The value of machine learning in privacy

Date of publication: December 13, 2018

Catalogue: Webinars 2018


Market research is embracing insightful new sources of data. Among them, behavioural data is one of the most promising. It has proved to have an edge over survey data by overcoming human memory limitations and lack of sincerity. The challenge, however, in sharing clickstream data with third parties is to avoid violating individual's privacy rights, as defined in the GDPR. To overcome this difficulty, we developed our first "PII Filter", based on an intuitive principle: public web sites can be accessed by anyone; therefore, those URLs should be visited by several people. As a result, a new PII Filter has been developed based on a much more Aristotelic principle, learning from experience. This new PII Filter relies on a supervised predictive classifier: a rule-based algorithm that learns from a labelled data set of URLs.

Carlos Ochoa


This is a long description of some author details.

Marcello Frattini


This is a long description of some author details.

This could also be of interest:

Research Papers



Machine learning and privacy

Catalogue: Fusion 2019

Author: Anna Bellido

Company: Netquest

November 10, 2019

  • Video