Table 1 Related work

Bing Liu [27]

In this study, the authors expand upon the current association rule model, enabling users to define multiple minimum supports to capture diverse natures and frequencies of items. More specifically, users have the capability to designate distinct minimum item support levels for each individual item.

√

Multiple min sup

----

Number of large item set

Number of Candidate item set

This model allows for the discovery of rare item rules without generating an excessive number of irrelevant rules associated with frequent items. The experimental and practical demonstrations underscore the effectiveness of this novel model.

In the past, users were only required to adjust a single minimum support (MS) threshold, but now they find themselves needing to fine-tune multiple MS thresholds.

Ya-Han Hu [35]

In this research, the emphasis is placed on the upkeep of the MIS-tree. This ensures that following any adjustments made to the item supports, the MIS-tree remains in a correct and accurate state without necessitating a rescan of the database.

√

Multiple min sup

----

Runtime

They observe that their CFP-growth algorithm outperforms the MSapriori algorithm by approximately an order of magnitude across all datasets. The development of their efficient algorithm for mining frequent patterns with multiple minimum support values is a notable achievement.

They address a limitation in the MSapriori algorithm, which necessitates a post-processing phase for the generation of association rules. Additionally, they introduce an efficient maintenance algorithm for updating the MIS-tree when the user adjusts the MIS values of items. However, the process of determining the minimum support still involves user intervention.

R. Uday [36]

In this study, the authors investigated the concept of "item-to-pattern difference" and expanded its application to the minimum constraint model. This extension allows the model to effectively prune patterns during the process of mining rare association rules.

√

Multiple min sup

----

Number of Frequent patterns

The findings indicate that, in comparison to both the single minimum support (minsup) model and the minimum constraint model, the proposed model effectively eliminates a higher number of uninteresting rules during the mining of rare association rules.

It shares the same issues as the minimum support (MS) model.

There is still parameter setting by the user

Salam [29]

They introduce an innovative approach to efficiently retrieve the top few maximal frequent patterns in order of significance, eliminating the need for the minimum support parameter.

----

Automatic minsup

----

Runtime

Their approach involves a single pass through the database and the generation of length two itemsets. The association ratio graph is introduced as a compact structure that encapsulates concise information and is constructed in time proportional to the square of the database size.

They only need to designate a parameter that is more easily understandable for humans, namely the desired number of itemsets, denoted as "k."

Kuo [37]

This research aims to introduce an innovative algorithm for association rule mining with the goal of enhancing computational efficiency and automating the identification of appropriate threshold values. The proposed method employs the particle swarm optimization algorithm, which initially seeks the optimal fitness value for each particle.

√

Automatic minsup

Borland C++ Builder 6

Runtime

Number of frequently generated patterns

This research has illustrated that the utilization of the PSO algorithm enables the rapid and objective determination of these two parameters. As a result, it enhances the mining performance for large databases, as evidenced by its application to the FoodMart2000 database.

varying product items may carry distinct levels of significance. Exploring a weighted PSO mining algorithm could offer additional practical approaches for industries.

Azzeddine Dahbi [32]

In this paper, the authors introduce two primary contributions. The initial contribution involves the automated computation of the minimum support (minsup) based on each dataset, rather than relying on a predetermined constant value set by users. The second contribution of their proposed method is the dynamic update of this minsup at each level, achieved by utilizing the means of support for all itemsets with a single item.

√

Automatic minsup

----

Number of Association Rule

Time-consuming

Quality of the Extracted Rules

The primary benefit of the proposed approach lies in the automatic determination of support at multiple levels. It yields results that encompass the desired rules with maximum interestingness in a short timeframe. The number of rules generated by the proposed algorithm is considerably lower when compared to the APRIORI Algorithm.

It employs the Mean for the automated calculation of the minimum support, a method that is susceptible to outlier data. This sensitivity may impact the algorithm's overall performance.

Azzeddine Dahbi [33]

The primary innovation in their paper lies in the automatic calculation of the minsup threshold, which is tailored to each dataset, departing from the conventional approach of using a user-predefined constant value. To accomplish this objective, the authors employ a set of statistical measures, encompassing central tendency measures like mean, mode, and median, as well as dispersion measures such as range, standard deviation, quartile 1, and quartile 3.

√

Automatic minsup

----

Number of Association Rule

Time-consuming

Quality of the Extracted Rules

They achieve results that encompass the desired rules with maximum interestingness in a short period. The proposed algorithm generates significantly fewer rules in comparison to the Apriori algorithm.

In spite of advancements in enhancing min support calculation compared to the preceding method, the impact of variable importance has been overlooked.