A A A
Feed icon

SpaceTimeLab News

ming_tsiang_tsou_stlab

Big Data and Human Dynamics

Published: Jul 18, 2016 2:05:43 PM

cpc_caulcott

CPC Closing Workshop - Big Data and Intelligent Policing

Published: Jun 22, 2016 10:09:38 AM

Big Data for Intelligent Policing

Published: Jun 8, 2016 5:13:04 PM

SpaceTimeLabLogo

GIScience 2016 Workshop

Machine Learning Methods for Spatial and Temporal Analysis

We are pleased to announce that SpaceTimeLab for Big Data Analytics will be running a workshop on Machine Learning Methods for Spatial and Temporal Analysis at GIScience 2016 on the 27th of September. If you are attending the conference and would like to participate you can register through the conference website: http://giscience.geog.mcgill.ca/. Resources will be published on this site in the near future. We look forward to seeing you there.

Preliminary details
Description of topics Machine learning (ML) methods have gained popularity in the GIScience community over the past few decades due to their success in dealing with the nonlinearities and heterogeneities of spatial and temporal datasets. However, the uptake of such methods is limited by their perceived difficulty. This workshop aims to provide a gentle, practical introduction to ML methods for addressing two common problems in spatial and temporal analysis: classification and regression. Attendees will be taught the key concepts underpinning a range of ML algorithms, including support vector machines and random forests. They will then be taught the essential skills necessary to train and test ML models using R statistical package. A number of real world datasets will be used as examples, including GPS tracks, road traffic data and environmental data. The workshop will conclude with a discussion of some the limitations of ML methods, advanced topics and future research directions.
Format of workshop The workshop will be delivered over two sessions of three hours. Each session will be a combination of introductory lectures (1 hour) and computer based tutorial sessions (2 hours). Attendees will be introduced to the key concepts and tools in the lecture before applying their knowledge in the tutorial.
Workshop outline

Lecture 1 (1 hour): An introduction to machine learning (ML) for spatial and temporal analysis

· An overview of ML

· ML methods for classification

· Applications in spatial and temporal analysis

Tutorial 1 (2 hours): ML methods for classification

· Brief introduction/refresher on R statistical package

· ML classification algorithms in R

· Training classifiers

Lecture 2 (30 minutes): ML methods for Regression

· ML methods for regression

· Applications in spatial and temporal analysis

Tutorial 2 (2 hours): ML methods for Regression

· ML regression algorithms in R

· Training regression models

Summary Lecture (30 minutes): Recap, advanced topics in ML and spatio-temporal analysis, SpaceTimeLab showcase

· Limitations of ML methods

· Examples of ML and ST-analysis at SpaceTimeLab

· Research directions in ML for ST-analysis

Intended audience The course is intended to be introductory. The aim is to teach the core algorithms to give attendees practical experience, and leave them with the skills and ideas to extend their knowledge further. This course will be suitable for undergraduate students, postgraduate students, or other researchers with a background in geographic information science and/or computer science. Some knowledge of R statistical package or a similar (e.g. Matlab or a programming language) is desirable.
Maximum number of participants 30

Page last modified on 11 feb 16 15:13