In this talk, the audience will learn how to build from scratch with Python machine learning models for predicting spatiotemporal activities in cities using location data.
The growing availability of data from cities (e.g., traffic flow, human mobility and geographical data) opens new opportunities for predicting and thus optimizing human activities. These may include human mobility, land use classification, event detection and location recommendation.
- How do we characterize the main activity of an area (e.g. Business,
- How to detect and forecast unusual spatiotemporal events (e.g. protest, sport game, concert)?
- How to understand why people are moving between places?
This talk will try to answers these questions explaining how, making use of Python, it is possible to (i) collect and aggregate geo-located data; and (ii) build machine learning models to predict spatiotemporal activities. For example, an unusual activity in a local area at a specific time can be predicted by analyzing the text from geo-located tweets.
Firstly, we will present an organized picture of geo-located data such as mobile phone data, geo-located texts from social networks (e.g. Foursquare and Twitter), census and Open Street Map. Then, we will briefly discuss how to collect and aggregate such data using Python packages such as GeoPandas and OSMnx. Finally, we will cover topics related to cities activities including event detection and forecasting, human mobility, and location recommendation and prediction.
As location data, we will mostly use samples from Open Street Map and Foursquare. The example code will make use of (basics) Pandas and GeoPandas. However, this is a beginner talk and it will be self-contained as much as possible.
in __on domenica 22 aprile at 12:45 **See schedule**