Christos Giatsidis is currently a Post-doctoral researcher in the Computer Science Laboratory at Ecole
Polytechnique in France with responsibilities regarding:
- Research on Graph Mining
- Applied machine learning
- Big Data management, Big Data and NoSQL technologies
- Teaching assistance on the topics of Machine Learning, Big Data, Data Science, SQL and NoSQL Data Bases
In 2014 he received a thesis prize for his thesis entitled "Graph Mining and Community Detection with Degeneracy"
His research interests include data/graph mining, community detection and evaluation, influence in social networks,methods and frameworks for mining Big Data, methods for embedding graph structure properties on high dimensional spaces. He has experience in both the research and industrial domain. Summarized work on the industrial domain :
- 2016 -Today
- Incorporating D-core metrics for the Aminer platform – visiting researcher at Tsinghua University. The D-core metric is based on published work on the evaluation of an individual’s contribution to the general community. Aminer is an academic platform for the evaluation of authors based on popular metrics (e.g. H-Index).
- Revising component failure for aircrafts with a new dataset. With a new data source of 278 aircrafts, over 13 years of log recordings and novel dimensions, an updated data analysis was conducted to re-evaluate the predictive potential of a richer and denser set of data.
- Data analysis over a 2 year period history of call occurrence data with interest in predicting future incoming calls. The nature of the prediction involves multiple topics and a nationwide span of call centers of AXA.
- 2014-2016
- Automatic summarization module for video/audio conferences. In-development module for a P2P tele-conference system that will detect topics in real time conversations and provide recommendations on searches related to the topics.
- Design methods to manage terabyte-sized medical historical data intended for intense analysis in multiple prediction tasks. The target goal is to have an efficient schema for storage in order to minimize the cost of feature engineering in a large scale.
- Prediction models for email services. The task at hand was to predict user reaction on sent emails with in-depth data cleaning and feature engineering from the raw source data and various external sources.
- 2012-2014
- Prediction model for component failure for a big aeronautics company. Data used span a ~1400 feature space for a 3 year history for 20 aircrafts. Intensive data preprocessing and feature selection and creation was employed.
- Obsessive behavior model learning and prediction in online gambling environment for a large operator in the French gambling industry. Predictive models learned from over 2 M user x 70 features x 2 year long history. Intensive data preprocessing and feature selection and creation approaches were employed