\relax 
\citation{Wong2005,Menzies2006}
\citation{Menzies97object-orientedpatterns:}
\citation{Shortliffe1975,Weiss1977,Reboh1983}
\citation{Menzies1996}
\citation{Menzies97object-orientedpatterns:}
\citation{Menzies1996}
\@writefile{toc}{\contentsline {section}{\numberline {1}Introduction}{\thepage }}
\@writefile{toc}{\contentsline {section}{\numberline {2}background}{\thepage }}
\citation{Clancey1985}
\citation{riesbeck96}
\citation{Menzies1996}
\citation{riesbeck96}
\citation{Clancey1985}
\citation{Wielinga1992}
\citation{Menzies97object-orientedpatterns:}
\citation{Wielinga1992}
\citation{Menzies2009}
\citation{Wielinga1992}
\@writefile{toc}{\contentsline {subsection}{\numberline {2.1}Library and Toolbox For Design Patterns}{\thepage }}
\newlabel{section:library}{{2.1}{\thepage }}
\@writefile{toc}{\contentsline {section}{\numberline {3}Methodology}{\thepage }}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.1}Datasets}{\thepage }}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.2}Using Train and Test Sets for Anomaly Detection}{\thepage }}
\citation{Walter}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.3}Greedy Agglomerative Clustering Simulation Engine}{\thepage }}
\newlabel{gac-simulation-engine}{{3.3}{\thepage }}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.4}Experiments}{\thepage }}
\@writefile{toc}{\contentsline {section}{\numberline {4}Our Model}{\thepage }}
\@writefile{toc}{\contentsline {subsection}{\numberline {4.1}Building up the model}{\thepage }}
\@writefile{toc}{\contentsline {subsubsection}{\numberline {4.1.1}Select a template from the library}{\thepage }}
\@writefile{lof}{\contentsline {figure}{\numberline {1}{\ignorespaces Heuristic classification takes raw data and applies and abstraction. Then on the data abstraction heuristic match methods are run, which in our case is a likelihood-based two step procedure. Then hypothesis coming from the heuristic match is evaluated and the solution is reached.}}{\thepage }}
\newlabel{fig:heuristic}{{1}{\thepage }}
\@writefile{toc}{\contentsline {subsubsection}{\numberline {4.1.2}Pick up methods from toolbox}{\thepage }}
\@writefile{toc}{\contentsline {subsubsection}{\numberline {4.1.3}Insert domain specific heuristics}{\thepage }}
\@writefile{toc}{\contentsline {subsection}{\numberline {4.2}Executing the Model}{\thepage }}
\@writefile{lof}{\contentsline {figure}{\numberline {3}{\ignorespaces PD and PF results for 5\% cut in low likelihood values. In every treatment we have 2 different experiments: Anomaly and normality. In 9 of 10 datasets our model attains very high PD rates (1). Furthermore, for the detection of anomalies, PF rates are also very low.However, the PF rates for normal instances are usually very high. In only 2 (contact lense and cloud) out of 10 datasets we have very low pf rates. Therefore, our model is quite successful in terms of detecting anomalies, yet suffers from high PF rates in normal case identification.}}{\thepage }}
\newlabel{fig:results}{{3}{\thepage }}
\@writefile{toc}{\contentsline {section}{\numberline {5}Results}{\thepage }}
\@writefile{lof}{\contentsline {figure}{\numberline {2}{\ignorespaces Pseudocode for generic anomaly detector. Test-set can be a normal or an anomaly test set, i.e. it can consist of all abnormal test instances or all normal test instances. The anomalous instances are corrected and at the end of each era correction is controlled with Wilcoxon test to see whether it was successful or not.}}{\thepage }}
\newlabel{figure:trainPseudocode}{{2}{\thepage }}
\@writefile{lof}{\contentsline {figure}{\numberline {4}{\ignorespaces PD and PF results for 10\% cut in low likelihood values. Getting rid of more difficult instances (low likelihood valued instances) improves the anomaly detection, i.e. we have PD's of 1 and PF's of 0. However, it increases the PF rates for normal instance identification.}}{\thepage }}
\newlabel{fig:results2}{{4}{\thepage }}
\@writefile{toc}{\contentsline {section}{\numberline {6}Anomaly Correction}{\thepage }}
\bibstyle{abbrv}
\bibdata{references}
\bibcite{Clancey1985}{1}
\bibcite{Menzies1996}{2}
\bibcite{Menzies97object-orientedpatterns:}{3}
\bibcite{Menzies2009}{4}
\bibcite{Reboh1983}{5}
\bibcite{riesbeck96}{6}
\bibcite{Shortliffe1975}{7}
\bibcite{Menzies2006}{8}
\bibcite{Weiss1977}{9}
\bibcite{Wielinga1992}{10}
\bibcite{Wong2005}{11}
\@writefile{lof}{\contentsline {figure}{\numberline {5}{\ignorespaces Sample anomalies and their corrected versions. Corrected version of anomaly 1 is shown with 1' and so on. The corrected features are shown with bold face.}}{\thepage }}
\newlabel{fig:correction}{{5}{\thepage }}
\@writefile{toc}{\contentsline {section}{\numberline {7}Conclusion}{\thepage }}
\@writefile{toc}{\contentsline {section}{\numberline {8}References}{\thepage }}