Itemset Mining Algorithm Research Articles

A Proposed Frequent Itemset Discovery Algorithm Based on Item Weights and Uncertainty

Most frequent itemset mining algorithms (FIMA) discover hidden relationships from unrelated items. They find the most frequent itemsets depending only on the frequency of the item's existence in the dataset. These algorithms give all items the same importance, and neglect the differences in importance of the items. They assume the full certainty of data, but in most cases, real word data may be uncertain. As a result, the data could be incomplete and/or imprecise. These two problems are the most common challenges that face FIMA algorithms. Some new algorithms proposed some solutions to face these two issues separately. In other words, some algorithms handle item importance only, and others handle uncertainty only. Few algorithms dealt with the two issues together. In this article, the single scan for weighted itemsets over the uncertain database (SSU-Wfim) is proposed. It depends on the single scan frequent itemsets algorithm (SS_FIM), and enhances it to deal with weighted items in an uncertain database. SSU_WFIM deals with the uncertainty of data by giving each item in a transaction an additional value to indicate occurrence likelihood. It gives the items different values to define the weight of them. It uses a table called Ptable to save the items and their probability values. This table is used to generate all possible candidates itemsets. The results indicate the high performance in aspects of runtime, memory consumption and scalability of SSU-Wfim comparing with the UApriori algorithm. The proposed algorithm saves time and memory with a percentage exceeds 70% for all tested datasets.

Mining Productive Itemsets in Dynamic Databases

Discovering frequent itemsets is a data analysis task used in numerous domains. It consists of finding sets of items (itemsets) that frequently appear in a set of database records (also called transactions). Though discovering frequent itemsets is useful, it can produce a large amount of spurious patterns. As a result, the user may spend a great amount of time to analyze the itemsets found by a frequent itemset mining algorithm to find truly interesting patterns. Hence, in recent years, a key research topic has emerged which is to discover statistically significant patterns in databases. The most popular model for identifying itemsets that are statistically significant is to discover non-redundant productive itemsets. The state-of-the-art algorithm to extract this set of patterns is OPUS-Miner. A key drawback of that algorithm is that it is designed to be applied to a static database. Moreover, a second drawback of OPUS-Miner is that it discovers all patterns in a database. In other words, the user cannot search for itemsets containing some specific items. This paper addresses these issues by defining the novel problem of discovering targeted non redundant productive itemsets in dynamic databases. An algorithm named IDPI+ (Interactive Discovery of Productive Itemsets) is presented, storing transactions in a tree structure, which can then be interactively queried to identify productive and non redundant itemsets containing specific items. A structure named Query-Tree is also introduced to process many queries at the same time. Moreover, to handle dynamic databases, efficient transaction insertion and deletion algorithms are provided to update the tree. It was observed in an experimental evaluation on benchmark datasets containing various types of data that IDPI+ can handle thousands of queries per second on a desktop computer. Moreover, it was found that IPDI+ is more than an order of magnitude faster than a baseline algorithm.

Efficient Mining of Frequent Itemsets Using Only One Dynamic Prefix Tree

Frequent itemset mining is a fundamental problem in data mining area because frequent itemsets have been extensively used in reasoning, classifying, clustering, and so on. To mine frequent itemsets, previous algorithms based on a prefix tree structure have to construct many prefix trees, which is very time-consuming. In this paper, we propose a novel frequent itemset mining algorithm called DPT (Dynamic Prefix Tree) which uses only one prefix tree. We first introduce the concept of the post-conditional database of an itemset, and analyze the distribution of an itemset’s post-conditional database in a prefix tree representing a database. Subsequently, we illuminate how DPT adjusts the prefix tree to mine frequent itemsets and give three optimization techniques. An interesting advantage of DPT is that the algorithm can directly output a prefix tree representing all frequent itemsets after slight modifications. Using only one dynamic prefix tree, DPT avoids the high cost of constructing many prefix trees and thus gains significant performance improvement. Experimental results show that DPT remarkably outperforms previous algorithms with respect to running time and memory usage, and that a prefix tree representing all frequent itemsets DPT outputs can be used more efficient than a list representing them previous algorithms output.

이진 논리 연산을 활용한 하이 BHUI: 유틸리티 마이닝 기법에 대한 연구

Pattern mining finding frequent patterns are widely used to search useful information from large scale data. General pattern mining, however, have the problem of not applying the characteristics of the individual items because all items have the same importance. High utility itemset mining that solves this problem retrieves meaningful information by applying a relative weight to each item. In this paper, we propose a new high utility itemset mining algorithm based on bitwise operations. The bitwise operation predetermines the likelihood of joining two itemsets to reduce the join operation. Since the join operation is the most time consuming operation in the high utility itemset mining, the proposed algorithm significantly reduces the mining time. We analyzed the performance of the proposed algorithm with execution time and the number of generated utility itemsets and compared the performance with HUI-Miner. Our experimental results show that the proposed algorithm outperforms HUI-Miner.

Business Intelligence for Evaluating the Intangible Benefits of On-Shelf High Utility Itemset from the Temporal Transaction Database

Utility Mining is the progression of identifying High Utility Itemsets (HUI’s) from enormous transaction data. Utility mining plays a decisive role in the inspection of the data or giving actionable information to help managers, sales executives, and other commercial end-users to generate versed business decisions. In the hypermarkets, the showcase period of every item in display will vary such as new products, seasonal products, and so on. Itemsets with time period are not retrieved by existing utility mining algorithms. Hence, On-Shelf Utility Mining algorithms were proposed to discover HUI’s and a general on-shelf period of all items in temporal databases is considered. Research work aims to propose an algorithm called LOSUM (List On-Shelf Utility Mining) to retrieve on-shelf HUI’s from a temporal transaction database by reducing the data stores scan. The algorithm is enhanced by implementing a list structure to store utility information of every itemset. The candidate itemsets are generated from the list itself. This reduces the supplementary scan of a database. The LOSUM is compared with FOSHU using Chess, Accident, Kosarak, and Mushroom datasets. The experimental results illustrate that the LOSUM is efficient than the existing algorithm FOSHU (Fast On-Shelf High Utility itemset mining) algorithm.

A data structure perspective to the RDD-based Apriori algorithm on Spark

During the recent years, a number of efficient and scalable frequent itemset mining algorithms for big data analytics have been proposed by many researchers. Initially, MapReduce-based frequent itemset mining algorithms on Hadoop cluster were proposed. Although, Hadoop has been developed as a cluster computing system for handling and processing big data, but the performance of Hadoop does not meet the expectation for the iterative algorithms of data mining, due to its high I/O, and writing and then reading intermediate results in the disk. Consequently, Spark has been developed as another cluster computing infrastructure which is much faster than Hadoop due to its in-memory computation. It is highly suitable for iterative algorithms and supports batch, interactive, iterative, and stream processing of data. Many frequent itemset mining algorithms have been re-designed on the Spark, and most of them are Apriori-based. All these Spark-based Apriori algorithms use Hash Tree as the underlying data structure. This paper investigates the efficiency of various data structures for the Spark-based Apriori. Although, the data structure perspective has been investigated previously, but for MapReduce-based Apriori, and it must be re-investigated in the distributed computing environment of Spark. The considered underlying data structures are Hash Tree, Trie, and Hash Table Trie. The experimental results on the benchmark datasets show that the performance of Spark-based Apriori with Trie and Hash Table Trie are almost similar but both perform many times better than Hash Tree in the distributed computing environment of Spark.

Performance Comparison of Association Rule Algorithms with SPMF on Automotive Industry Data

By the recent developments about the information technologies, companies can store their data faster and easier with lower costs. All transactions (sales, current card, invoicing, etc.) performed in companies during the day combine at the end of the day to form big datasets. It is possible to extract valuable information through these datasets with data mining. And this has become more important for companies in terms of today's conditions where the competition in the market is high. In this study, a dataset of a company selling car maintenance and repair products in Turkey is used. Association Rules are applied on this dataset for determining the items which are bought together by the customers. These rules, which are calculated specifically for the company, can be used to redefine the sales and marketing strategies, to revise the storage areas efficiently, and to create sales campaigns suitable for the customers and regions. These algorithms are also called Frequent Itemset Mining Algorithms. The most recent 11 algorithms from these are applied to this dataset in order to compare the performances according to metrics like memory usage and execution times against varying support values and varying record numbers by using SPMF platform. Three different datasets are created by using the whole dataset like 6-months, 12-months and 22-months. According to the experiments, it can be said that executon times generally increases inversely with the support values as nearly all algorithms have higher execution time values for the lowest support value of 0.1. dEclat_bitset algorithm has the most efficient performance for 6-months and 12-months dataset. But Eclat algorithm can be said to be the most efficient algorithm for 0.7 and 0.3 support values; on the other hand dEclat_bitset is the most efficient algorithm for 0.3 and 0.1 support values on 22-months dataset.

An efficient dynamic switching algorithm for mining colossal closed itemsets from high dimensional datasets

The abundant data across a variety of domains including bioinformatics has led to the formation of dataset with high dimensionality. The conventional algorithms expend most of their time in mining a large number of small and mid-sized itemsets which does not enclose complete and valuable information for decision making. The recent research is focused on Frequent Colossal Closed Itemsets (FCCI), which plays a significant role in decision making for many applications, especially in the field of bioinformatics. The state-of-the-art algorithms in mining FCCI from datasets consisting of a large number of rows and a large number of features are computationally expensive, as they are either pure row or feature enumeration based algorithms. Moreover, the existing preprocessing techniques fail to prune the complete set of irrelevant features and irrelevant rows. The proposed work emphasizes an Effective Improvised Preprocessing (EIP) technique to prune the complete set of irrelevant features and irrelevant rows, and a novel efficient Dynamic Switching Frequent Colossal Closed Itemset Mining (DSFCCIM) algorithm. The proposed DSFCCIM algorithm efficiently switches between row and feature enumeration methods based on data characteristics during the mining process. Further, the DSFCCIM algorithm is integrated with a novel Rowset Cardinality Table, Itemset Support Table, two efficient methods to check the closeness of rowset and itemset, and two efficient pruning strategies to cut down the search space. The proposed DSFCCIM algorithm is the first dynamic switching algorithm to mine FCCI from datasets consisting of a large number of rows and a large number of features. The performance study shows the improved effectiveness of the proposed EIP technique over the existing preprocessing techniques and the improved efficiency of the proposed DSFCCIM algorithm over the existing algorithms.

An efficient approach based on selective partitioning for maximal frequent itemsets mining

We present a maximal frequent itemset (MFI) mining algorithm based on selective partitioning called SelPMiner. It makes use of a novel data format named Itemset-count tree—a compact and optimized representation in the form of partition that reduces memory requirement. It also does selective partitioning of the database, which reduces runtime to scan database. As the algorithm progressively searches for longer frequent itemsets in a depth-first manner, it creates new partitions with even smaller sizes having less dimensions and unique data instances, which results in faster support counting. SelPMiner uses a number of optimizations to prune the search space. We also prove upper bounds on the amount of memory consumed by these partitions. Experimental comparisons of the SelPMiner algorithm with popular existing fastest MFI mining algorithms on different types of datasets show significant speedup in computation time for many cases. SelPMiner works especially well when the minimum support is low and consumes less memory.

COOC: Visual Exploration of Co-Occurrence Mobility Patterns in Urban Scenarios

A co-occurrence pattern is an interesting pattern in human mobility, which has essential values in business intelligence, social activities, and urban planning. However, due to the deluge and complexity of mobile big data, as well as the complicated intrinsic features of the co-occurrence pattern, mining and analyzing the co-occurrence pattern are computationally highly expensive. Therefore, in this paper, we propose a framework to mine co-occurrence event data from mobile data and to explore the urban co-occurrence pattern visually. Our framework contains two modules: data modeling, to obtain the co-occurrence event data effectively utilizing frequent itemsets mining algorithm based on traffic GPS records, and visualization, to explore the co-occurrence pattern in urban scenarios from global, regional, statistical, and location perspectives. Our visualization system has been demonstrated using case studies with a real-world data set.

A novel multi-core algorithm for frequent itemsets mining in data streams

Data streams are modern data sources that are gaining attention as a consequence of their many practical applications (they can be found in data transmission, eCommerce, and intrusion detection system among others). Nevertheless, the efforts to obtain insights from data streams are limited due to their massive information volume and the time needed to process them. In this paper, a new approach for Frequent Itemsets Mining on data streams based on prefix trees which takes advantage of multi-core systems is proposed. This approach uses the Gearman framework as the interface for multi-core processing, and it allows to exploit their scalability efficiently. Experimental results show that the proposed method obtains the same patterns compared with similar approaches reported in the state-of-the-art and outperforms them concerning the processing time required. Also, it is proved that the proposed method is insensitive to variations in the support threshold value, and its efficiency depends on the size of the transactions and not on the size of the alphabet, which is a significant issue in other Frequent Itemsets Mining algorithms.

A Novel Integrated Approach for Companion Vehicle Discovery Based on Frequent Itemset Mining on Spark

Companion vehicle discovery received much attention from the research community. It has been widely adopted by traffic management departments in many aspects such as the involved vehicle tracking. Since there are a massive amount of traffic data that have complex and inaccurate accompanying vehicle relationships, companion vehicle discovery has become a challenge yet hot research topic. Several algorithms have been proposed to solve this issue on transactional datasets some of which are based on the frequent item mining algorithms that are used to extract knowledge from data in several real-world applications, such as market basket analysis, crime detection/prevention, and crowd mining. However, most of those algorithms mostly fail on large-scale datasets since it needs to scan the datasets iteratively for several times, which makes them unfeasible and time-consuming while dealing with big data. To this end, we proposed a novel HD-FIM algorithm to extract the companion vehicles from a massive amount of traffic data with the best execution efficiency on spark platform. It works in a hybrid approach between depth first and breadth first to handle the big data in distributed clusters. Experiment results show that the proposed algorithm, HD-FIM, outperforms the existing typical frequent itemset mining algorithms through practical vehicle set extraction calculations and it can be applied in any applicable traffic big data.

Parallel Algorithm for Frequent Itemset Mining on Intel Many-core Systems

Frequent itemset mining leads to the discovery of associations and correlations among items in large transactional databases. Apriori is a classical frequent itemset mining algorithm, which employs iterative passes over database combining with generation of candidate itemsets based on frequent itemsets found at the previous iteration, and pruning of clearly infrequent itemsets. The Dynamic Itemset Counting (DIC) algorithm is a variation of Apriori, which tries to reduce the number of passes made over a transactional database while keeping the number of itemsets counted in a pass relatively low. In this paper, we address the problem of accelerating DIC on the Intel Xeon Phi many-core system for the case when the transactional database fits in main memory. Intel Xeon Phi provides a large number of small compute cores with vector processing units. The paper presents a parallel implementation of DIC based on OpenMP technology and thread-level parallelism. We exploit the bit-based internal layout for transactions and itemsets. This technique reduces the memory space for storing the transactional database, simplifies the support count via logical bitwise operation, and allows for vectorization of such a step. Experimental evaluation on the platforms of the Intel Xeon CPU and the Intel Xeon Phi coprocessor with large synthetic and real databases showed good performance and scalability of the proposed algorithm.

Mining local and peak high utility itemsets

A major limitation of traditional High Utility Itemset Mining (HUIM) algorithms is that they do not consider that the utility of itemsets may vary over time. Thus, traditional HUIM algorithms cannot find itemsets that do not yield a high utility when considering the whole database, but still have a high utility during specific time periods. Discovering such itemsets is useful, as a product may sell exceptionally well during specific time periods but not during the rest of the year. This paper addresses this limitation of HUIM by defining the problem of mining local high utility itemsets (LHUI), and an extension to mine peak high utility itemsets (PHUI), which consists of finding the time periods where an itemset generates a utility that is much higher than usual (a peak). Algorithms named LHUI-Miner and PHUI-Miner are proposed to mine these patterns. Moreover, because the set of PHUIs can be large and some items in PHUIs don’t contribute much to their peaks, a third algorithm named NPHUI-Miner is proposed to discover a smaller set of patterns called Non-redundant Peak High Utility Itemsets (NPHUIs). Experimental results show that the proposed algorithms are efficient and can find useful patterns.

Efficiently Mining Frequent Itemsets on Massive Data

Frequent itemset mining is an important operation to return all itemsets in the transaction table, which occur as a subset of at least a specified fraction of the transactions. The existing algorithms cannot compute frequent itemsets on massive data efficiently, since they either require multiple-pass scans on the table or construct complex data structures which normally exceed the available memory on massive data. This paper proposes a novel precomputation-based frequent itemset mining (PFIM) algorithm to compute the frequent itemsets quickly on massive data. PFIM treats the transaction table as two parts: the large old table storing historical data and the relatively small new table storing newly generated data. PFIM first pre-constructs the quasi-frequent itemsets on the old table whose supports are above the lower-bound of the practical support level. Given the specified support threshold, PFIM can quickly return the required frequent itemsets on the table by utilizing the quasi-frequent itemsets. Three pruning rules are presented to reduce the size of the involved candidates. An incremental update strategy is devised to efficiently re-construct the quasi-frequent itemsets when the tables are merged. The extensive experimental results, conducted on synthetic and real-life data sets, show that PFIM has a significant advantage over the existing algorithms and runs two orders of magnitude faster than the latest algorithm.

A Low-Complexity, Low-Memory Frequent Itemset Mining Algorithm for Transactions With Sorted Items

A Low-Complexity, Low-Memory Frequent Itemset Mining Algorithm for Transactions With Sorted Items

A Data Structure Perspective to the RDD-based Apriori Algorithm on Spark

During the recent years, a number of efficient and scalable frequent itemset mining algorithms for big data analytics have been proposed by many researchers. Initially, MapReduce-based frequent itemset mining algorithms on Hadoop cluster were proposed. Although, Hadoop has been developed as a cluster computing system for handling and processing big data, but the performance of Hadoop does not meet the expectation for the iterative algorithms of data mining, due to its high I/O, and writing and then reading intermediate results in the disk. Consequently, Spark has been developed as another cluster computing infrastructure which is much faster than Hadoop due to its in-memory computation. It is highly suitable for iterative algorithms and supports batch, interactive, iterative, and stream processing of data. Many frequent itemset mining algorithms have been re-designed on the Spark, and most of them are Apriori-based. All these Spark-based Apriori algorithms use Hash Tree as the underlying data structure. This paper investigates the efficiency of various data structures for the Spark-based Apriori. Although, the data structure perspective has been investigated previously, but for MapReduce-based Apriori, and it must be re-investigated in the distributed computing environment of Spark. The considered underlying data structures are Hash Tree, Trie, and Hash Table Trie. The experimental results on the benchmark datasets show that the performance of Spark-based Apriori with Trie and Hash Table Trie are almost similar but both perform many times better than Hash Tree in the distributed computing environment of Spark.

CIBS: A biomedical text summarizer using topic-based sentence clustering.

Automatic text summarizers can reduce the time required to read lengthy text documents by extracting the most important parts. Multi-document summarizers should produce a summary that covers the main topics of multiple related input texts to diminish the extent of redundant information. In this paper, we propose a novel summarization method named Clustering and Itemset mining based Biomedical Summarizer (CIBS). The summarizer extracts biomedical concepts from the input documents and employs an itemset mining algorithm to discover main topics. Then, it applies a clustering algorithm to put the sentences into clusters such that those in the same cluster share similar topics. Selecting sentences from all the clusters, the summarizer can produce a summary that covers a wide range of topics of the input text. Using the Recall-Oriented Understudy for Gisting Evaluation (ROUGE) toolkit, we evaluate the performance of the CIBS method against four summarizers including a state-of-the-art method. The results show that the CIBS method can improve the performance of single- and multi-document biomedical text summarization. It is shown that the topic-based sentence clustering approach can be effectively used to increase the informative content of summaries, as well as to decrease the redundant information.

The concept of frequent itemset mining for text

Frequent itemset mining is one of popular data mining technique with frequent pattern or itemset as representation of data. However, most of frequent itemset mining research was conducted for structured data. In this paper, we did literature review of the frequent itemset mining algorithm that suitable for unstructured data such as text data. We reviewed several frequent itemset mining algorithm that had already used in text mining research, among others Apriori algorithm; Pattern-growth algorithm; and various algorithm for itemset mining problem such as based on representation, database changes, and richer database type. The result showed that from year to year research on text data using frequent itemset mining had increased, including the development of frequent itemset mining algorithms. Although, still rarely new algorithms were implemented in text data

TKEH: an efficient algorithm for mining top-k high utility itemsets

High utility itemsets mining is a subfield of data mining with wide applications. Although the existing high utility itemsets mining algorithms can discover all the itemsets satisfying a given minimum utility threshold, it is often difficult for users to set a proper minimum utility threshold. A smaller minimum utility threshold value may produce a huge number of itemsets, whereas a higher one may produce a few itemsets. Specification of minimum utility threshold is difficult and time-consuming. To address these issues, top-k high utility itemsets mining has been defined where k is the number of high utility itemsets to be found. In this paper, we present an efficient algorithm (named TKEH) for finding top-k high utility itemsets. TKEH utilizes transaction merging and dataset projection techniques to reduce the dataset scanning cost. These techniques reduce the dataset when larger items are explored. TKEH employs three minimum utility threshold raising strategies. We utilize two strategies to prune search space efficiently. To calculate the utility of items and upper-bounds in linear time, TKEH utilizes array-based utility technique. We carried out some extensive experiments on real datasets. The results show that TKEH outperforms the state-of-the-art algorithms. Moreover, TKEH always performs better for dense datasets.