Università della Svizzera italiana Home

Predictive inference and computational statistics

Statistics underlies every part of the data science value chain.  Starting with pre-processing, filtering and mining the data, statistically significant relationships between variables are uncovered and then used to create models that can effectively predict future outcomes. Learning is introduced into the process by estimating and restructuring these predictive models multiple times, updating prior information using the Bayesian paradigm.

Computational statistics involves designing algorithms that implement intensive statistical methodologies using high performance computers.  Among the methods employed are Monte Carlo algorithms, which resolve the problems intrinsic in estimating complex Bayesian predictive models by simulating from the posterior distribution of the model parameters. With the advent of big data, this knowledge area has become an essential part of scientific investigation across multiple disciplines.  The many applications include extracting actionable knowledge from hidden patterns, discovering unknown correlations, predicting market trends, identifying customer preferences and analyzing information related to social behavior that is buried in the digital crumbs we leave behind as we increasingly entrust digital devices to manage our lives.

Research topics include:

  • Performing sentiment analysis on geo-referenced social media data to define real-time social well-being indices that allow nowcasting of official statistics
  • Understanding economic complexity to develop a basis for competitive advantage in a country
  • Analyzing geographical and meteorological information together with paramedic data to improve medical emergency response time

Link to Knowledge Area:

Social Network Analysis Research Center (SoNAR-C)

Knowledge Area Leader: Antonietta Mira


    All events