| |
My research endeavour broadly revolves around (A) Systems sensing and mining & (B) Social and complex networks mining.
(A) Systems sensing: Mobile affective computing is one of my significant and unique research contribution. In this project, we developed smartphone based emotion detection framework, which determines emotions (say, stress, depression etc) based on the users’ smartphone engagement (say, keystrokes, facial images, recorded audio etc.) patterns. Moreover, we leverage on the users' smartphone & wearable (say, smartwatch) engagements to detect her physical activities, as well as mine her ambience and mobility contexts. For instance, in an organization, individuals prefer to form various formal and informal groups for mutual interactions.
We developed a lightweight, unsupervised, yet near-accurate methodology to detect
various interacting groups based on collective sensing through users' smartphones. The gained expertise in systems allowed me to develop smartphone based assistive technology for commuters in public & private transport systems. This includes the development of (i) navigation system in public buses in the absence of GPS, (ii) generate a rich annotated city transit map with minimal GPS usage, (iii) route recommendation systems for commuters based on their own comfort definition. I have developed a unique expertise to work at the interface of data science and system science. For instance, sudden slowdown or shutdown of large scale enterprise applications affect a large population of customers. We proposed AI driven methodologies for early detection of anomalies & troubleshooting in large-scale systems. A brief summary of my systems research can be found here.
(B) Social network mining: We investigated various structural and dynamical behaviors of social network through the lens of data science and network science. For instance, we explored the information diffusion dynamics taking place in various online social networks, say in Twitter, Yelp etc. Long chain of retweets from the original tweet leads to the formation of a cascade, leading to a rapid information diffusion over the underlying social (follower) network. We developed analytical frameworks for modelling the cascade formation considering both retweet and mention activities in Twitter. Location-based social networks (say, Yelp) allow customers to share the locations that they have visited in a number of ways (say, via 'tips' and 'reviews'). We proposed a graph based recommendation framework, which leverages on geo-social correlations to recommend new locations to the customers to visit. Moreover, in the context of event-based social network (say, Meetup), we developed a deep learning based venue recommendation system, which provides context driven venue recommendations for the Meetup event-hosts to organize the popular upcoming events. Additionally, we have made few fundamental contributions in complex network and data science. For instance, we developed a fast and scalable network embedding framework, which is capable of generating high quality embeddings for billion scale networks. A brief summary of my social network research can be found here.
Most of my research projects are funded by grants from DST, MHRD, NetApp, TCS Innovation Lab, Huawei, DAAD etc.
|