Sequential Mining Research Articles

Efficient mining of concept-hierarchy aware distinguishing sequential patterns

Distinguishing sequential patterns are sequential patterns that have much higher frequencies in one target group of sequences (concerning a given phenomenon of interest) than in a contrasting group of sequences. Distinguishing sequential patterns are useful for many machine learning tasks, as well as for the explanation and characterization of the phenomenon underlying the target group of sequences. However, previous studies on mining distinguishing sequential patterns did not consider the hierarchical relationship among elements in sequences. To fill the gap, this paper investigates the mining of distinguishing sequential patterns in the presence of concept hierarchies among sequence elements. The associated patterns will be called concept-hierarchy aware distinguishing sequential patterns (hDSPs). After presenting the challenges on mining hDSPs, we present hDSP-Miner, a method with effective pruning techniques, for mining hDSPs. Our empirical study using real-world protein sequences demonstrates that hDSP-Miner is effective and efficient, and it can discover more novel distinguishing sequential patterns than previous algorithms for mining distinguishing sequential patterns.

Sequential Pattern Mining with Multidimensional Interval Items

Sequential Pattern Mining with Multidimensional Interval Items

Expression profiling and cellular localization of myxozoan minicollagens during nematocyst formation and sporogenesis

In free-living cnidarians, minicollagens are major structural components in the biogenesis of nematocysts. Recent sequence mining and proteomic analysis demonstrate that minicollagens are also expressed by myxozoans, a group of evolutionarily ancient cnidarian endoparasites. Nonetheless, the presence and abundance of nematocyst-associated genes/proteins in nematocyst morphogenesis have never been studied in Myxozoa. Here, we report the gene expression profiles of three myxozoan minicollagens, ncol-1, ncol-3, and the recently identified noncanonical ncol-5, during the intrapiscine development of Myxidium lieberkuehni, the myxozoan parasite of the northern pike, Esox lucius. Moreover, we localized the myxozoan-specific minicollagen Ncol-5 in the developing myxosporean stages by Western blotting, immunofluorescence, and immunogold electron microscopy. We found that expression of minicollagens was spatiotemporally restricted to developing nematocysts within the myxospores during sporogenesis. Intriguingly, Ncol-5 is localized in the walls of nematocysts and predominantly in nematocyst tubules. Overall, we demonstrate that despite being significantly reduced in morphology, myxozoans retain structural components associated with nematocyst development in free-living cnidarians. Furthermore, our findings have practical implications for future functional and comparative studies as minicollagens are useful markers of the developmental phase of myxozoan parasites.

Using the Diagnostic Odds Ratio to Select Patterns to Build an Interpretable Pattern-Based Classifier in a Clinical Domain: Multivariate Sequential Pattern Mining Study.

BackgroundIt is important to exploit all available data on patients in settings such as intensive care burn units (ICBUs), where several variables are recorded over time. It is possible to take advantage of the multivariate patterns that model the evolution of patients to predict their survival. However, pattern discovery algorithms generate a large number of patterns, of which only some are relevant for classification.ObjectiveWe propose to use the diagnostic odds ratio (DOR) to select multivariate sequential patterns used in the classification in a clinical domain, rather than employing frequency properties.MethodsWe used data obtained from the ICBU at the University Hospital of Getafe, where 6 temporal variables for 465 patients were registered every day during 5 days, and to model the evolution of these clinical variables, we used multivariate sequential patterns by applying 2 different discretization methods for the continuous attributes. We compared 4 ways in which to employ the DOR for pattern selection: (1) we used it as a threshold to select patterns with a minimum DOR; (2) we selected patterns whose differential DORs are higher than a threshold with regard to their extensions; (3) we selected patterns whose DOR CIs do not overlap; and (4) we proposed the combination of threshold and nonoverlapping CIs to select the most discriminative patterns. As a baseline, we compared our proposals with Jumping Emerging Patterns, one of the most frequently used techniques for pattern selection that utilizes frequency properties.ResultsWe have compared the number and length of the patterns eventually selected, classification performance, and pattern and model interpretability. We show that discretization has a great impact on the accuracy of the classification model, but that a trade-off must be found between classification accuracy and the physicians’ capacity to interpret the patterns obtained. We have also identified that the experiments combining threshold and nonoverlapping CIs (Option 4) obtained the fewest number of patterns but also with the smallest size, thus implying the loss of an acceptable accuracy with regard to clinician interpretation. The best classification model according to the trade-off is a JRIP classifier with only 5 patterns (20 items) that was built using unsupervised correlation preserving discretization and differential DOR in a beam search for the best pattern. It achieves a specificity of 56.32% and an area under the receiver operating characteristic curve of 0.767.ConclusionsA method for the classification of patients’ survival can benefit from the use of sequential patterns, as these patterns consider knowledge about the temporal evolution of the variables in the case of ICBU. We have proved that the DOR can be used in several ways, and that it is a suitable measure to select discriminative and interpretable quality patterns.

Sequential Pattern Mining Approach for Personalized Fraudulent Transaction Detection in Online Banking

Financial institutions face challenges of fraud due to an increased number of online transactions and sophisticated fraud techniques. Although fraud detection systems have been implemented to detect fraudulent transactions in online banking, many systems just use conventional rule-based approaches. Rule-based detection systems have a difficulty in updating and managing their rules and conditions manually. Additionally, generated from the few fraud cases, the rules are general rather than specific to each user. In this paper, we propose a personalized alarm model to detect frauds in online banking transactions using sequence pattern mining on each user’s normal transaction log. We assumed that a personalized fraud detection model is more effective in responding to the rapid increase in online banking users and diversified fraud patterns. Moreover, we focused on the fact that fraudulent transactions are very different from each user’s usual transactions. Our proposed model divides each user’s log into transactions, extracts a set of sequence patterns, and uses it to determine whether a new incoming transaction is fraudulent. The incoming transaction is divided into multiple windows, and if the normal patterns are not found in the consecutive windows, an alarm is sounded. We applied the model to a real-world dataset and showed that our model outperforms the rule-based model and the Markov chain model. Although more experiments on additional datasets are needed, our personalized alarm model can be applied to real-world systems.

Online Course Recommendation Using Deep Convolutional Neural Network with Negative Sequence Mining

Massive Open Online Course (MOOC) has been criticized for low completion rates, and one of the major reasons is that it fails to offer personalized course recommendations for different users with different demands. To solve this problem, this paper proposes a personalized course recommendation model based on convolutional neural network combined with negative sequence pattern mining. The model first models the course-learning sequence as a negative sequence pattern according to the user’s course registration, degree of completion, and final grades, in which, the negative term means that students should not choose and misoperate the principle of courses. Then, it employs a convolutional neural network structure to extract the internal features of negative sequence patterns for representation learning. Finally, through the convolutional sequence-embedding model, each user is recommended with a course list that includes the user’s maximized needs in recent temporal terms and the courses that are easy to be misselected. Experiment results show that the recommended model achieves higher recommendation performance with lower course dropout rate compared to baselines, which provides a new insight for both online and offline course recommendation.

Characterizing clinical pediatric obesity subtypes using electronic health record data.

In this work, we present a study of electronic health record (EHR) data that aims to identify pediatric obesity clinical subtypes. Specifically, we examine whether certain temporal condition patterns associated with childhood obesity incidence tend to cluster together to characterize subtypes of clinically similar patients. In a previous study, the sequence mining algorithm, SPADE was implemented on EHR data from a large retrospective cohort (n = 49 594 patients) to identify common condition trajectories surrounding pediatric obesity incidence. In this study, we used Latent Class Analysis (LCA) to identify potential subtypes formed by these temporal condition patterns. The demographic characteristics of patients in each subtype are also examined. An LCA model with 8 classes was developed that identified clinically similar patient subtypes. Patients in Class 1 had a high prevalence of respiratory and sleep disorders, patients in Class 2 had high rates of inflammatory skin conditions, patients in Class 3 had a high prevalence of seizure disorders, and patients in Class 4 had a high prevalence of Asthma. Patients in Class 5 lacked a clear characteristic morbidity pattern, and patients in Classes 6, 7, and 8 had a high prevalence of gastrointestinal issues, neurodevelopmental disorders, and physical symptoms respectively. Subjects generally had high membership probability for a single class (>70%), suggesting shared clinical characterization within the individual groups. We identified patient subtypes with temporal condition patterns that are significantly more common among obese pediatric patients using a Latent Class Analysis approach. Our findings may be used to characterize the prevalence of common conditions among newly obese pediatric patients and to identify pediatric obesity subtypes. The identified subtypes align with prior knowledge on comorbidities associated with childhood obesity, including gastro-intestinal, dermatologic, developmental, and sleep disorders, as well as asthma.

Cause analysis of pit wall collapses in Kumtor Mine

In 2002 and 2006, pit walls collapsed in Kumtor Mine. The collapse of 2002 was accompanied by negative consequences. The volume of the collapsed rock mass was about 3 Mm3. The height of the collapse was 280 meters, and the width of the collapse at the base was about 270 meters. It took a significant amount of cost and time to resume the work of the open pit. In all cases of pit wall collapses, it is important to perform a cause analysis in order to reduce the risk of re-collapse in the course of further mining. The investigation analyzed data from monitoring of pit wall displacements, structural mapping, rock properties, meteorological and seismic data, groundwater level data, mining sequence prior to collapse, etc. Explanations of eyewitnesses of the collapse were heard. Invited experts from various consulting firms considered various versions of the possible causes and processes that led to the collapse, but so far the final opinion among experts about the reasons for the collapse in 2002 has not been formed. It is believed that the main reason for the collapse was the unfavorable three-dimensional orientation of the fault, which was not identified during the period of engineering and geological surveys. In support of this version, the results of numerical simulations in a 2D formulation were presented. Numerical modeling showed that a similar picture of pit wall collapse is possible in the presence of a gently dipping fault, which cuts the exposed steeply dipping fault. At the same time, the gently dipping fault did not have an outlet to the surface of the wall and extended below the bottom of the open pit. To improve the safety of mining operations, it was decided to clarify the structural-tectonic and geological model of the open pit and update the field observation program by use a laser monitoring system. Thanks to the updated monitoring system, it was possible to avoid the tragic consequences of the collapse of the pit wall in 2006. At the same time, the collapse occurred and it was necessary to analyze the causes of the collapse. The main direction of the analysis of the causes of collapse is the search for structural-tectonic faults, along which the collapse surface was formed. The search for the failure surface, taking into account the exposed surfaces of structural-tectonic faults, indicates that the failure surface is not always formed according to the classical schemes adopted for fractured rock masses. The pit wall collapses that occurred in Kumtor Mine in 2002 and 2006 indicate that the failure surface was formed by several intersecting structural-tectonic faults. The analysis of the causes of pit wall failure should be performed taking into account the three-dimensional orientation of structural faults on the basis of numerical modeling. 3D numerical modeling was carried out in FLAC 3D. The numerical simulation results for the collapse confirmed that the 2006 failure surface was formed from several structural faults. The obtained results allowed the development of recommendations for safe mining operations for 15 years.

Mining relations between personality traits and learning styles

Current teaching methods gradually use technology-assisted tools such as online assessment, online teaching, and e-books. Digital learning methods give students more flexibility in learning without the limitations of time and location. Aside from using technology, the timely adjustment of teaching methods to match students’ learning styles may achieve better learning effectiveness. However, previous studies about digital learning either lacked factors related to students’ personality traits or were biased by using scores on homework or exams to make predictions of students’ learning effectiveness. Therefore, this study analyzes the correlation between learning effectiveness and learning characteristics, including personality traits, behavioral styles, and learning strategies of students using association rule mining and sequential pattern mining. The results can be used to identify key factors that affect students’ learning styles to determine students in need and provide a teaching reference for instructors to assist students in achieving better learning results.

Research on Joint Protection Layers and Gas Prevention Technology in Outburst Coal Seams

Outburst closed distance coal seam distances are extremely difficult to mine, and are commonly involved in safety accidents in the process of mining. Based on the engineering background of the Wulan Coal Mine at the western margin of the Ordos Basin, this paper presents a theoretical analysis, engineering analogies, and numerical simulations to determine the optimal mining sequence for the Wulan Coal Mine #7, #8, #2, and #3 coal seams. Floor penetration drilling was used to eliminate outbursts in the #7 and #8 coal seams, and gas control in the #2 and #3 coal seams was achieved using ground drilling to pump and release the pressured gas. We established a comprehensive management technology system for gas emissions from the short-distance joint protection layer in the outburst coal seams of Wulan Coal Mine, where pressure relief and drainage were carried out by drilling through the floor and surface drilling pressure relief extraction. Through field tests, the pre-drainage rate of the #2 coal seam in the Wulan Coal Mine was 66.8%, and the pre-drainage rate of the #3 coal seam was 68.1%. This shows that protective layer mining of the #7 and #8 coal seams combined with surface drilling to extract pressure-relief gas can protect the #2 and #3 coal seams.

Dichotomic Pattern Mining Integrated With Constraint Reasoning for Digital Behavior Analysis.

Sequential pattern mining remains a challenging task due to the large number of redundant candidate patterns and the exponential search space. In addition, further analysis is still required to map extracted patterns to different outcomes. In this paper, we introduce a pattern mining framework that operates on semi-structured datasets and exploits the dichotomy between outcomes. Our approach takes advantage of constraint reasoning to find sequential patterns that occur frequently and exhibit desired properties. This allows the creation of novel pattern embeddings that are useful for knowledge extraction and predictive modeling. Based on dichotomic pattern mining, we present two real-world applications for customer intent prediction and intrusion detection. Overall, our approach plays an integrator role between semi-structured sequential data and machine learning models, improves the performance of the downstream task, and retains interpretability.

Efficient Uncertain Sequence Pattern Mining Based on Hadoop Platform

In the Internet of Things (IoT) era, information is collected by sensor devices, resulting in data loss or uncertain data and other consequences. We need to represent the uncertain data collected using probabilities to extract the useful information for production and application from a huge indeterminate data warehouse. The data in the database has a particular order in time or space, so the High-Utility Probability Sequential Pattern Mining (HUPSPM) has become a new investigation and analysis topic in data processing. After the progress of timestamp, many efficient algorithms for sequential mining have been developed. However, these algorithms have a limitation: they can only be executed in a stand-alone environment and are only suitable for small datasets. Therefore, introducing an advanced graph framework for processing large datasets addresses the shortcomings of the existing methods. The proposed algorithm can avoid repeated database searching, splitting the database, and improve the parallel computing capability. The initial database is pruned according to the existing pruning strategy to effectively reduce the number of candidate sets effectively. Experiments show that the algorithm presented in this paper has excellent advantages in mining high-utility probability sequences in large datasets.

Analysing the evolution of student interaction patterns in a Massive Private Online Course

ABSTRACT Recently, researchers have proposed to leverage technology-supported data (log files) to investigate temporal and sequential patterns of interaction behaviors in learning processes. There are two major challenges to be addressed: clarifying the positioning of interaction levels and identifying the evolution of the interaction action patterns in learning processes, particularly for students with differing achievements. This paper explores the use of sequential pattern mining to address the evolution of student action patterns in Massive Private Online Courses (MPOCs) and compare these patterns between different achievement groups. The study was conducted with first-year undergraduate computer science students enrolled in a computer application course at a traditional open university in one of the Chinese provinces (N = 1375). The results showed the development of various action patterns in each phase of the course and the distinct action patterns for high-achieving and low-achieving students. The findings of study provide a new perspective for instructors and students to understand interaction patterns at the fine-grained level, and can help instructional designers develop learner-cantered courses and platforms to improve online learning.

Incremental sequential patterns for multivariate temporal association rules mining

Most temporal association rule mining algorithms can mine the temporal relationship between items, but the sequential relations between successive items remain unknown, while sequential pattern mining algorithms can discover the sequential relationship between successive items but the quantitative time interval of sequential patterns remains unknown. Furthermore, they only focus on mining a historical temporal database, which ignores the fact that temporal databases are continually appended or updated. To address these problems, by integrating the advantages of these two algorithms, we extend the sequential pattern mining algorithm—PrefixSpan by introducing the concept of fuzzy temporal pattern and incremental learning, a new algorithm called Incremental sequential patterns for multivariate temporal association rules mining (ISPTAR) is proposed. First, the temporal transaction dataset obtained by fuzzy discretization of multivariate time series is converted into a temporal sequence dataset. Second, based on the PrefixSpan algorithm, the fuzzy temporal sequential patterns are mined by taking the valid time interval of the sequential pattern and the temporal relationship between items in the sequential pattern into account. Third, fuzzy temporal association rules can be constructed to mine the association between different attributes based on the mined fuzzy temporal sequential patterns. In addition, when new data are added, the proposed algorithm can update the sequential patterns without rescanning the historical database. We compare the ISPTAR algorithm with other sequential patterns mining and association rules mining algorithms on four real datasets from the UCI machine learning data repository. Experimental results demonstrate that ISPTAR, in most cases, outperformed the other algorithms in execution time and the number of generated rules. In addition, the rules generated by ISPTAR contain more temporal information, which is helpful to assist important decision-making.

Classification Based Neural Network Modelling with Continuous and Sequential data

Data mining is a potential process that involves finding useful data from an existing database. The main core work of the paper is classification with Neural Networks and Pattern mining. The study is involved in figuring out the efficient and accurate algorithm from the existing mining algorithms. The sequential mining algorithms are incorporated in areas where the data are in the form of transactions. However, in this paper, the closed and frequent sequential mining algorithms are looked at deeper with the neural network Based Multi-Layer Perceptron to pick out a productive algorithm. There are a lot of potential sequential pattern mining algorithms that can be used on different types of data. And some are good with results when they are worked with certain datasets. The frequent sequential algorithms produce the data based on the ‘minsup’ value provided. Unlike the other mining algorithms, the closed pattern mining algorithms are compact and lossless when producing results. In the frequent sequential pattern mining category, the lattice-based CM-SPADE algorithm stands out with high accuracy in terms of memory management in megabytes and execution time in milliseconds. The CM-ClaSP algorithm performs better than the ClaSP, CloFAST and BIDE+ algorithms in the closed sequential pattern mining algorithm. Now the performance of the CM-SPADE has been added with the classification based format Multi-Layer Perceptron to boost up the Perceptron Networks Classification Style with an outcome of 93% of accuracy.

Classification of Ni Levels for Determination Cut-Off Grade in Region X

The need for ore with levels above 1.8% presents the author's initiative to conduct research to produce ore products with the required levels. The aim is to find out the lowest grade of ore that can be mined and used as a cut-off grade in area x. The research method used is the blending method. Some of the data needed include the drill point coordinates, the borehole's depth, the borehole, the slope of the borehole, lithology, and the value of Ni content. The distribution of nickel in the x region is dominated by ore which has a grade of 1 to 2% with a total volume of 140,100 Bcm, and there is a small amount of nickel that has a rate above 2% with an importance of 15,650 Bcm. The mining sequence has been divided into three to three stages, where the tonnage obtained from each location is 8,9937 tons, 442361 tons, and 100,554 tons. The simulation of blending steps was carried out starting from the classification of nickel with the highest grade, namely nickel with the lowest rate of 1.8% with an average rise of 2.07%, to the last stage, which produces a middle grade of 1.9%, namely in the classification of nickel with rates above 1.6 %. From the results of the blending simulation that has been carried out, it can be concluded that the lowest grade that can be used as a cut-off grade in area x is ore with a Ni content of 1.6%.

Geochemical features of altered carbonaceous mudstones from Troyanovo-3 mine borehole (Maritsa Iztok lignite field, Bulgaria)

We sampled and studied a borehole located in the southern part of Maritsa Iztok lignite field (Bulgaria) which comprises Late Miocene-Pliocene sediments of Maritsa Formation. The current report of Troyanovo-3 Mine sequence is the first to our knowledge. It represents a lignite-mudstone succession with variations in organic matter (OM) maturity. Thirty eight coal and mudstone core samples showed predictable ash and sulfur content. The XRD study showed illite/muscovite, illite/smectite, kaolinite, chlorite and quartz as main constituents that argue for a detrital origin. Additionally, we identified pyrite and gypsum and some peaks of marcasite and elemental sulfur. A set of eight mudstone samples rich in OM was selected for further geochemical analyses. The OM was classified as type III and II/III kerogen based on Rock-Eval pyrolysis data. The low amounts of extractable organic matter (0.01–2.25%) we found are consistent with the samples' low maturity and kerogen type. The yields of two samples were too low for further fractionation. Asphaltene amount (26.1–79.7%) predominates in the extracts. Diterpenoids (neutral fraction) and their polar analogues (slightly polar fraction) are highly abundant. In slightly polar fractions we registered ketotriterpenoids (mainly lupenones) in addition to the linear n-alkan-2-ones.OM elevated maturity of the deepest sample (59.60 m depth) was evidenced mainly by hopane distributions, i.e. Ts/(Tm + Ts), homohopane H31αβ index, hopane ratios (H29αβ/H30αβ) and H27β/H27α. The organic geochemistry results do not fit the ordinary regional maturity trend. OM maturity variation, superimposed over natural low coalification degree, could be ascribed to an interaction with ascending hot reductive fluids. The sulfide mineralization found in the fractures also gives indications about the fluid circulation. Another thermal alteration affecting the middle part samples (24.0–28.3 m) was suggested due to elevated Tmax values and HI and OI relation for a charcoal piece and carbonaceous mudstone. Smoldering processes affecting these samples were presumed.Low-temperature alteration processes were identified by n-alkanes, n-alkan-2-ones, and n-alcohols expulsion and downwards migration (from 18.0 m to 30.0 m) accompanied by a clear terpenoids depletion. The 17,21-seco-hopane-17,21-dione found for the first time for Maritsa Iztok Basin (MIB) samples further suggested processes of water seepage and wash-out. Microbial degradation activities usually accompanying water-washing possibly led to a specific n-alkan-2-one distribution. All the features we report point to either an even-numbered n-alkanoic acid decarboxylation or to a selective microbial degradation of odd-numbered n-alkanes to n-alkan-2-ones. In all, the history of the coal-mudstone sequence is complex, overprinted by variable alteration processes, first ever reported for MIB samples.

Evaluation of Students' Online Learning Behavior and Perception of Psychological State Based on Data Mining.

This paper first analyzes the characteristics of students' OLB and stimulates students' online learning motivation based on the locus of control (LoC) stimulation model. Through data mining technology, this paper puts forward the analysis model of OLB, the sequence mining model of OLB, and the interactive model of OLB. Finally, a psychological state perception model based on OLB analysis is constructed; MAE and MSE are used as indexes to compare the fitting effects of GBRT, AdaBoostRegression, LinearSVR, and LinearRegression algorithms on the model. The results show that the relationship between students' psychological state and OLB is very close. While the model constructed has certain predictive ability for students' psychological state perception of OLB, which has certain practical significance for the regulation of students' learning behavior.

Seq2Pat: Sequence-to-Pattern Generation for Constraint-Based Sequential Pattern Mining

Pattern mining is an essential part of knowledge discovery and data analytics. It is a powerful paradigm, especially when combined with constraint reasoning. In this paper, we present Seq2Pat, a constraint-based sequential pattern mining tool with a high-level declarative user interface. The library finds patterns that frequently occur in large sequence databases subject to constraints. We highlight key benefits that are desirable, especially in industrial settings where scalability, explainability, rapid experimentation, reusability, and reproducibility are of great interest. We then showcase an automated feature extraction process powered by Seq2Pat to discover high-level insights and boost downstream machine learning models for customer intent prediction.

Evaluation of Atopic Dermatitis and Cutaneous Infectious Disorders Using Sequential Pattern Mining: A Nationwide Population-Based Cohort Study

According to previous studies, the increased risk of cutaneous infectious disorders in patients with atopic dermatitis (AD) is related to impaired epidermal function, abnormal systemic immune function, and lower antimicrobial peptides. In this study, we analyzed the association between AD and cutaneous infectious disorders in the real world using sequential pattern mining (SPM). We analyzed National Health Insurance data from 2010–2013 using SPM to identify comorbid cutaneous infectious diseases and the onset durations of comorbidities. Patients with AD were at greater risk for molluscum contagiosum (adjusted odds ratio (aOR), 5.273), impetigo (aOR, 2.852), chickenpox (aOR, 2.251), otitis media (aOR, 1.748), eczema herpeticum (aOR, 1.292), and viral warts (aOR, 1.105). In SPM analysis, comorbidity of 1.06% shown in molluscum contagiosum was the highest value, and the duration of 77.42 days documented for molluscum contagiosum was the shortest onset duration among all the association rules. This study suggests that AD is associated with an increased risk of cutaneous infectious disorders. In particular, care should be taken regarding its high relevance with impetigo, molluscum contagiosum, and otitis media, which may help in preventing AD from worsening through appropriately preventing and managing the condition.