Supervised and Semi-supervised approaches to sample growing and online prediction of customer intent

The company collects survey data from websites which is combined with behavioural data from survey respondents. This combined data set is information rich but can be too sparse for modelling purposes in a straight forward supervised learning context. As such, on-going research concerns optimizing the process by which non-survey related behavioural data can be leveraged to improve the robustness and efficacy of supervised models built using the combined survey/behavioural data alone. In addition, these models can be used to optimize and improve online predictive algorithms, where behavioural data from web site users is modelled in real-time to create profiles or to label traffic as indicative as one group/class or another.

Faculty Supervisor:

Layachi Bentabet

Student:

Tegan Maharaj

Partner:

iPerceptions

Discipline:

Computer science

Sector:

Information and communications technologies

University:

Bishop's University