IS LAB Series

Machine Learning on Educational and Social Data with HCI Analysis

Cristian Mihaescu

University of Craiova - Romania

17 April 2018

DI-A1, 10h30

Theme:

Data analysis has lately become one of the hottest research areas due to the wide range of algorithms and implementations that are currently used for solving tasks from various application areas. At University of Craiova, Romania we are a group that has developed Tesys e-Learning platform and implemented various data analysis pipelines on its activity data, as well as on social activity data. Finally, we have managed to perform HCI usability studies in order to further improve Tesys. Among the tackeled tasks there are: building an advanced classifier and its integration with Weka, extracting patterns from educational traces, MTree clustering algorithm in Weka, exploring the perceived ease of use in Tesys, generating alerts for drops in activity levels, detection of "spam" and "don't care users". Our current and future research consists in improving MTree clusterer, defining new methods for obtaining optimal K in datasets and application of ML in brain-computer interface.

The Convergence of Foster's Methodology for Parallel Program Development and MDA

Mihai Mocanu

University of Craiova - Romania

17 April 2018

DI-A1, 10h30

Theme:

Today, to build from scratch or change a software system in a correct manner, it is more important to keep the rational fundaments behind design decisions than the decisions themselves or their results. This is a clear sign for the increase in importance of models, but also for a paradigm shift, to a process centered on MDA (Model Driven Architecture) and associated standards, that aims to build platform-independent applications. Could this be true for parallel systems development as well? If so, we may have additional benefits, for example: we can omit programming language details and delay or even automate decisions related to the specific of the execution platform. In an era when the interest in high performance computing (HPC) is worldwide, and growing in many new markets, the stakes are huge. In this respect, adoption of the well-established methodology of Ian Foster that underlies a software development process (related to parallel algorithm design), revised based on the convergence with MDA principles, could be very rewarding."

Poster

Banner