### Table 3.6: Numbers of rules and computation times for various datasets

### Table 6. Examples of objective (interestingness) measures for the itemset {i1,i2}.

"... In PAGE 9: ....2.1. Examples of Association Measures We begin by describing association measures, which are called interestingness or objective measures, that have been developed to address the issue that as- sociation measures with properties different from that of traditional confidence can be useful. A list of many such measures is given in Table6 . Note that these measures are given in terms of two binary variables i1 and i2 and the frequencies with which their 0 and 1 values co-occur, as defined in Table 7.... In PAGE 9: ...t al. 2004, and Tan et al. 2005. More specifically, the measures in Table6 can be characterized in a number of different ways. For example, there are several, including confidence, that have an interpretation in terms of conditional probability or related notions involving statistical independence or information theory.... In PAGE 9: ... Mutual information is an information theoretic measure that quantifies the de- gree of information that the first itemset (the antecedent) provides about second itemset (the consequent). As an example of measures of a different sort, some of the other measures in Table6 correspond to similarity measures, namely, the cosine measure, correlation, and the Jaccard coefficient. Indeed, when the pattern evaluation vectors vX and vY are continuous, it seems especially useful to consider confidence measures based on some of the simi- larity or distance measures that have been developed for evaluating the strength of a connection between two continuous vectors.... In PAGE 9: ... Indeed, when the pattern evaluation vectors vX and vY are continuous, it seems especially useful to consider confidence measures based on some of the simi- larity or distance measures that have been developed for evaluating the strength of a connection between two continuous vectors. Some of these measures are the continuous analogues of similarity measures in Table6 , such as the cosine, correlation, and the extended Jaccard (Strehl et al. 2000) measures.... In PAGE 10: ... Although some binary measures, including those of Table 6, have natural extensions to the case where X and Y contain multiple items, many do not. Second, using the framework, the measures of Table6 can be automat- ically applied for non-traditional patterns, such as ETIs. As we show below, a generalized version of confidence for ETIs seems quite useful, and thus, gener- alized versions of other association measures for ETIs may also prove valuable.... ..."

### Table 5. 2-itemsets

2004

"... In PAGE 9: ...et to 0.5, and the minimum interest to 0.07. In Table5 and Table 6, the first column presents the results when our ap- proach was used. The second column uses the algorithm from[4], while in the third one the results are obtained using the approach in [5].... In PAGE 10: ... The itemsets BCD and ABC are not discovered by SRM because none of its subsets of two items are generated as concrete during the process. From the itemsets that were shown in Table5 and Table 6 a set of association rules can be generated. Here we show, some of the rules that were generated from the itemsets that were discovered by one algorithm, but not by others.... ..."

Cited by 8

### Table 2: Itemsets and their support in T .

"... In PAGE 2: ... An item set X T is frequent if its support is greater than or equal to a user- de ned threshold . Table2 shows all frequent item-sets in... ..."

### Table 6. 3-itemsets

2004

"... In PAGE 9: ...et to 0.5, and the minimum interest to 0.07. In Table 5 and Table6 , the first column presents the results when our ap- proach was used. The second column uses the algorithm from[4], while in the third one the results are obtained using the approach in [5].... In PAGE 10: ...5 it is discovered by our algorithm, but not by SRM. In Table6 there are differences for both, the positive and the negative ones. The algorithm that uses the minimum interest parameter discovers only the ABC itemset because it is the only one that has all the pairs X,Y of the item ABC where ABC = X[Y above the parameter.... In PAGE 10: ... The itemsets BCD and ABC are not discovered by SRM because none of its subsets of two items are generated as concrete during the process. From the itemsets that were shown in Table 5 and Table6 a set of association rules can be generated. Here we show, some of the rules that were generated from the itemsets that were discovered by one algorithm, but not by others.... ..."

Cited by 8

### Table 4: Frequent Itemsets

2006

"... In PAGE 7: ... We continue this process until no more frequent patterns are found. Table4 presents a por- tion of the frequent patterns of length 1 and length 2 for the... ..."

Cited by 3

### Table 4: Frequent Itemsets

2006

"... In PAGE 7: ... We continue this process until no more frequent patterns are found. Table4 presents a portion of the frequent patterns of length 1 and length 2 for the transformed path database of Table 3. Pruning of infrequent candidates.... ..."

Cited by 3

### Table 3. Frequent itemsets of experiment 1 1-Itemset 2-Itemset 3-Itemset 2659 76 1

"... In PAGE 2: ... DBLP collection has 328858 bibliography of different publications (article, book, PhD thesis, etc) expressed in XML format. We use rule template introduced in [Figure 3] and input parameters in [Table 2], [Figure 5] and [ Table3 ] display some of results: Table 2. Experiment 1 Parameters ... ..."

### Table 1 Number of frequent closed itemsets and frequent itemsets for the MUSHROOMS example Minsupp (%) # Frequent closed itemsets # Frequent itemsets

"... In PAGE 8: ... The MUSHROOMS example shows that iceberg concept lattices are suitable especially for strongly correlated data. In Table1 , the size of the iceberg lattice (i.e.... ..."