This project investigates the susceptibility of jobs to computerisation and particularly which features of a job determine the probability of computerisation. This is achieved by using Gaussian Process Classification. A set of labelled occupations is used to train and test the model and the effect of using different feature sets is explored. Feature selection in the form of greedy selection is used to find the feature set that achieves the best classification and thereby find the features that are most significant when determining if a job can be computerised. It is concluded that the most important feature is Originality and the best feature set for classifying the data in this project consists of Originality and Service Orientation. Furthermore, experiments are performed using linear embedding methods for feature learning. However, these experiments fail to prove that better classification can be achieved using this method.
Other Recent Journal Article / Working Papers
(Inter)nationalising the antibiotic research and development pipeline
The technological and economic prospects for CO2 utilization and removal
Nudging Immunity: The Case for Vaccinating Children in School and Day Care by Default
Emergence of the Asian lineage of Zika virus in Angola: an outbreak investigation
Measuring productivity dispersion: a parametric approach using the Lévy alpha-stable distribution
Net zero for agriculture