A huge amount of data are being collected these days in almost every sector of life. These data need to be analysed for knowledge discovery and future prediction. Decision trees and forests being a white box approach are suitable for knowledge discovery from these data. Discovered knowledge can be used for future prediction. Individual accuracy of each base tree and the diversity among the base trees of a forest play an important role in achieving a high ensemble accuracy of a forest. While accuracy is an important goal for forests it may not be a good indicator of the usefulness of a forest for cost sensitive and/or class imbalanced datasets. Decision forest algorithms may need to be fine tuned further to achieve a better outcome for these cost sensitive or class imbalanced datasets or even for datasets that require preserving data privacy. In this presentation, we will discuss at a high level various approaches we take for increasing the prediction accuracy and knowledge discovery of decision forests for static and incremental datasets.

Speaker(s): Zahid Islam ,

Agenda:
Short Bio: Zahid Islam is a Professor of Computer Science at the School of Computing and Mathematics at Charles Sturt University. He is serving as the Director of the [Data Science Research Unit (DSRU)](http://bjbs.csu.edu.au/research/data-science/home), Charles Sturt University. He is also serving as a Theme Lead and the University Lead for the [Cyber Security CRC](https://cybersecuritycrc.org.au/our-research). He was privileged to get the opportunity to work with many brilliant colleagues including fantastic PhD candidates and Postdocs on various research topics including data mining, privacy issues related to data mining and applications of data mining.

Virtual: https://events.vtools.ieee.org/m/274797