Doorley, Ronan & Noyman, Ariel & Sakai, Yasushi & Larson, Kent. (2019). What's your MoCho? Real-time Mode Choice Prediction Using Discrete Choice Models and a HCI Platform. UrbComp SIGKDD 2019

Abstract

The impact of city-planning on mobility habits of urban dwellers has been proven crucial to well functioning cities. Nevertheless, the correlation between discrete urban interventions and metropolitan scale mobility mode-choices (MC) is challenging to predict and communicate. This paper presents the design and deployment of 'MoCho', a real-time MC modelling, prediction and collaboration platform. MoCho aims to predict and simulate MC of individuals in a metro region in response to real-time urban design iterations. The prediction models consider individual characteristics and attributes of available alternatives and are calibrated using survey data. To explore MoCho MC predictions, users interact with CityScope, a compu-tangible user-interface which triggers new MC predictions and their impacts based on interactive design of land-use, density or spatial proximity. Finally, a distributed computational system delivers real-time predictions onto a web-based user-interface. In 2018, a MoCho instance has been developed and deployed to simulate MC for the Boston metro area, focusing on a 14 acres development site in Kendall Sq. Cambridge, MA. The choice model was well fitted and the parameters showed significant associations with a range of explanatory variables including travel times, residential and employment densities and personal attributes like age, gender , education-level and home-ownership. Such a combination of an intuitive TUI and well-calibrated prediction models can allow experts and non-experts alike to participate in an evidence-based urban design process. Code for MoCho MC model and front-end is available here: https://github.com/CityScope/CS_choiceModels