Pattern Mining Approach Research Articles

Deviance mining with treatment learning and declare-based encoding of event logs

Business Process Deviance Mining is a research area that aims to characterize deviations of a business process from its expected outcomes. Techniques within this area discover which features of a set of process executions are associated to changes in process performance, providing insights on which process behavior leads to the best performance and also revealing behaviors that result in undesired process outcomes. In this sense, performance may refer to time, cost, resource dimensions or to any domain-specific performance indicator. Existing techniques for business process deviance mining are based on the extraction of patterns from event logs, using different pattern mining approaches. Up to date these extraction patterns have limited expressiveness, since they are not able to capture complex relationships that may be present in highly-flexible processes. In this work, we propose a new encoding technique for vector-based representation of process instances, and then apply Treatment Learning as a novel approach in the context of Deviance Mining to identify the characteristics of a process that mostly impact its performance. The proposed encoding technique is based on the fulfillment of Declare constraint templates, which makes it able to discover more expressive treatments. We compare our proposal with current process encoding techniques in a series of experiments with publicly available event logs from real-life processes. The results showed that treatment learning, together with our proposed Declare-based encoding, produced relevant and more expressive insights from the event logs, being a practical application for process analysis.

User behavior pattern mining and reuse across similar Android apps

Nowadays, Android apps have penetrated all aspects of our lives. Despite their popularity, understanding their behaviors is still a challenging task. Considering that many Android apps are in the same category and share similar workflows, in this paper, we propose a user behavior pattern mining and reuse approach across similar Android apps, thereby reducing the cost of understanding new apps. Particularly, for a specific new app, to figure out its typical behaviors, the behavior patterns that refer to the frequently-occurring workflows can be obtained from another similar app and transferred to this app. Moreover, to reuse the behavior patterns on this app, a semantic-based event fuzzy matching strategy and continuous workflow generation strategy are raised to generate workflows for this app. To evaluate our approach’s effectiveness and rationality, we conduct a series of experiments on 25 Android apps in five categories. Furthermore, the experimental results show that 88.3% of behavior patterns can be completely reused on similar apps, and the generated workflows cover 89.1% of the top 20% of important states.

Intra-Annual Identification of Local Deforestation Hotspots in the Philippines Using Earth Observation Products

Like many other tropical countries, the Philippines has suffered from decades of deforestation and forest degradation during and even after the logging era. Several open access Earth Observation (EO) products are increasingly being used for deforestation analysis in support of national and international initiatives and policymaking on forest conservation and management. Using a combination of annual forest loss and near-real time forest disturbance products, we provide a comprehensive analysis of the deforestation events in three forest frontiers of the Philippines. A space-time pattern mining approach was used to map quarterly deforestation hotspots at 1 km pixel size (100 hectares), where hotspots are classified according to the spatial and temporal variability of the 2000–2020 deforestation in the study area. Our results revealed that 79–81% of the hotspots overlap with primary forests and 27–29% are inside the state-declared protected areas. The intra-annual analysis of deforestation in 2020 revealed an alarming trend, where most deforestation occurred between the 1st and 2nd quarter (92–94% in hotspot forests; 87–89% in non-hotspot forests), highly overlapping within the slash-and-burn farming season. We also found “new” hotspots (2020) formed mostly from landslide scars and partly from selective logging, the latter is believed to be underestimated. Our study paves the way for rapid and regular assessment of the country’s deforestation, useful for the respective environmental institutions who convene several times a year. Moreover, our findings assert the imperative of alternative livelihoods to upland farmers, efficient forest protection activities, and even the mitigation of landslide risks.

An ontology matching approach for semantic modeling: A case study in smart cities

AbstractThis paper investigates the semantic modeling of smart cities and proposes two ontology matching frameworks, called Clustering for Ontology Matching‐based Instances (COMI) and Pattern mining for Ontology Matching‐based Instances (POMI). The goal is to discover the relevant knowledge by investigating the correlations among smart city data based on clustering and pattern mining approaches. The COMI method first groups the highly correlated ontologies of smart‐city data into similar clusters using the generic k‐means algorithm. The key idea of this method is that it clusters the instances of each ontology and then matches two ontologies by matching their clusters and the corresponding instances within the clusters. The POMI method studies the correlations among the data properties and selects the most relevant properties for the ontology matching process. To demonstrate the usefulness and accuracy of the COMI and POMI frameworks, several experiments on the DBpedia, Ontology Alignment Evaluation Initiative, and NOAA ontology databases were conducted. The results show that COMI and POMI outperform the state‐of‐the‐art ontology matching models regarding computational cost without losing the quality during the matching process. Furthermore, these results confirm the ability of COMI and POMI to deal with heterogeneous large‐scale data in smart‐city environments.

Leveraging semantic for community mining in multilayer networks

Community detection in multilayer networks aims to identify groups of well-connected nodes across multiple layers. While existing methods have been developed to deal with large graphs with few layers (typically less than 10), many real-world datasets are structured by transitive relationships that give rise to networks with thousands of extremely dense layers (e.g. co-citation networks, IMDb actor graphs, co-reference information network, social net- works). In addition, in these datasets the layers are often associated with textual summaries which provide important and hitherto unexploited information on the nature of the relation encoded by the layer. In this paper, we propose a new method which exploits the text associated with the layers in order to identify communities grouping together nodes connected through several semantically close layers. The method consists in embedding layer textual information in an Euclidean space, and to use it to group together, in the same community, nodes belonging to semantically close layers. To that end, we develop a pattern mining approach that extracts communities from numerical data. This approach, which mixes both symbolic and numeric techniques, is particularly well suited to identify communities in multilayer graphs. Indeed, we show that it obtains more diverse and better quality communities than those obtained by state-of-the-art competitors on datasets where ground truth is known. We also show that taking into account the semantic information improves the quality of the communities.

An Efficient Method for Mining Rare Association Rules: A Case Study on Air Pollution

Most pattern mining techniques almost singularly focus on identifying frequent patterns and very less attention has been paid to the generation of rare patterns. However, in several domains, recognizing less frequent but strongly related patterns have greater advantage over the former ones. Identification of compelling and meaningful rare associations among such patterns may proved to be significant for air quality management that has become an indispensable task in today’s world. The rare correlations between air pollutants and other parameters may aid in restricting the air pollution to a manageable level. To this end, efficient and competent rare pattern mining techniques are needed that can generate the complete set of rare patterns, further identifying significant rare association rules among them. Moreover, a notable issue with databases is their continuous update over time due to the addition of new records. The users requirement or behavior may change with the incremental update of databases that makes it difficult to determine a suitable support threshold for the extraction of interesting rare association rules. This paper, presents an efficient rare pattern mining technique to capture the complete set of rare patterns from a real environmental dataset. The proposed approach does not restart the entire mining process upon threshold update and generates the complete set of rare association rules in a single database scan. It can effectively perform incremental mining and also provides flexibility to the user to regulate the value of support threshold for generating the rare patterns. Significant rare association rules representing correlations between air pollutants and other environmental parameters are further extracted from the generated rare patterns to identify the substantial causes of air pollution. Performance analysis shows that the proposed method is more efficient than existing rare pattern mining approaches in providing significant directions to the domain experts for air pollution monitoring.

Prediction of In-Hospital Mortality from Administrative Data: A Sequential Pattern Mining Approach.

Study of trajectory of care is attractive for predicting medical outcome. Models based on machine learning (ML) techniques have proven their efficiency for sequence prediction modeling compared to other models. Introducing pattern mining techniques contributed to reduce model complexity. In this respect, we explored methods for medical events' prediction based on the extraction of sets of relevant event sequences of a national hospital discharge database. It is illustrated to predict the risk of in-hospital mortality in acute coronary syndrome (ACS). We mined sequential patterns from the French Hospital Discharge Database. We compared several predictive models using a text string distance to measure the similarity between patients' patterns of care. We computed combinations of similarity measurements and ML models commonly used. A Support Vector Machine model coupled with edit-based distance appeared as the most effective model. Indeed discrimination ranged from 0.71 to 0.99, together with a good overall accuracy. Thus, sequential patterns mining appear motivating for event prediction in medical settings as described here for ACS.

Data mining approaches to understanding the formation of secondary organic aerosol

This research used data mining approaches to better understand factors affecting the formation of secondary organic aerosol (SOA). Although numerous laboratory and computational studies have been completed on SOA formation, it is still challenging to determine factors that most influence SOA formation. Experimental data were based on previous work described by Offenberg et al. (2017), where volume concentrations of SOA were measured in 139 laboratory experiments involving the oxidation of single hydrocarbons under different operating conditions. Three different data mining methods were used, including nearest neighbor, decision tree, and pattern mining. Both decision tree and pattern mining approaches identified similar chemical and experimental conditions that were important to SOA formation. Among these important factors included the number of methyl groups for the SOA precursor, the number of rings for the SOA precursor, and the presence of dinitrogen pentoxide (N2O5).

A parallel approach for high utility-based frequent pattern mining in a big data environment

A parallel approach for high utility-based frequent pattern mining in a big data environment

Towards alignment strategies in human-agent interactions based on measures of lexical repetitions

Alignment of communicative behaviour is an important feature of Human–Human interaction that directly affects the collaboration and the social connection of conversational partners. With the aim of improving the communicative abilities of a virtual agent, and in particular its strategies related to (lexical) verbal alignment, this article focuses on the alignment of linguistic productions of dialogue participants in task-oriented dialogues. We propose a new framework to quantify both the lexical alignment and the self-repetition behaviours of dialogue participants from dyadic dialogue transcripts. The framework involves easily computable measures based on repetition of lexical patterns automatically extracted via a sequential pattern mining approach. These measures allow the characterisation of the nature of these processes by addressing various informative aspects such as their variety, complexity, and strength. This framework is implemented in the freely available and open-source software dialign. Using these measures, we present a contrastive study between Human–Human and Human–Agent dialogues on various corpora that reveals major differences in the lexical alignment and self-repetition behaviours. Lastly, we address the challenge of integrating lexical alignment capabilities in artificial agents. To this end, we describe guidelines and we discuss the integration of the proposed framework in a real-time dialogue system.

Understanding the Impact of COVID-19 Pandemic Outbreak on Grocery Stocking Behaviour in India: A Pattern Mining Approach

The outbreak and worldwide spread of the COVID-19 pandemic is unprecedented in recent human memory and has affected normal lifestyle of people worldwide. The lockdown restrictions imposed by countries across the globe created concerns about food availability and this affected grocery purchase and stocking behaviour of households. Researchers worldwide started analysing the nature and trend of changes in the behaviour and effect of socio-demographic factors on the change. However, limited studies have been undertaken in the Indian context. Also, how family structure influences panic buying behaviour has not been studied in detail so far. This work studies in detail whether the pandemic outbreak affected grocery purchase trends of Indian households in terms of the duration for which grocery purchase was being made, amount spent on grocery shopping per purchase and type of grocery items preferred using an online household survey of 506 families from all over India. Also, the influence of household income, family structure and grocery availability and delivery issues on grocery stocking behaviour has been studied in detail using Apriori algorithm. It was observed that panic buying is a reality, and quantity stocked and amount spent per purchase changed significantly after outbreak of the pandemic. Also, family income, size and presence of elderly and children in the household were found to affect buying patterns. Though it is difficult to predict long-term effect on behaviour change, with fear of prolonged pandemic situation and many countries resuming lockdown for a second time, this study is expected to be effective for design of supply policies of essentials in the new normal situation.

An Approach for Seasonally Periodic Frequent Pattern Mining in Retail Supermarket

Periodic-Frequent pattern mining is an empirical data mining technique that is used to investigate the customer purchase behavior pattern in the retail market. It provides the most extreme advantages to a retailer for product placements, product segmentation, store layout, selling strategies and different merchandising decisions to improve customer satisfaction as well as increase the sales. Traditionally, the single minimum support based frequent pattern mining approaches missed many lower supported items for high support threshold value, and on the other sites, the number of frequent patterns explode for low support threshold value. The retail market is highly seasonal, a number of items those are not periodic-frequent at all but those are periodic and frequent at a specific season, as results a huge number of seasonal periodic-frequent itemsets are remains out of reach. To resolve this problem, we propose an approach for extracting seasonally periodic-frequent patterns with user specified minimum support and maximum periodicity thresholds on multilevel retail supermarket dataset in this paper. The dataset has been created by collecting transaction slips in a Supermarket of Bangladesh. The results of our proposed approach extract many hidden seasonal patterns in transactional database for easy merchandising decisions and it shows the superiority over the traditional systems.

Pattern Mining Approaches Used in Social Media Data

Social media conveys a reachable platform for users to share information. The inescapable practice of social media has produced remarkable volumes of social data. Social media gathers the data in both structured-unstructured and formal-informal ways as users are not concerned with the exact grammatical structure and spelling when interacting with each other by means of various social networking websites (Twitter, Facebook, YouTube, LinkedIn, etc.). People are increasingly involved in and dependent on social media networks for data, news and opinions of other handlers on a variety of topics. The strong dependence on social media network sites contributes to enormous data generation characterized by three issues: scale, noise, and variety. Such problems also hinder social network data to be evaluated manually, resulting in the correct use of statistical analytical methods. Mining social media data can extract significant patterns that can be advantageous for consumers, users, and business. Pattern mining offers a wide variety of methods to detect valuable knowledge from huge datasets, such as patterns, trends, and rules. In this work, data was collected comprised of users’ opinions and sentiments and then processed using a significant number of pattern mining methods. The results were then further analyzed to attain meaningful information. The aim of this paper is to deliver a summary and a set of strategies for utilizing the ubiquitous pattern mining approaches, and to recognize the challenges and future research guidelines of dealing out social media data.

Mining Frequent Patterns Partially Devoid of Dissociation with Automated MMS Specification Strategy

Mining frequent patterns with single minimum support is quite unpredictable as it may miss rare itemsets depending upon the high value of minimum threshold. The phenomenon is commonly known as “rare item problem” dilemma. In this connection, frequent pattern mining under multiple minimum supports (MMS) is considered as an adequate solution of the problem. Existing approaches in this domain require user-defined minimum item support (MIS), and consider only the support in recognizing the importance of the items and ignore other factors including the inter-item relationships. The generation of huge number of itemsets including the itemsets with low associability or high dissociation is another problem in the existing MMS-based approaches. Keeping view of the above issues, this paper introduces AutoMMS-FPM, an automated MMS specification approach for frequent pattern mining with partial devoid of dissociation. The proposed approach not only takes account of support but also some other factors such as influence (inf), dissociation (d) and length importance factor (ρ) to boost up the mining process. Influence refers to the inter-item relationship and it is formulated based on the concept of degree centrality of the corresponding item network. MIS for the items are calculated automatically using MBS and RIF factors where the first one relates to the item support and the second one relates to the item influence. Experimental results on both of the synthetic and real datasets show that the proposed algorithm outperforms than existing approaches in terms of number of generated itemsets, run time, memory usages, dissociation and scalability.

Approximate high utility itemset mining in noisy environments

High utility pattern mining has been proposed to overcome the limitations of frequent pattern mining which cannot reflect the unique profits of items. High utility pattern mining has been actively conducted because it can find more valuable patterns than previous fields of pattern mining. However, its traditional approaches are designed to perform on the assumption that the data stored in databases is faultless. If there are unknown errors, such as noises, in a given database, the mining results traditional high utility pattern mining approaches mined in this database cannot be fully trusted. In this paper, a novel technique considering the noises is suggested in order to overcome this limitation. The proposed technique calculates the ranges of trustworthy utilities for patterns using a utility tolerance factor. By using this factor, the robust high utility patterns, called as approximate high utility patterns, can be extracted from a noisy database. To evaluate the performance of the proposed algorithm, various experiments are designed and conducted in terms of runtime, memory usage, and scalability. The experimental results show that the proposed algorithm outperforms than competitors, an apriori-based approach and UP-Growth.

An improved hidden behavioral pattern mining approach to enhance the performance of recommendation system in a big data environment

The proposed work aims to solve data sparsity problem in the recommendation system. It handles two-level pre-processing techniques to reduce the data size at the item level. Additional resources like items genre, tag, and time are added to learn and analyse the behaviour of the user preferences in-depth. The advantage of the proposed method is to recommend the item, based on user interest pattern and avoid recommending the outdated items. User information are grouped based on similar item genre and tag feature. This effectively handle overlapping conditions that exist on item’s genre, as it has more than one genre at initial level. Further, based on time, it analyses the user non-static interest. Overall it reduces the dimensions which is an initial way to prepare data, to analyse hidden pattern. To enhance the performance, the proposed method utilized Apache’s spark Mllib FP-Growth and association rule mining approach in a distributed environment. To reduce the computation cost of constructing tree in FP-Growth, the candidate data set is stored in matrix form. The experiments were conducted using MovieLens data set. The observed results shows that the proposed method achieves 4% increase in accuracy when compared to earlier methods.

Discovering spatial interaction patterns of near repeat crime by spatial association rules mining

Urban crime incidents always exhibit a structure of spatio-temporal dependence. Exploration of the spatio-temporal interactions of crime incidents is critical to understanding the occurrence mechanism and spatial transmission characteristics of crime occurrences, therefore facilitating the determination of policing practices. Although previous researches have repeatedly demonstrated that the crime incidents are spatially clustered, the anisotropic characteristics of spatial interaction has not been fully considered and the detailed spatial transmission of crime incidents has rarely been explored. To better understand the spatio-temporal interaction patterns of crime occurrence, this study proposes a new spatial association mining approach to discover significant spatial transmission routes and related high flow regions. First, all near repeat crime pairs are identified based on spatio-temporal proximity. Then, these links between close pairs are simplified by spatial aggregation on spatial grids. Based on that, measures of the spatio-temporal interactions are defined and a spatial association pattern mining approach is developed to discover significant spatial interaction patterns. Finally, the relationship between significant spatial transmission patterns and road network structure is analyzed. The experimental results demonstrate that our approach is able to effectively discover spatial transmission patterns from massive crime incidents data. Our results are expected to provide effective guidance for crime pattern analysis and even crime prevention.

A two-phase sequential pattern mining framework to detect stealthy P2P botnets

The botnet has been one of the most common threats to network security. Among all emerged botnet, Peer to Peer (P2P) botnets are more perilous and resistant due to their distributed nature. In addition to their resiliency against takedown strategies, modern P2P botnets are stealthier in the way they perform fraudulent activities. One of the main challenges to detect P2P bots/botnets is the presence of benign P2P traffic. The botnet traffic can blend in with legitimate P2P traffic, and it makes the P2P bots stealthier. However, the problem of detecting P2P botnets in the presence of legitimate P2P traffic has received little attention from the research community.In this paper, a novel P2P botnet detection framework resilient to the presence of legitimate P2P traffic is proposed based on a two-phase Sequential Pattern Mining (SPM) approach. The proposed framework is evaluated in many different cases of the coexistence of P2P malicious and legitimate traffics, using real-world network traffic. Our experimental results show that the proposed framework is capable of detecting P2P bots in the presence of legitimate P2P traffic with a detection rate of 99.2%. Besides its accurate detection, our proposed framework is highly scalable and can detect even one bot in the network or different bots from different bot families.

Cascading Failure Pattern Identification in Power Systems Based on Sequential Pattern Mining

Cascading failure simulation data contain many fault chains (FCs), which can present cascading failure propagation paths. In cascading failure process, some component outages play dominating roles in propagation, indicating the common characteristics among different FCs. The commonness is embodied as combinations of components which are vulnerable to trip and cause serious blackout consequences. In addition, a component outage will increase the outage probability of relevant components and induce dependent outage in subsequent stage. Such relevance between two components can be called component outage causality. A combination of sequential component outages with outage causalities, which exists in different FCs and leads to system load loss, can be regarded as a cascading failure pattern (CFP). Statistical characteristics of FCs indicate that CFPs are variously distributed in different FCs, can present propagation paths and cause different impacts on system blackouts. This paper proposes a cascading failure pattern (CFP) identification method based on sequential pattern mining approach. The proposed method focuses on mining CFPs from massive FCs, quantifies the influence of CFP on system blackout and identifies the critical ones. The proposed method is verified with FCs data on IEEE 39-bus and 118-bus test systems.

Efficient list based mining of high average utility patterns with maximum average pruning strategies

High average utility pattern mining is the concept proposed to complement drawbacks of high utility pattern mining by considering lengths of patterns along with the utilities of the patterns. High average utility pattern mining should be able to gratify the anti-monotone property like other pattern mining techniques. Many high average utility pattern mining studies to satisfy the anti-monotone property have been proposed in order to improve various upper-bounds because the performance of pattern mining can be improved efficiently by satisfying the anti-monotone property. Although those upper-bounds can effectively reduce the search space, they still take a lot of cost to calculate all unpromising patterns or cannot find them in advance. Therefore, in this paper, a novel high average utility pattern mining approach is proposed by employing two novel upper-bounds called tight maximum average utility upper-bound and maximum remaining average utility upper-bound. Moreover, a newly suggested list-based structure, TA-List, is designed to adopt two pruning strategies. The proposed technique can efficiently extract high average utility patterns by reducing search space. To evaluate the performance of the proposed method, various experiments using real and synthetic datasets are conducted in terms of runtime, memory usage and scalability and the proposed approach is compared with the state-of-the-art high average utility pattern mining algorithms. The results of experiments show that the suggested algorithm has better performance with regard to runtime, memory usage and scalability.