Self learning pattern detection and deviation indication
(2017) In LU-CS-EX 2017-01 EDA920 20162Department of Computer Science
- Abstract
- This thesis work produces and evaluates a self learning algorithm that detects user access patterns in a Physical Access Control System (PACS). Self learning algorithms and pattern detection are both well researched areas with numerous methodologies. However, no one method can be used for all problems and this project focuses on hierarchical clustering for a particular problem where no solutions have been documented.
The work was done by analyzing the input data, iteratively implementing and evaluating the clustering algorithm. The evaluation is based on generated data due to lack of real world data. To compensate, a number of scenarios were put together and parameters were extensively tested.
The results show that depending on what... (More) - This thesis work produces and evaluates a self learning algorithm that detects user access patterns in a Physical Access Control System (PACS). Self learning algorithms and pattern detection are both well researched areas with numerous methodologies. However, no one method can be used for all problems and this project focuses on hierarchical clustering for a particular problem where no solutions have been documented.
The work was done by analyzing the input data, iteratively implementing and evaluating the clustering algorithm. The evaluation is based on generated data due to lack of real world data. To compensate, a number of scenarios were put together and parameters were extensively tested.
The results show that depending on what parameters are used the subjective accuracy of the produced hierarchical algorithm will vary greatly. The results also show that the combination of parameters affect how much data the algorithm needs before it can produce useful results.
It can be concluded that hierarchical clustering is applicable on the generated data for a PACS. The variety of parameters and attributes of the input data makes the possibilities of future work based on this project numerous. (Less) - Popular Abstract (Swedish)
- I DAGENS KONTORSBYGGNADER TAR MAN SIG OFTAST IN MED PERSONLIGA PASSERKORT OCH PIN-KOD. MED HJÄLP AV MASKININLÄRNING KAN MAN FINNA BAKOMLIGGANDE MÖNSTER OCH UPPTÄCKA OREGELBUNDNA PASSERINGAR SAMT OLOVLIGA INTRÅNG.
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/8902813
- author
- Siljeholm, Sofia LU and Rönnåker, Johan
- supervisor
- organization
- course
- EDA920 20162
- year
- 2017
- type
- H3 - Professional qualifications (4 Years - )
- subject
- keywords
- Clustering, Deviation, Machine learning, PACS, Patterns
- publication/series
- LU-CS-EX 2017-01
- report number
- LU-CS-EX 2017-01
- ISSN
- 1650-2884
- language
- English
- id
- 8902813
- date added to LUP
- 2017-02-09 17:31:34
- date last changed
- 2017-02-09 17:31:34
@misc{8902813, abstract = {{This thesis work produces and evaluates a self learning algorithm that detects user access patterns in a Physical Access Control System (PACS). Self learning algorithms and pattern detection are both well researched areas with numerous methodologies. However, no one method can be used for all problems and this project focuses on hierarchical clustering for a particular problem where no solutions have been documented. The work was done by analyzing the input data, iteratively implementing and evaluating the clustering algorithm. The evaluation is based on generated data due to lack of real world data. To compensate, a number of scenarios were put together and parameters were extensively tested. The results show that depending on what parameters are used the subjective accuracy of the produced hierarchical algorithm will vary greatly. The results also show that the combination of parameters affect how much data the algorithm needs before it can produce useful results. It can be concluded that hierarchical clustering is applicable on the generated data for a PACS. The variety of parameters and attributes of the input data makes the possibilities of future work based on this project numerous.}}, author = {{Siljeholm, Sofia and Rönnåker, Johan}}, issn = {{1650-2884}}, language = {{eng}}, note = {{Student Paper}}, series = {{LU-CS-EX 2017-01}}, title = {{Self learning pattern detection and deviation indication}}, year = {{2017}}, }