WFIM: Weighted Frequent Itemset Mining with a weight range and a minimum weight

Abstract

Researchers have proposed weighted frequent itemset mining algorithms that reflect the importance of items. The main focus of weighted frequent itemset mining concerns satisfying the downward closure property. All weighted association rule mining algorithms suggested so far have been based on the Apriori algorithm. However, pattern growth algorithms are more efficient than Apriori based algorithms. Our main approach is to push the weight constraints into the pattern growth algorithm while maintaining the downward closure property. In this paper, a weight range and a minimum weight constraint are defined and items are given different weights within the weight range. The weight and support of each item are considered separately for pruning the search space. The number of weighted frequent itemsets can be reduced by setting a weight range and a minimum weight, allowing the user to balance support and weight of itemsets. WFIM generates more concise and important weighted frequent itemsets in large databases, particularly dense databases with low minimum support, by adjusting a minimum weight and a weight range.

Handling Dynamic Weights in Weighted Frequent Pattern Mining

“…Table 1 shows an example of a retail database in which normalized weight values are assigned to items based on their prices. A normalization process is required to adjust the differences between data from various sources to create a common basis for comparison [3]- [5]. According to the normalization process, the final item weights can be determined to be within a specific weight range.…”

“…WFIM [3] is the first FP-tree-based weighted frequent pattern algorithm using two database scans over a static database. It makes use of a minimum weight and a weight range.…”

“…The WFIM [3] and WIP [5] algorithms show that the main challenge of weighted frequent pattern mining is that the weighted frequency of an itemset (or a pattern) does not have the downward closure property. Consider that item "a" has weight 0.6 and frequency 4, item "b" has weight 0.2 and frequency 5, and itemset "ab" has frequency 3.…”

“…For example, in retail applications, an expensive product may contribute a large portion of overall revenue even though it does not appear in many transactions. For this reason, WFP mining [3]- [8], [28], [29] was proposed as a way to discover more useful knowledge by considering the different weights of each item. The weight-based pattern mining approach can be applied in many areas, such as market data analysis where the prices of products are important factors, web traversal pattern mining where each web page has a different strength of impact, and biomedical data analysis where most diseases are not caused by a single gene but by a combination of genes.…”

“…Even though WFP mining can consider diverse application-specific weights during the mining process, it still cannot reflect a real world environment where the significance (weight) of an item varies with time. Most of the existing WFP [3]- [8], [28] mining algorithms are based on static databases. Zhang et al [29] proposed a strategy to maintain association rules in dynamic databases by weighting.…”