III: Small: Adopting Machine Learning Techniques for Big Spatial and Spatio-temporal Data and Applications

Motivated by the ubiquity and magnitude of spatial and location data collected from myriad of devices, including, cell phones, GPS-enabled vehicles, and satellites, this award aims to take advantage of such data to enhance the scalability, accuracy, and usability of various artificial intelligence and machine learning techniques. This will advance the state-of-the-art of these techniques not only to accommodate more applications where spatial data is predominant (e.g., transportation, meteorology, and agriculture), but also to current applications that deal with spatial data (e.g., census, knowledge-base construction).

This award adopts machine learning techniques for big spatial and spatio-temporal data and applications in two orthogonal, but related, directions. First, injecting the spatial awareness inside machine learning techniques and applications, which would result in a higher accuracy for such applications. Second, taking advantage of the recent advances in machine learning techniques to boost the usability, scalability, and accuracy of long lasting spatial and spatio-temporal data analysis techniques. Towards its goal, this award exploits three main research topics: (1) Developing the concept and building Spatial Markov Logic Network (SMLN) by injecting the spatial awareness inside Markov Logic Network. (2) Leveraging the developed SMLN to build the first open-source system for a spatially-aware knowledge base construction system. (3) Leveraging the developed SMLN to increase the accuracy, scalability, and usability of several long lasting spatial analysis problems, including spatial autologistic regression, spatial classification, and Bayesian networks.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.