Overview: Machine learning (ML) is a set of methods to recognize patterns and make predictions from large datasets. City governments are increasingly using ML methods to help serve their citizens better. This workshop provides staff with an understanding of the nuts-and-bolts of ML in a smart city context. Focused on a simple but powerful family of models – supervised classification – we will walk through the key steps to build and deploy a predictive tool for city authorities. The workshop will familiarize participants with key ML concepts including training and test data, identifying important features for prediction, and model evaluation metrics. Having surveyed recent applied examples, from prioritizing building inspections to targeting energy efficiency upgrades, participants will be invited to brainstorm potential applications in a city of their choice.
Audience: The workshop is open to all Bank staff and consultants with no technical background needed.
Objectives:(i) understand the key steps to build and deploy a machine learning model for city authorities; (ii) identify untapped datasets and use cases where Machine Learning can help your clients.
Agenda:
12.00pm
Lunch
12.30pm
Block 1: Introduction and Interactive Example
Families of ML algorithms; identifying a business case. Case study: should the World Bank build more parking spaces or bike racks?
1.00pm
Block 2: Building a model
Get data; design predictive features; separate training and test data; select and train model.
1.30pm
Block 3: Evaluating a Model
Select a performance metric; measure and fine-tune performance; deploy the model.
1.30pm –
2.00pm
Block 4: Imagining Urban Use Cases
Example use cases; group exercise to identify potential uses and suitable urban data
About the Presenters
Jon Kastelan, Machine Learning Specialist, has advised governments and city agencies on analytics and machine learning across North America and Asia Pacific. He worked with PwC's Analytics and Urbanisation practice to support development of the Smart City blueprint for Hong Kong. He has also taught data science and analytics at New York University and worked with urban-focused start-ups.
Nick Jones, DRM Specialist at GFDRR, spent 2017-2018 as a CUSP Fellow at New York University. He led a graduate seminar program on urban data analysis, and worked on data science methods at the NYC Department of Buildings and Mayor’s Office of Data Analytics.
Calendar » [Presentation & Recording posted] Workshop on Machine Learning for Cities: From Key Concepts to Smart City Applications
[Presentation & Recording posted] Workshop on Machine Learning for Cities: From Key Concepts to Smart City Applications
Find the presentation and outcome slides of this event attached.
Access the source code to produce the figures in the slides and illustrates how the concepts are implemented.
Watch the recording of the event.
Opening Remarks
Hyoung Gun Wang, Senior Economist & Smart Cities KSB Lead, WBG
Presenters
Jon Kastelan, Machine Learning Specialist
Nick Jones, DRM Specialist, WBG
Overview: Machine learning (ML) is a set of methods to recognize patterns and make predictions from large datasets. City governments are increasingly using ML methods to help serve their citizens better. This workshop provides staff with an understanding of the nuts-and-bolts of ML in a smart city context. Focused on a simple but powerful family of models – supervised classification – we will walk through the key steps to build and deploy a predictive tool for city authorities. The workshop will familiarize participants with key ML concepts including training and test data, identifying important features for prediction, and model evaluation metrics. Having surveyed recent applied examples, from prioritizing building inspections to targeting energy efficiency upgrades, participants will be invited to brainstorm potential applications in a city of their choice.
Audience: The workshop is open to all Bank staff and consultants with no technical background needed.
Objectives: (i) understand the key steps to build and deploy a machine learning model for city authorities; (ii) identify untapped datasets and use cases where Machine Learning can help your clients.
Agenda:
12.00pm
Lunch
12.30pm
Block 1: Introduction and Interactive Example
Families of ML algorithms; identifying a business case. Case study: should the World Bank build more parking spaces or bike racks?
1.00pm
Block 2: Building a model
Get data; design predictive features; separate training and test data; select and train model.
1.30pm
Block 3: Evaluating a Model
Select a performance metric; measure and fine-tune performance; deploy the model.
1.30pm –
2.00pm
Block 4: Imagining Urban Use Cases
Example use cases; group exercise to identify potential uses and suitable urban data
About the Presenters
Jon Kastelan, Machine Learning Specialist, has advised governments and city agencies on analytics and machine learning across North America and Asia Pacific. He worked with PwC's Analytics and Urbanisation practice to support development of the Smart City blueprint for Hong Kong. He has also taught data science and analytics at New York University and worked with urban-focused start-ups.
Nick Jones, DRM Specialist at GFDRR, spent 2017-2018 as a CUSP Fellow at New York University. He led a graduate seminar program on urban data analysis, and worked on data science methods at the NYC Department of Buildings and Mayor’s Office of Data Analytics.