算法1：基于改进Eclat的反常项集挖掘算法

输入：数据集D的垂直数据表示形式 $D_{\text{vert}}\left[P\right]$ ，提升度阈值 $\lambda$

输出：反常项集AbNI

1

procedure $\text{AbNIMiner}\left(D_{\text{vert}}\left[P\right],P\subseteq \mathcal{I},\lambda \right)$

2

$\text{AbNI}\Leftarrow \varnothing$

3

for all $m\in \mathcal{I}$ occurring in $D_{\text{vert}}\left[P\right]$ in reverse order do//depth-first search

4

$N\Leftarrow P\cup \left\{n\right\}$ //expanding the itemset N

5

if not isGenerator(N) then

6

continue

7

storeGenerator(N)

8

if $\left|N\right|>1$ and $\text{wSup}\left(N,D\right)\le \frac{\lambda }{2-\lambda }$ then

9

if a subset of Nhas been pruned then

10

continue

11

if $\text{wHlift}\left(N,D\right)\le \lambda$ then

12

$\text{AbNI}\Leftarrow \text{AbNI}\cup \left\{N\right\}$

13

If ${\lambda }^{-1}>\underset{S\subset N}{\min }\left\{\sqrt{\text{wSupCnt}\left(S,D\right)\ast \text{wSupCnt}\left(N\backslash S,D\right)}\right\}$ then

14

continue

15

if $\text{wSupCnt}\left(N,D\right)\ge \frac{1}{{\left({\text{wSupCnt}}_N^{\max }\ast \lambda \right)}^2}$ then

16

continue

17

$D_{\text{vert}}\left[N\right]\Leftarrow \varnothing$ //generating the next layer of the searching tree

18

for all $m\in \mathcal{I}$ in such that $m>n$ do

19

$C\Leftarrow \mathrm{cov}\left(\left\{n\right\},D\right)\cap \mathrm{cov}\left(\left\{m\right\},D\right)$

20

$M\Leftarrow N\cup \left\{m\right\}$

21

$\text{wSupCnt}\left(M,D\right)\Leftarrow {\omega }_C\left|C\right|$

22

if $\sqrt{\text{wSupCnt}\left(N,D\right)}-\sqrt{\text{wSupCnt}\left(M,D\right)}\ge {\lambda }^{-\frac{1}{2}}-\sqrt{{\text{wSupCnt}}_N^{\max }}$ then