www-ai.cs.tu-dortmund.de/LEHRE/FACHPROJEKT/SS12/paper/outlier/marascu2009.pdf
on real Web usage data and Section 7 gives our conclusion.
2. RELATED WORKS In this paper, we focus on clustering-based outlier detection al-
gorithms [11, 16, 9, 7, 14]. Such techniques rely on the assumption [...] -0.5, -1 2 2.25, 11.75 -1.25, -2.25 1 7 -4.75
Then, we keep only the most two significant coefficients and we make the others zero. In our series of coefficients ([7, −4, 75,
−1.25, −2.25, 0, −1.5, −0.5 [...] 5, −1]) the most two significant ones are 7 and −4, 75, meaning that the series becomes [7, −4, 75, 0,
0, 0, 0, 0, 0]. In the following step, the inverse operation is cal- culated and we obtain an approximation …