Brownian motion based abnormality detection in time based processes.
The research is based on a patent whose development started in 2009. The basic idea is to detect abnormality in time-based processes that have signature of Brownian motion equations. Appendix A explains the mathematical background. Application of the above principle is carried out in the following areas -
Identification of contamination in drinking water.
Identification of contamination in sewage systems.
Finding equipment malfunctions of solar photovoltaic technology.
Current methodology of abnormality detection in the literature is based on clustering or by comparing actual value visa predicted value of some equation based model.
The following gives a short description how BM can be used for abnormality detection in two domains.
Detection of abnormalities in Email messages Brownian motion may be used to detect abnormalities in mail or in text in general.
This issue is presented in an article currently under construction. Seminar about this subject was presented in the Faculty of Technology Management on 11/01/2015.
The challenge is how to perform such a task without violating privacy of the sending / receiving e-mail. The basic idea is to generate numerical measurements based on the meta-data of the email system. Such measurements include:
How many email messages were sent, in what time intervals, how many are within the organization, and how many outside the organization, and more. These measurements can be used in order to perform the algorithm of Brownian motion. When an employee decides to leave the organization, his/her behavior changes and reflected the above variables measured using Brownian motion. A summary article about the above was sent to the conference on Production Engineering dated 02/24/2015. Full article in English will be sent after the conference to a leading magazine in the field of data mining.
Current analysis is based on fixed values of the BM and by examining full text.
The next step in the research is to apply machine learning in order to apply self learning of the BM parameters as depicted in appendix A.
Detection of water contamination Detection of water contamination using BM was performed during commercial work done by me between 2008 and 2013. The idea is that online measurements can be translated into BM. Contamination generates disruption to the BM. The challenge in the future research is to classify the type of contamination based on the shape of the disruption curve. The shape of the disruption can be measured as shown in figure 1 (in the Appendix).
Most of the time (horizontal axis), the value of L (see equation 3) is below the value of X. During disruption this condition is violated. Assuming the g is the maximum value of L from equation (3) during normal behavior, during disruption this value goes up. The shape of the disruption can be denoted by the following parameters: d, S, RB and RA.
The research will try to classify different types of abnormalities as they are shown in the BM curve.
Binstock and E. Brill
Detecting employee's abnormalities behavior using mail meta data monitoring.
Proceedings of the 19th Industrial Engineering and Management Conference.
February 24, Tel Aviv, 2015
M. Green and E. Brill
Managing the grid while optimizing revenue from small PV systems
Proceedings of the 19th Sede Boqer Symposium on Solar Electricity Production
February 23-25, 2015
E. Brill, J. Sack
Water contamination detection – Methodology and empirical results
EnviroGeoChemica Acta J. (EGCA)
Special Issue, 6, 2014 (390-396)
[presented at IPN – Israel Water Week (I2W2),
Mexico City, Mexico, 17-19 September 2014]
Implementing machine learning algorithms for water quality event detection:
Theory and practice
In: Securing Water and Wastewater Systems
Series: Protecting Critical Infrastructure
Robert M. Clark, Simon Hakim (eds.)
Springer, 2, 2014 (107-122)
E. Brill, E. Hochman, D. Zilberman
Allocation and pricing at the water district level
In: "Economics of Water Resources", Volume II
R. Quentin Grafton (ed.)
Edward Elgar Publishing, 2009, Part IV, Ch.27 (385-396)
E. Brill, G. Hochman, E. Hochman
Privatization and regulation of multi-source water usage
In: "Conflict and Cooperation on Trans-Boundary Water Resources"
.E. Just, S. Netanyahu (eds.)
Kluwer Academic Publishers, 1998 (249-266)
E. Brill, U. Chakravorty, E. Hochman
Trans-boundary water allocation between Israel and the Gaza Strip:
Desalination, recycling, and fresh water
In: "Conflict and Cooperation on Trans-boundary Water Resources"
R. Just, S. Netanyahu (eds.)
Boston: Kluwer Academic Publishers, 1998 (381-400)
E. Hochman, E. Brill
Homogeneity and heterogeneity of bio-resources in economic models
In: "Modern Agriculture and the Environment"
D. Rosen, E. Tal-Or, Y. Hadar, Y. Chen (eds.)
The Hebrew University of Jerusalem
Great Britain: Kluwer Academic Publishers, 1997 (567-582)