Sequential Pattern Mining Problem Research Articles

A new tree-based approach to mine sequential patterns

Generic sequential pattern mining problem aims to mine the set of sequential patterns from a sequential database that satisfies a minimum support or occurrence threshold constraint. The main challenges that affect the efficiency of a solution lie in reducing the pattern search space, early detecting the infrequent patterns, representing the database in an efficient format, etc. Also, additional challenges get included when the problem environment transitions from static to incremental database leading to not to re-mine but efficiently tracking the effect of the incremental portion over the complete updated database. In this article, we introduce a new tree-based solution to the sequential pattern mining problem, including two sets of novel solutions for static and incremental sequential databases. We propose two new structures, SP-Tree and IncSP-Tree, and design two efficient algorithms, Tree-Miner and IncTree-Miner to mine the complete set of sequential patterns from static and incremental databases respectively. The proposed novel structures provide an efficient manner to store the complete sequential database maintaining “build-once-mine-many” property and giving scope to perform interactive mining. Additionally, we also design a new breath-first based support counting technique to efficiently identify the infrequent patterns at early stages and a new heuristic pruning strategy to reduce pattern search space. We also design a new pattern storage structure BPFSP-Tree to store the frequent patterns during successive iterations in incremental mining to reduce the number of database scans and to remove the infrequent patterns efficiently. A novel structure named Sequence Summarizer is also introduced to efficiently calculate and update the co-occurrence information of the items, especially in an incremental environment. Experimental results from various real-life and synthetic datasets demonstrate the efficiency of our work in comparison with the related state-of-the-art approaches.

Mining Top-k High Average-Utility Sequential Patterns for Resource Transformation

High-utility sequential pattern mining (HUSPM) helps researchers find all subsequences that have high utility in a quantitative sequential database. The HUSPM approach appears to be well suited for resource transformation in DIKWP graphs. However, all the extensions of a high-utility sequential pattern (HUSP) also have a high utility that increases with its length. Therefore, it is difficult to obtain diverse patterns of resources. The patterns that consist of many low-utility items can also be a HUSP. In practice, such a long pattern is difficult to analyze. In addition, the low-utility items do not always reflect the interestingness of association rules. High average-utility pattern mining is considered a solution to extract more significant patterns by considering the lengths of patterns. In this paper, we formulate the problem of top-k high average-utility sequential pattern mining (HAUSPM) and propose a novel algorithm for resource transformation. We adopt a projection mechanism to improve efficiency. We also adopt the sequence average-utility-raising strategy to increase thresholds. We design the prefix extension average utility and the reduced sequence average utility by incorporating the average utility into the utility upper bounds. The results of our comparative experiments demonstrate that the proposed algorithm can achieve sufficiently good performance.

Targeted mining of contiguous sequential patterns

In recent years, sequential pattern mining (SPM) has been applied in many domains, including recommender systems, fraud detection, and other related industries. Target-oriented SPM (TSPM) has been proposed to improve the interpretability of SPM by addressing the issue that too many irrelevant or meaningless sequential patterns are often discovered, which facilitates data analysis. Nevertheless, current TSPM research does not consider contiguity to identify sequential patterns, despite the fact that it is an important feature of sequence data. Contiguity has significant implications in applications such as bioinformatics and network intrusion detection, as it can aid the detection of periodic patterns and anomalous behaviors. Therefore, this paper proposes a novel problem of target-oriented contiguous sequential pattern mining (TCSPM) and introduces two algorithms, called TCSPM and TCSPM+. TCSPM utilizes a compact data structure called sequence chain and relies on a sequence chain pruning strategy to reduce the search space. In addition, TCSPM applies a reverse matching technique to quickly search for targeted contiguous sequences. TCSPM+ is an optimized version of TCSPM that utilizes sequence segmentation operations and a query sequence pruning strategy to further improve performance. Extensive experiments on both real and synthetic datasets demonstrate that TCSPM is an efficient algorithm for discovering targeted contiguous sequential patterns (TCSPs). TCSPM+ substantially enhances the performance of TCSPM in terms of runtime, memory usage, and scalability, outperforming TCSPM by an order of magnitude on some datasets.

Data Analytic for Healthcare Cyber Physical System

Nowadays, a large number of AI-powered healthcare cyber-physical systems (CPSs) have been used in healthcare services. In order to provide better care, AI-powered healthcare CPSs analyze the data they collect using a variety of techniques. Data analysis for artificial intelligence (AI)-driven healthcare CPS is one of these approaches. However, none of the techniques in data analysis can provide a good representation of contiguous and negative information. Therefore, we are the first to introduce the problem of contiguous negative sequential pattern mining. A novel algorithm called Contiguous Negative Sequential Pattern Miner (CNSPM) is proposed to discover and analyze contiguous negative sequential patterns (CNSPs) from the data collected by healthcare CPSs. Finally, we select some real medical and non-medical datasets to conduct numerous experiments. We further analyze the discovered patterns and show how healthcare services can use meaningful patterns for medical decision-making. The performance results on these datasets demonstrate that the proposed algorithm can discover more valuable patterns efficiently and effectively from the collected and transformed medical data.

A CP-based approach for mining sequential patterns with quantities

This paper addresses the problem of mining sequential patterns (SPM) from data represented as a set ofsequences. In this work, we are interested in sequences of items in which each item is associated with its quantity.To the best of our knowledge, existing approaches don’t allow to handle this kind of sequences under constraints.In the other hand, several proposals show the efficiency of constraint programming (CP) to solve SPM problemdealing with several kind of constraints. However, in this paper, we propose the global constraint QSPM whichis an extension of the two CP-based approaches proposed in [5] and [7]. Experiments on real-life datasets showthe efficiency of our approach allowing to specify many constraints like size, membership and regular expressionconstraints.

Discovering Constraint-based Sequential Patterns from Medical Datasets

The problem of mining sequential patterns from medical data has received a lot of attention as it aims to discover the causal relationship between different diseases or symptoms that are present in the patient’s body. Medical data contains the records pertaining to the information of the diseases or the symptoms of the patients besides the patients’ personal information. The records are ordered in accordance with the time and date of the patients’ visit in the hospital. Such data may offer precious information related to the cause and effect of a disease on the human body. Although, the date and time gives us chronological ordering of the occurrence of the diseases in the human body, it does not provide the information about the time intervals within which the successive diseases may occur. If the time gap of cause and effect is found to be too large, the concerned sequential pattern would be un-realistic. Considering, the time attributes of medical data, we try to address the above-mentioned problem on the sequential patterns. In this paper, we propose a method of extracting sequential patterns from medical dataset, with time-restrictions. The method extracts all sequences of diseases which occur within user-specified time intervals. The efficacy of our method is established with an experiment conducted on real life medical datasets.

Counting the frequency of time-constrained serial episodes in a streaming sequence

As a representative sequential pattern mining problem, counting the frequency of serial episodes from a streaming sequence has drawn continuous attention in academia due to its wide application in practice, e.g., telecommunication alarms, stock market, transaction logs, bioinformatics, etc. Although a number of serial episodes mining algorithms have been developed recently, most of them are neither stream-oriented, as they require processing the whole dataset multiple times, nor time-aware, as they fail to take into account the time constraint of serial episodes. In this paper, we propose two novel one-pass algorithms, ONCE and ONCE+, each of which can respectively compute two popular frequencies of given episodes satisfying predefined time-constraint as signals in a stream arrives one-after-another. ONCE is only used for non-overlapped frequency where the occurrences of a serial episode in sequence are not intersected. ONCE+ is designed for the distinct frequency where the occurrences of a serial episode do not share any event. Theoretical study proves that our algorithm can correctly mine the frequency of target time constraint serial episodes in a given stream. Experimental study over both real-world and synthetic datasets demonstrates that the proposed algorithm can work, with little time and space, in signal-intensive streams where millions of signals arrive within a single second. Moreover, the algorithm has been applied in a real stream processing system, where the efficacy and efficiency of this work are tested in practical applications.

Efficiently Approximating Top-$k$ Sequential Patterns in Transactional Graphs

In real-world networks, an edge (or vertex) may be associated with a transaction database, where each transaction consists of a set of items to describe the attributes or behaviors of the edge (or vertex). This type of graph is named as a transactional graph, and the problem of mining sequential patterns from transactional graphs is to discover the subsequences of transaction sequences that frequently appear in many paths of the graph. Solving this problem can facilitate broad applications ranging from social network analysis to urban computing. However, this task is not trivial due to a large number of sequences existed in the graph. In particular, each path of the graph may induce multiple sequences of transactions, and gathering the transaction sequences induced by all the paths may lead to an extremely large volume. To efficiently approximate the top-k patterns, we propose a Parallelized Sampling-based Approach For Mining Top-k Sequential Patterns, PSMSP, which involves two key techniques: (a) a parallelized unbiased sequence sampling approach based on a sequence balanced graph partition strategy, and (b) a novel PSP-Tree structure to efficiently mine the patterns based on the anti-monotonicity properties. The experimental results on synthetic and real-world datasets demonstrate that the PSMSP can successfully find the sequential patterns with superior efficiency.

The concept of sequential pattern mining for text

Sequential pattern mining is one of popular data mining technique with sequential pattern as representation of data. However, most of sequential pattern mining research was conducted for structured data. In this paper, we did literature review of the sequential pattern mining algorithm that suitable for unstructured data such as text data. We reviewed several sequential pattern mining algorithm that had already used in text mining research, among others GSP, Spade, PrefixSpan, Spam, Lapin, SM-Spam, CM-Spade, BIDE, and another various algorithm based on sequential pattern mining problem such as concise representation and how to extract more rich pattern. The result showed that that from year to year research on text data using sequential pattern mining had increased. Although, not many algorithm were developed and also still rarely new algorithms were implemented in text data.

A novel mapreduce algorithm for distributed mining of sequential patterns using co-occurrence information

Sequential Pattern Mining (SPM) problem is much studied and extended in several directions. With the tremendous growth in the size of datasets, traditional algorithms are not scalable. In order to solve the scalability issue, recently few researchers have developed distributed algorithms based on MapReduce. However, the existing MapReduce algorithms require multiple rounds of MapReduce, which increases communication and scheduling overhead. Also, they do not address the issue of handling long sequences. They generate huge number of candidate sequences that do not appear in the input database and increases the search space. This results in more number of candidate sequences for support counting. Our algorithm is a two phase MapReduce algorithm that generates the promising candidate sequences using the pruning strategies. It also reduces the search space and thus the support computation is effective. We make use of the item co-occurrence information and the proposed Sequence Index List (SIL) data structure helps in computing the support at fast. The experimental results show that the proposed algorithm has better performance over the existing MapReduce algorithms for the SPM problem.

Mining sequential patterns with itemset constraints

Mining sequential patterns is used to discover all the frequent sequences in a sequence database. However, the mining may return a huge number of patterns, while the users are only interested in a particular subset of these. In this paper, we consider the problem of mining sequential patterns with itemset constraints. In order to solve this problem, we propose a new algorithm named MSPIC-DBV, which is a pattern-growth algorithm that uses prefixes and dynamic bit vectors. This algorithm prunes the search space at the beginning and during the mining process. Moreover, it reduces the number of candidates that need to be checked. The experimental results show that the proposed algorithm outperforms the previous methods.

A Constraint Programming Approach for Web Log Mining

Mining web access patterns consists in extracting knowledge from server log files. This problem is represented as a sequential pattern mining problem (SPM) which allows to extract patterns which are sequences of accesses that occur frequently in the web log file. There are in the literature many efficient algorithms to solve SMP (e.g., GSP, SPADE, PrefixSpan, WAP-tree, LAPIN, PLWAP). Despite the effectiveness of these methods, they do not allow to express and to handle new constraints defined on patterns, new implementations are required. Recently, many approaches based on constraint programming (CP) was proposed to solve SPM in a declarative and generic way. Since no CP-based approach was applied for mining web access patterns, the authors introduce in this paper an efficient CP-based approach for solving the web log mining problem. They bring back the problem of web log mining to SPM within a CP environment which enables to handle various constraints. Experimental results on non-trivial web log mining problems show the effectiveness of the authors' CP-based mining approach.

Top-k Closed Sequential Graph Pattern Mining

Graphs have become increasingly important in modeling structures with broad applications like Chemical informatics, Bioinformatics, Web page retrieval and World Wide Web. Frequent graph pattern mining plays an important role in many data mining tasks to find interesting patterns from graph databases. Among different graph patterns, frequent substructures are the very basic patterns that can be discovered in a collection of graphs. We extended the problem of mining frequent subgraph patterns to the problem of mining sequential patterns in a graph database. In this paper, we introduce the concept of Sequential Graph-Pattern Mining and proposed two novel algorithms SFG(Sequential Frequent Graph Pattern Mining) and TCSFG(Top-k Closed Sequential Frequent Graph Pattern Mining). SFG generates all the frequent sequences from the graph database, whereas TCSFG generates top-k frequent closed sequences. We have applied these algorithms on synthetic graph database and generated top-k frequent graph sequences. Index Terms—Data mining, graph mining, frequent sequential patterns, closed sequential patterns.

An apriori-based algorithm for mining semi-order-preserving submatrix

Order-preserving submatrices OPSMs find objects that exhibit a coherent pattern with the same linear ordering in subspace. In general, this problem can be reducible to a special case of the sequential pattern mining problem, where a pattern and its supporting sequences uniquely specify an OPSM. In this paper, we extend the idea of order-preserving submatrix and define a new model semi-order-preserving submatrix or SOPSM that can be generalised to cover most existing bicluster models, and then propose a novel exact algorithm for mining all significant SOPSMs. To reduce the computational costs, we further propose a pruning technique and design an improved data structure for prefix tree to speed up the running time of the algorithm. A set of extensive experiments have been performed which demonstrate the effectiveness and efficiency of our method in mining SOPSMs.

An apriori-based algorithm for mining semi-order-preserving submatrix

Order-preserving submatrices (OPSMs) find objects that exhibit a coherent pattern with the same linear ordering in subspace. In general, this problem can be reducible to a special case of the sequential pattern mining problem, where a pattern and its supporting sequences uniquely specify an OPSM. In this paper, we extend the idea of order-preserving submatrix and define a new model semi-order-preserving submatrix (or SOPSM) that can be generalised to cover most existing bicluster models, and then propose a novel exact algorithm for mining all significant SOPSMs. To reduce the computational costs, we further propose a pruning technique and design an improved data structure for prefix tree to speed up the running time of the algorithm. A set of extensive experiments have been performed which demonstrate the effectiveness and efficiency of our method in mining SOPSMs.

Mining sequential patterns from probabilistic databases

This paper considers the problem of sequential pattern mining (SPM) in probabilistic databases. Specifically, we consider SPM in situations where there is uncertainty in associating an event with a source, model this kind of uncertainty in the probabilistic database framework and consider the problem of enumerating all sequences whose expected support is sufficiently large. We give an algorithm based on dynamic programming to compute the expected support of a sequential pattern. Next, we propose three algorithms for mining sequential patterns from probabilistic databases. The first two algorithms are based on the candidate generation framework—one each based on a breadth-first (similar to GSP) and a depth-first (similar to SPAM) exploration of the search space. The third one is based on the pattern-growth framework (similar to PrefixSpan). We propose optimizations that mitigate the effects of the expensive dynamic programming computation step. We give an empirical evaluation of the probabilistic SPM algorithms and the optimizations and demonstrate the scalability of the algorithms in terms of CPU time and the memory usage. We also demonstrate the effectiveness of the probabilistic SPM framework in extracting meaningful sequences in the presence of noise.

On the Deep Order-Preserving Submatrix Problem: A Best Effort Approach

Order-preserving submatrix (OPSM) has been widely accepted as a biologically meaningful cluster model, capturing the general tendency of gene expression across a subset of experiments. In an OPSM, the expression levels of all genes induce the same linear ordering of the experiments. The OPSM problem is to discover those statistically significant OPSMs from a given data matrix. The problem is reducible to a special case of the sequential pattern mining problem, where a pattern and its supporting sequences uniquely specify an OPSM. Unfortunately, existing methods do not scale well to massive data sets containing thousands of experiments and hundreds of thousands of genes, which are common in today's gene expression analysis. In particular, deep OPSMs, corresponding to long patterns with few supporting sequences, incur explosive computational costs in their discovery and are completely pruned off by existing methods. However, it is of particular interest of biologists to determine small groups of genes that are tightly coregulated across many experiments, and some pathways or processes may require as few as two genes to act in concert. In this paper, we study the discovery of deep OPSMs from massive data sets. We propose a novel best effort mining framework Kiwi that exploits two parameters k and w to bound the available computational resources and search a selected search space, and does what it can to find as many as possible deep OPSMs. Extensive biological and computational evaluations on real data sets demonstrate the validity and importance of the deep OPSM problem, and the efficiency and effectiveness of the Kiwi mining framework.

The Sequential Pattern Mining Algorithm MHSP Based on MH

Pattern growth approach is an important method in sequential pattern mining. Projection database based on the method is introduced in PrefixSpan, and the PrefixSpan algorithm can solve the problem of mining sequential patterns. But relative to large projection database, the performance of PrefixSpan is affected. Inspired by the prefix-divide method and MH structure, this paper proposed a new algorithm MHSP for sequential pattern mining. Based on the real datasets, experimental results show that the performance of MHSP algorithm is more than twice as fast as PrefixSpan.

Contiguous item sequential pattern mining using UpDown Tree

In this paper the problem of Contiguous Item Sequential Pattern (CISP) Mining is presented as a sequential pattern mining problem under two constraints. First, each element in a sequence consists of only one item. Second, items appearing in the sequences that contain a pattern must be adjacent with respect to the underlying order as they appear in the pattern. Even though the problem of CISP mining can be solved by using previous approaches on sequential pattern mining under a general constraint description framework, this may lead to poor performance due to the large searching space. To efficiently solve this problem, a new data structure, UpDown Tree, is proposed for CISP mining. UpDown Tree based approach can greatly improve the efficiency of CISP mining in terms of both time and memory comparing to previous approaches. An extensive experimental study has shown promising results with our approach.

SQUIRE: Sequential pattern mining with quantities

Discovering sequential patterns is an important problem for many applications. Existing algorithms find qualitative sequential patterns in the sense that only items are included in the patterns. However, for many applications, such as business and scientific applications, quantitative attributes are often recorded in the data, which are ignored by existing algorithms. Quantity information included in the mined sequential patterns can provide useful insight to the users. In this paper, we consider the problem of mining sequential patterns with quantities. We demonstrate that naive extensions to existing algorithms for sequential patterns are inefficient, as they may enumerate the search space blindly. To alleviate the situation, we propose hash filtering and quantity sampling techniques that significantly improve the performance of the naive extensions. Experimental results confirm that compared with the naive extensions, these schemes not only improve the execution time substantially but also show better scalability for sequential patterns with quantities.