Luis Torgo

Canada Research Chair and Professor

Faculty of Computer Science, Dalhousie University

I am a Canada Research Chair (Tier 1) on Spatiotemporal Ocean Data Analytics and a Professor at the Faculty of Computer Science of the Dalhousie University, Canada. I also hold appointments as an Associate Professor of the Department of Computer Science of the Faculty of Sciences of the University of Porto, Portugal, and as an invited professor of the Stern Business School of the New York University where I have been teaching in recent years at the Master of Science in Business Analytics.

At Dalhousie University I carry out my research in the context of the Institute for Big Data Analytics, where I act as Deputy Director of the institute and lead the Ocean Data Analytics research lab. I am the Head of the Research Cluster on Big Data Analytics, AI and Machine Learning at the Faculty of Computer Science of the Dalhousie University. I am also a senior researcher of LIAAD / INESC Tec and an associate member of the Artificial Intelligence Institute of the University of Waikato.

My research revolves around the general area of Data Science, with a strong focus on Predictive Analytics for data with spatial, temporal or spatiotemporal dependencies. Recently, I’ve been focusing my work on modelling rare events, with applications to fraud detection, prediction of extreme values and monitoring activities for anticipating anomalous behavior. I love to see my research being applied and thus I try to maintain a network of collaborations with researchers from other areas.

My favourite tool is the R programming language and environment. I am the author of the widely acclaimed Data Mining with R book published by CRC Press in 2010 with a strongly revised second edition that appeared in 2017.

I am also the CEO and one of the founding partners of KNOYDA a company devoted to training and consulting within data science.

In case you want to know more about me you may have a look at my full CV.

Interests

Machine Learning and Data Science
Utility-based Predictive Analytics
Spatiotemporal Analytics
Ocean Data Analytics
Rare events
Forecasting

Education

PhD in Computer Science, 2000

University of Porto, Portugal
Systems and Informatics Engineering, 1989

University of Minho, Portugal

Main Research Lines

Ocean Data Analytics

Applying Data Science Methods to Ocean Data

2018-10-16

Modeling Dependent Data

Forecasting models for dealing with data with dependencies (e.g. temporal, spatial, spatiotempora, etc.)

2018-10-16

Imbalanced Distributions

Methods for handling predictive tasks with imbalanced distributions of the target variable

2018-10-16

Explainable Machine Learning

Methods for helping end users to better understand complext machine learning models

2018-10-16

Featured Publications

Vitor Cerqueira, Luis Torgo, Igor Mozetic

January 2020 Machine Learning

Evaluating time series forecasting models: An empirical study on performance estimation methods

This paper describes an empirical study and comparison of different performance estimation methods in the context of time series forecasting. published in Machine Learning journal Abstract. Performance estimation aims at estimating the loss that a predictive model will incur on unseen data.

PDF

Vitor Cerqueira, Luis Torgo, F. Pinto, C. Soares

January 2019 Machine Learning

Arbitrage of Forecasting Experts

This paper describes our award-winning (Best Student Machine Learning Paper Award given by the Machine Learning Journal at the European Conference on Machine Learnig (ECML/PKDD’2017)) work on ensembles for time series forecasting.

PDF Code Source Document DOI

Luís Torgo

January 2017 Chapman and Hall/CRC

Data Mining with R: Learning with Case Studies, Second Edition

This is the second edition of my widely acclaimed Data Mining with R book initially published by CRC Press in 2010, with this strongly revised second edition appearing in 2017.

Code Source Document

Paula Branco, Luis Torgo, Rita Ribeiro

January 2016 ACM Comput. Surv.

A Survey of Predictive Modeling on Imbalanced Domains

This paper presents an extensive survey of predictive analytics methods for handling problems where the distribution of the target variable (norminal/classification or numeric/regression) is highly imbalanced, and moreover the less frequent values are the more relevant for the end-user.

PDF Source Document