Sascha Krstanovic

Data Science Professional · Stuttgart, Germany · mail@saschakrstanovic.com


I'm a Sr. Data Scientist at Daimler AG's Finance department and an external researcher at the Data and Web Science Group of the University of Mannheim.
My topics of interest lie in the fields of artificial intelligence and data science. My research deals with time series forecasting, ensemble learning, and the combination of the two.

Experience

Sr. Data Scientist

Daimler AG

Working towards the AI-powered transformation of controlling and reporting in the Daimler Trucks unit. Besides heavy process automation, we focus on estimating the future development of relevant KPIs, collaborating closely with the respective business departments in an international environment. Tasks and tools include:

  • machine learning: time series models, artificial neural networks, classification, visualization
  • information retrieval and web mining
  • Python, R, RShiny, SPSS, Tableau
  • project management

April 2017 - Present

Consultant Big Data & Advanced Analytics

PricewaterhouseCoopers AG

Tasks and projects (excerpt):

  • customer churn modeling
  • product portfolio optimization via customized clustering techniques & association analysis
  • fraud detection in accounting data
  • project management

August 2015 - March 2017

Master Thesis, Time Series Analysis

Robert Bosch GmbH, R&D

Title: "Dealing with Missing Values in Time Series"

Writing my master thesis about dealing with missing values in time series data with a focus on multivariate sensor data that was collected at irregularly spaced time intervals. Besides the massive amount of data, a major challenge is finding appropriate ways of imputing missing data gaps in the series.

January 2015 - June 2015

Intern, Information Risk Management

KPMG AG

  • Working with various BI applications and reporting systems
  • Basic data analytics and software engineering

August 2014 - November 2014

Education

University of Mannheim

Master of Science, Business Informatics

Specialization track: Data & Web Science with a focus on Data Mining, Machine Learning, Information Retrieval/Extraction and Business Intelligence.

Master Thesis: "Dealing with Missing Values in Time Series" in cooperation with the Robert Bosch GmbH.

September 2013 - July 2015

University of Mannheim

Bachelor of Science, Business Mathematics

Specialization track: Economics (focus on statistics, econometrics, macroeconomics).

The bachelor thesis reviewed methods and applications of data compression techniques, such as used in the ZIP file format.

September 2010 - August 2013

Research

Ensemble Learning for Time Series Forecasting

Currently investigating state of the art forecasting models like recurrent neural networks (more precisely, LSTMs). Figuring out how to design ensembles of such models in order to harness the "wisdom of the crowds". The objective is to design an ensemble framework that is robust to the underlying data generation process. Major challenges are:

  • Designing base learners s.t. the models are sufficiently diverse, e.g. via parameter variation, batch configuration, and different time horizons
  • Combining individual model forecasts effectively, especially under consideration of multi step ahead forecasting
  • Ensuring robustness of the approaches across time series data of different domains (sensor data, finance, natural sciences, ...)


References

Sascha Krstanovic and Heiko Paulheim Ensembles of recurrent neural networks for robust time series forecasting. In: Lecture notes in computer science. Artificial Intelligence XXXIV : 37th SGAI International Conference on Artificial Intelligence, AI 2017, Cambridge, UK, December 12-14, 2017, proceedings; 34-46. Springer, Cham, 2017.

Contact

Please drop a mail to mail*at*saschakrstanovic.com for any inquiries.



Imprint

Sascha Krstanovic
Klugestr. 10
70197 Stuttgart