Spatio-temporal Transportation Forecasting

The team of Amazon EU Transport Network Planning is responsible for one of the world’s largest transportation operations, planning a network that on the busiest day is being used by trucks that depart every minute across Europe.

We analyse enormous volumes of data to help Amazon think through how best to organize a quickly growing and changing network. We are constantly engaging with new kinds of problems, and use a wide range of analytical methods ranging from Econometrics, Machine Learning and Simulation, as well as tools to reach the right solutions, including Tableau, Quicksight, SQL/Redshift, Python and R.

One of our most recent developments is a zip-code level, spatiotemporal forecast that allows transportation stakeholders, such as commercial carrier managers, transport network planners and network engineers to take strategic decisions on the development of the rapidly expanding Amazon Transportation Network. The current model combines existing short- and long-term transportation, supply chain and finance forecasts, in a unique way; however, there is great potential for improvement by employing statistical and/or Machine Learning algorithms, better visualization tools and AWS technologies.

The goal of this project is:

1. To investigate the use of spatial econometric models such as Spatial Error Model, Spatial Durbin Model, Geographically Weighted Regression, spatial clustering, Getis-Ord index, spatial classification and/or Long short-term memory (LSTM) recurrent neural networks (RNN) in order to improve the forecasting accuracy of the model.
2. Employ AWS technologies to increase speed and accessibility to the final product
3. To build an R Shiny App.