a collaborative work between the Applied Computing and Queen Charlotte and Chelsea Hospital, Imperial College London and KU Hospital Leuven, Belgium. Dongxu’s current work is one important component of the proposed DSS framework.