Deepti Joshi

Research

Big Data Analytics and Processing

Confluence of global positioning system-based sensor systems, satellite technology, and motivation to monitor the Earth’s natural and human resources has resulted in massive explosion of geospatial datasets (i.e., data with associated explicit or implicit geographic coordinates). In addition, widespread use of location aware devices (e.g., smart phones) has facilitated a new mode of data collection by volunteers and resulted in a new paradigm called Volunteered Geographic Information (VGI) computing (Goodchild, 2007). This form of crowd sourcing for data collection has resulted in massive data being collected that fills gaps in the information collected by institutional sources (e.g., governmental and research) at spatial and temporal resolutions that were not feasible before. The resultant emergent collective intelligence from these data will play an increasingly greater role in the understanding of the Earth’s processes and their interactions and has the potential to fundamentally transform how humans interface with their natural and managed ecosystems. These datasets provide new pathways to understand the ecosystems, biological and epidemiological processes, and human activities and their impact. This kind of data-driven scientific inquiry can accelerate the progress of discovery and lead to improved decision making. My interest lies in maximizing the information potential of these datasets by developing new scalable computational techniques based on new mathematical and statistical foundations to examine these disparate, diverse(in collection, dissemination, and usage) datasets as a whole (not individually) and to exploit the unique properties of the geographic space.

Spatio-temporal data mining

Every entity in this world has a geospatial location that allows us to answer questions such as where it is, where it occured, where it belongs to, etc. Similary everything also has time attached to it. It may be a single point in time when it occured, or a window of time which may denote its lifetime. Making use of the spatio-temporal inofrmation related to the objects or entities that exist in the world, in addition with everything else that we know of the objects to study the relationships that may exist in them is a very challenging problem. My contributions have been specifically in incorporating the geospatial properties of objects within clustering algorithms thus producing more robust and scalable geospatial clustering algorithms. In addition, I have also developed a clustering algorithm that treats both time and space as first-class citizens, and produces clusters that span across both the spatial and temporal dimensions.

Currently, I am working to further develop algorithms for the analysis of spatio-temporal clusters that will allow us to identy the movements of clusters along both the spatial and temporal dimensions.

New Project: TWIG

Population Density KML file: KML