An Application of Machine Learning Methods for Anomaly Detection in Internet Advertising

  • Marko Živanović School of Electrical and Computer Engineering, Belgrade
  • Svetlana Štrbac-Savić
  • Zlatogor Minchev Joint Training Simulation and Analysis Center, Institute of ICT, Institute of Mathematics and Informatics, Bulgarian Academy of Sciences


This research deals certain issues regarding downloading data from the Internet, i.e. Internet page advertising, and certain mechanisms to take care of the integrity of the data that is put into the dedicated processing context afterwards. The work also relates to e-commerce as well as to the economy in general, as some advertising scenarios provide high error-rates with pricing, which may be unacceptable in various scenarios, such as renting or selling a home. This paper presents a brief overview of the outlier detection methods and machine learning-based classifiers which are used to determine the number of anomalies in the analyzed dataset. This work contributes to the operation of the organizations which are dealing with data accuracy and integrity, such as home renting or selling agencies.