SPATIAL DATA SCIENCE CONFERENCE

Machine learning on spatial data is hard. This talk will look at some of the unique issues when working with this data, how Dstillery cleans and classifies this data in real time a quarter trillion times a day, and why we ended up building two parallel spatial featurizers to maximize the utility of geo data in our machine learning stack.

In this talk, Dr. Shan Jiang will discuss geo-computing, modeling and visualization methods to extract from massive and passive spatiotemporal mobile phone data to advance understanding of human mobility in urban environments. She will provide a range of examples from Boston to Singapore, to demonstrate the usage of big data for travel behavior modeling and analysis, which will allow cities to monitor and forecast travel timely and efficiently, and make informed decisions on sustainable infrastructure planning and investment.

The nuances of statistical regression modelling with spatial data will be discussed, and solutions through “spatial” regression modelling options will be summarized. A particular solution for a public health issue will then be illustrated.

As cities grow in size, density, and complexity, companies need to re-think their distribution strategies to operate profitably in these constrained environments. In this talk, I will discuss how our team at the MIT Megacity Logistics Lab is leveraging different sources of data, including geo-spatial datasets, to better inform strategic and planning decisions in urban logistics. At its core, logistics is about moving goods between different locations. Thus, geo-spatial data plays a key role in enabling better decision-making. I will present several case studies to illustrate the combined use of logistics planning methods, data science algorithms and varied sources of data to support decisions in last-mile logistics.