Unit:
Department of Informatics and TelecommunicationsΠληροφορική
Author:
KOSTAKONTI SOFIA
VASILOPOULOU THOMAIS
Supervisors info:
ΧΑΤΖΗΕΥΘΥΜΙΑΔΗΣ ΕΥΣΤΑΘΙΟΣ, ΚΑΘΗΓΗΤΗΣ, ΤΜΗΜΑ ΠΛΗΡΟΦΟΡΙΚΗΣ ΚΑΙ ΤΗΛΕΠΙΚΟΙΝΩΝΙΩΝ, ΕΘΝΙΚΟ ΚΑΙ ΚΑΠΟΔΙΣΤΡΙΑΚΟ ΠΑΝΕΠΙΣΤΗΜΙΟ ΑΘΗΝΩΝ
Original Title:
Event Correlation and Forecasting over High Dimensional Streaming Sensor Data
Translated title:
Event Correlation and Forecasting over High Dimensional Streaming Sensor Data
Summary:
As technology advances, the need to detect and predict events in real-time or near realtime intensifies. In this paper, we will discuss, what constitutes an event for every data
stream and how using different algorithms, future events may be predicted. These
predictions are feasible, due to the correlation between corresponding events and
become more frequent as the spectrum of previous events taken into account
increases. Moreover, in real-world applications, these events remain relevant as time
progresses with diminishing probability all the while, which is something that the
algorithms we developed take into account. Due to managing Big Data, a Python
implementation was considered the best approach, since both Pandas and NumPy
libraries provide ease of use and optimal run time for such problems. In order to present
as realistic results as possible, a variety of variables were differentiated so as to extract
the outcome with the best precision and recall.
Main subject category:
Technology - Computer science
Keywords:
Sensor Networks, Event Correlation, Event Forecasting, Change Detection
