Belief Nets for Mental Health

Abstract

A Bayesian network is a powerful tool for inference; its structure and parameters can be learnt from data. The objectives of this thesis are first to find the structure associated
with the suicide risk of a patient, to compare this structure with that of the expert and then to predict the risk itself. In order to do so, we first learn the structure using algorithms such as K2 or MWST. After having obtained this structure we try to
predict the risk of suicide, with this found structure, on testing data. The results, which are quite good, will be discussed in this thesis, but the lack of data for training part has a significant effect.

Date of Award	Sept 2005
Original language	English
Awarding Institution	Aston University