Data Stream Mining Techniques Research Articles

A COMPREHENSIVE REVIEW ON DATA STREAM MINING TECHNIQUES FOR DATA CLASSIFICATION; AND FUTURE TRENDS

Data Mining is a developing interdisciplinary control managing Data Reclamation and Data Stream Mining techniques, whose subject is gathering, overseeing, processing, breaking down, and visualizing the huge volume of organized or unstructured data. Data stream mining indicates how to look at Unknown patterns from a massive amount of data over algorithms. It has experienced quick improvement with significant progress in math, statistics, data science, and computer science domains. Data streams are commonly generated by various sources such as sensor networks, social media feeds, financial transactions, online retail, network traffic, and many other applications. The gathered data could be additionally utilized for various purposes, for example, execution assessment, irregularity discovery, change identification, or issue finding of the operating systems. This data stream analysis is done using different data stream mining techniques. This paper provides a broad overview of the distinct approaches used for data stream mining. Initially, we studied the different techniques of data stream mining. Next, we discuss the different clustering and classification techniques and their benefits. Then we examine the evaluation of different data stream mining techniques results that some techniques are feasible for real-time data streams and some of not. This study provides a complete understanding of techniques and their benefits. The studies done so far need to be sufficiently exhaustive for data mining techniques, so future work is needed to assess which technique is feasible for real-time data streams.

Similarity-Based Adaptive Window for Improving Classification of Epileptic Seizures with Imbalance EEG Data Stream.

Data stream mining techniques have recently received increasing research interest, especially in medical data classification. An unbalanced representation of the classification's targets in these data is a common challenge because classification techniques are biased toward the major class. Many methods have attempted to address this problem but have been exaggeratedly biased toward the minor class. In this work, we propose a method for balancing the presence of the minor class within the current window of the data stream while preserving the data's original majority as much as possible. The proposed method utilized similarity analysis for selecting specific instances from the previous window. This group of minor-class was then added to the current window's instances. Implementing the proposed method using the Siena dataset showed promising results compared to the Skew ensemble method and some other research methods.

A Frequent Pattern Conjunction Heuristic for Rule Generation in Data Streams

This paper introduces a new and expressive algorithm for inducing descriptive rule-sets from streaming data in real-time in order to describe frequent patterns explicitly encoded in the stream. Data Stream Mining (DSM) is concerned with the automatic analysis of data streams in real-time. Rapid flows of data challenge the state-of-the art processing and communication infrastructure, hence the motivation for research and innovation into real-time algorithms that analyse data streams on-the-fly and can automatically adapt to concept drifts. To date, DSM techniques have largely focused on predictive data mining applications that aim to forecast the value of a particular target feature of unseen data instances, answering questions such as whether a credit card transaction is fraudulent or not. A real-time, expressive and descriptive Data Mining technique for streaming data has not been previously established as part of the DSM toolkit. This has motivated the work reported in this paper, which has resulted in developing and validating a Generalised Rule Induction (GRI) tool, thus producing expressive rules as explanations that can be easily understood by human analysts. The expressiveness of decision models in data streams serves the objectives of transparency, underpinning the vision of ‘explainable AI’ and yet is an area of research that has attracted less attention despite being of high practical importance. The algorithm introduced and described in this paper is termed Fast Generalised Rule Induction (FGRI). FGRI is able to induce descriptive rules incrementally for raw data from both categorical and numerical features. FGRI is able to adapt rule-sets to changes of the pattern encoded in the data stream (concept drift) on the fly as new data arrives and can thus be applied continuously in real-time. The paper also provides a theoretical, qualitative and empirical evaluation of FGRI.

Mapping the Big Data Landscape: Technologies, Platforms and Paradigms for Real-Time Analytics of Data Streams

The `Big Data' of yesterday is the `data' of today. As technology progresses, new challenges arise and new solutions are developed. Due to the emergence of Internet of Things applications within the last decade, the field of Data Mining has been faced with the challenge of processing and analysing data streams in real-time, and under high data throughput conditions. This is often referred to as the Velocity aspect of Big Data. Whereas there are numerous reviews on Data Stream Mining techniques and applications, there is very little work surveying Data Stream processing paradigms and associated technologies, from data collection through to pre-processing and feature processing, from the perspective of the user, not that of the service provider. In this article, we evaluate a particular type of solution, which focuses on streaming data, and processing pipelines that permit online analysis of data streams that cannot be stored as-is on the computing platform. We review foundational computational concepts such as distributed computation, fault-tolerant computing, and computational paradigms/architectures. We then review the available technological solutions, and applications that pertain to data stream mining as case studies of these theoretical concepts. We conclude with a discussion of the field of data stream processing/analytics, future directions and research challenges.

Impact of injection attacks on sensor-based continuous authentication for smartphones

Given the relevance of smartphones for accessing personalized services in smart cities, Continuous Authentication (CA) mechanisms are attracting attention to avoid impersonation attacks. Some of them leverage Data Stream Mining (DSM) techniques applied over sensorial information. Injection attacks can undermine the effectiveness of DSM-based CA by fabricating artificial sensorial readings.The goal of this paper is to study the impact of injection attacks in terms of accuracy and immediacy to illustrate the time the adversary remains unnoticed. Two well-known DSM techniques (K-Nearest Neighbours and Hoeffding Adaptive Trees) and three data sources (location, gyroscope and accelerometer) are considered due to their widespread usage Results show that even if the attacker does not previously know anything about the victim, a significant attack surface arises – 1.35 min are needed, in the best case, to detect the attack on gyroscope and accelerometer and 7.27 min on location data. Moreover, we show that the type of sensor at stake and configuration settings may have a dramatic effect on countering this threat.

Lessons learned from data stream classification applied to credit scoring

The financial credibility of a person is a factor used to determine whether a loan should be approved or not, and this is quantified by a ‘credit score,’ which is calculated using a variety of factors, including past performance on debt obligations, profiling, amongst others. Machine learning has been widely applied to automate the development of effective credit scoring models over the years. Yet, studies show that the development of robust credit scoring models may take longer than a year, and thus, if the behavior of customers changes over time, the model will be outdated even before its deployment. In this paper, we made 3 anonymized real-world credit scoring datasets available alongside the results obtained. In each of these datasets, we verify whether the credit scoring task should be thought as an ephemeral scenario since many of the variables may drift over time, and thus, data stream mining techniques should be used since they were tailored for incremental learning and to detect and adapt to changes in the data distribution. Therefore, we compare both traditional batch machine learning algorithms with data stream algorithms in different validation schemes using both Kolmogorov–Smirnov and Population Stability Index metrics. Furthermore, we also provide insights on the importance of features according to their Information Value, Mean Decrease Impurity, and Mean Positional Gain metrics, such that the last depicts changes in the importance of features over time. For 2 of the 3 tested datasets, the results obtained by data stream learners are comparable to predictive models currently in use, thus showing the efficiency of data stream classification for the credit scoring task.

Responsible processing of crowdsourced tourism data

Online tourism crowdsourcing platforms, such as AirBnB, Expedia or TripAdvisor, rely on the continuous data sharing by tourists and businesses to provide free or paid value-added services. When adequately processed, these data streams can be used to explain and support businesses in the early identification of trends as well as prospective tourists in obtaining tailored recommendations, increasing the confidence in the platform and empowering further end-users. However, existing platforms still do not embrace the desired accountability, responsibility and transparency (ART) design principles, underlying to the concept of sustainable tourism. The objective of this work is to study this problem, identify the most promising techniques which follow these principles and design a novel ART-compliant processing pipeline. To this end, this work surveys: (i) real-time data stream mining techniques for recommendation and trend identification; (ii) trust and reputation (T&R) modelling of data contributors; (iii) chained-based storage of trust models as smart contracts for traceability and authenticity; and (iv) trust- and reputation-based explanations for a transparent and satisfying user experience. The proposed pipeline redesign has implications both to digital and to sustainable tourism since it advances the current processing of tourism crowdsourcing platforms and impacts on the three pillars of sustainable tourism.

FogMed: a Fog-based Framework for Disease Prognosis Based Medical Sensor Data Streams

Recently, an increasing number of works start investigating the combination of fog computing and electronic health (ehealth) applications. However, there are still numerous unresolved issues worth to be explored. For instance, there is a lack of investigation on the disease prediction in fog environment and only limited studies show, how the Quality of Service (QoS) levels of fog services and the data stream mining techniques influence each other to improve the disease prediction performance (e.g., accuracy and time efficiency). To address these issues, we propose a fog-based framework for disease prediction based on Medical sensor data streams, named FogMed. This framework aims to improve the disease prediction accuracy by achieving two objectives: QoS guarantee of fog services and anomaly prediction of Medical data streams. We build a virtual FogMed environment and conduct comprehensive experiments on the public ECG dataset to validate the performance of FogMed. The experiment results show that it performs better than the cloud computing model for processing tasks with different complexities in terms of time efficiency.

Data stream mining techniques: a review

A plethora of infinite data is generated from the Internet and other information sources. Analyzing this massive data in real-time and extracting valuable knowledge using different mining applications platforms have been an area for research and industry as well. However, data stream mining has different challenges making it different from traditional data mining. Recently, many studies have addressed the concerns on massive data mining problems and proposed several techniques that produce impressive results. In this paper, we review real time clustering and classification mining techniques for data stream. We analyze the characteristics of data stream mining and discuss the challenges and research issues of data steam mining. Finally, we present some of the platforms for data stream mining.

Secure and Usable User-in-a-Context Continuous Authentication in Smartphones Leveraging Non-Assisted Sensors.

Smartphones are equipped with a set of sensors that describe the environment (e.g., GPS, noise, etc.) and their current status and usage (e.g., battery consumption, accelerometer readings, etc.). Several works have already addressed how to leverage such data for user-in-a-context continuous authentication, i.e., determining if the porting user is the authorized one and resides in his regular physical environment. This can be useful for an early reaction against robbery or impersonation. However, most previous works depend on assisted sensors, i.e., they rely upon immutable elements (e.g., cell towers, satellites, magnetism), thus being ineffective in their absence. Moreover, they focus on accuracy aspects, neglecting usability ones. For this purpose, in this paper, we explore the use of four non-assisted sensors, namely battery, transmitted data, ambient light and noise. Our approach leverages data stream mining techniques and offers a tunable security-usability trade-off. We assess the accuracy, immediacy, usability and readiness of the proposal. Results on 50 users over 24 months show that battery readings alone achieve 97.05% of accuracy and 81.35% for audio, light and battery all together. Moreover, when usability is at stake, robbery is detected in 100 s for the case of battery and in 250 s when audio, light and battery are applied. Remarkably, these figures are obtained with moderate training and storage needs, thus making the approach suitable for current devices.

Dynamic adaptation of online ensembles for drifting data streams

The success of data stream mining techniques has allowed decision makers to analyze their data in multiple domains, ranging from monitoring network intrusion to financial markets analysis and online sales transactions exploration. Specifically, online ensembles that construct accurate models against drifting data streams have been developed. Recently, there has been a surge in interest in mobile (or so-called pocket) data stream mining, aiming to construct near real-time models for data stream mining applications that run on mobile devices. In such a setting, it follows that the computational resources are limited and that there is a need to adapt analytics to map the resource usage requirements. Consequently, the resultant models should not only be highly accurate, but they should also adapt swiftly to changes. In addition, the data mining techniques should be fast, scalable, and efficient in terms of resource allocation. It then becomes important to consider Return on Investment (ROI) issues such as storage requirements and memory utilization. This paper introduces the Adaptive Ensemble Size (AES) algorithm, an extension of the Online Bagging method, to address these issues. Our AES method dynamically adapts the sizes of ensembles, based on ROI usage patterns. We illustrate our approach by analyzing the performances against both synthetic and real-world data streams. The results, when comparing our AES algorithm with the state-of-the-art, indicate that we are able to obtain a high Return on Investment (ROI) and to swiftly adapt to change, without compromising on the predictive accuracy.

Online network traffic classification with incremental learning

Conventional network traffic detection methods based on data mining could not efficiently handle high throughput traffic with concept drift. Data stream mining techniques are able to classify evolving data streams although most techniques require completely labeled data. This paper proposes an improved data stream mining algorithm for online network traffic classification that is able to incrementally learn from both labeled and unlabeled flows. The algorithm uses the concept of incremental k-means and self-training semi-supervised method to continuously update the classification model after receiving new flow instances. The experimental results show that the proposed algorithm is able to classify 325 thousands flow instances per second and achieves up to 91–94 % average accuracy, even when using 10 % of labeled input flows. It is also able to maintain high accuracy even in the presence of concept drifts. Although there are drifts detected in the datasets evaluated using the Drift Detection Method, our proposed method with incremental learning is able to achieve up to 91–94 % accuracy compared to 60–69 % without incremental learning.

Online Incremental Learning for High Bandwidth Network Traffic Classification

Data stream mining techniques are able to classify evolving data streams such as network traffic in the presence of concept drift. In order to classify high bandwidth network traffic in real-time, data stream mining classifiers need to be implemented on reconfigurable high throughput platform, such as Field Programmable Gate Array (FPGA). This paper proposes an algorithm for online network traffic classification based on the concept of incrementalk-means clustering to continuously learn from both labeled and unlabeled flow instances. Two distance measures for incrementalk-means (Euclidean and Manhattan) distance are analyzed to measure their impact on the network traffic classification in the presence of concept drift. The experimental results on real datasets show that the proposed algorithm exhibits consistency, up to 94% average accuracy for both distance measures, even in the presence of concept drifts. The proposed incrementalk-means classification using Manhattan distance can classify network traffic 3 times faster than Euclidean distance at 671 thousands flow instances per second.

Online Fault-monitoring in Machine Tools Based on Energy Consumption Analysis and Non-invasive Data Acquisition for Improved Resource-efficiency

Improving the overall equipment effectiveness of machine tools will improve resource-efficiency and productivity in manufacturing. First step to achieve more effectiveness would require sensors for monitoring of machine availability and quality of machining processes. Abnormal machine conditions are characterized by fault-pattern, which can indicate failure and quality losses. Further, machine failure can shorten the remaining useful life of the components and affect the products. Therefore, it is essential to determine a valuable data source which will enable the extraction of fault-pattern and the allocation of these pattern to machining processes. However, this can be challenging due to lack of open source control architecture, different machine types and automation degree, changing operating loads, and dynamic failure rates in a real environment. Retrofit for online analysis of electrical power intake of machine tools seems to satisfy this challenge. A fault-monitoring framework for manufacturing equipment has been proposed in this paper, based on data stream mining techniques for online pattern matching in electrical power data streams. Complex event processing is applied to ensure scalable data processing for large data volumes and automate the reporting in order to assign the fault-patterns to machining processes and products. This concept is introduced as energy-based maintenance and validated for a powertrain machining line in milling and drilling machines.

Survey on Constrained based Data Stream Mining

In most of the real time applications data may arrive as continuous ordered sequence of items, called Data streams. The main challenge in dealing with the Data Stream is its voluminous, complex and dynamically arriving stream of data. There are certain techniques to deal with data streams, in particular, finding the frequent or sequential patterns that occur repeatedly. These results retrieve huge number of patterns, which are hard to analyze and use them, also difficult to store these results and its intermediate results. The traditional pattern mining techniques fail to give the relevant details to the user. In order to obtain that, some constraint based mining techniques, which acts as a filter to the large result set retrieved from traditional pattern mining techniques. This paper investigates different data stream mining techniques and constrained based stream mining techniques from which the user gets the required information from the data stream. General Terms Data Stream, Pattern.

Data stream mining for predicting software build outcomes using source code metrics

ContextSoftware development projects involve the use of a wide range of tools to produce a software artifact. Software repositories such as source control systems have become a focus for emergent research because they are a source of rich information regarding software development projects. The mining of such repositories is becoming increasingly common with a view to gaining a deeper understanding of the development process. ObjectiveThis paper explores the concepts of representing a software development project as a process that results in the creation of a data stream. It also describes the extraction of metrics from the Jazz repository and the application of data stream mining techniques to identify useful metrics for predicting build success or failure. MethodThis research is a systematic study using the Hoeffding Tree classification method used in conjunction with the Adaptive Sliding Window (ADWIN) method for detecting concept drift by applying the Massive Online Analysis (MOA) tool. ResultsThe results indicate that only a relatively small number of the available measures considered have any significance for predicting the outcome of a build over time. These significant measures are identified and the implication of the results discussed, particularly the relative difficulty of being able to predict failed builds. The Hoeffding Tree approach is shown to produce a more stable and robust model than traditional data mining approaches. ConclusionOverall prediction accuracies of 75% have been achieved through the use of the Hoeffding Tree classification method. Despite this high overall accuracy, there is greater difficulty in predicting failure than success. The emergence of a stable classification tree is limited by the lack of data but overall the approach shows promise in terms of informing software development activities in order to minimize the chance of failure.

Wafer-to-wafer process fault detection using data stream mining techniques

In this paper, we develop a wafer-to-wafer fault detection system using data stream mining techniques for a semiconductor etch tool. The system consists of two data stream mining modules: a trace segmentation module and a multivariate trace comparison module. Each time a wafer exits the processing chamber, the trace segmentation module extracts the traces of monitored tool parameters from raw sensor data streams. We propose a novel trace segmentation algorithm called multisensor-based trace segmentation. The algorithm finds the individual start and end times of monitored tool parameters in a wafer-to-wafer fashion. For analyzing faulty tool operations, the multivariate trace comparison module performs a new principal component analysis (PCA) called a trace structure-based PCA. For each tool parameter, the structural similarity distance between a template and the extracted trace is measured using a dynamic time warping algorithm. Then, the measurements are used to build the PCA model. This approach is contrasted with the traditional PCA procedure in which the trace means are used as the building blocks for the PCA model. Experiments using the data collected from a worksite reactive ion etch tool showed that the performance of the proposed system is very encouraging.

Adaptive Real Time Data Mining Methodology for Wireless Body Area Network Based Healthcare Applications

Since the population is growing, the need for high quality and efficient healthcare, both at home and in hospital, is becoming more important. This paper presents the innovative wireless sensor network based Mobile Real-time Health care Monitoring (WMRHM) framework which has the capacity of giving health predictions online based on continuously monitored real time vital body signals . Developments in sensors, miniaturization of low-power microelectronics, and wireless networks are becoming a significant opportunity for improving the quality of health care services. Physiological signals like ECG, EEG, SpO2, BP etc. can be monitor through wireless sensor networks and analyzed with the help of data mining techniques. These real-time signals are continuous in nature and abruptly changing hence there is a need to apply an efficient and concept adapting real-time data stream mining techniques for taking intelligent health care decisions online. Because of the high speed and huge volume data set in data streams, the traditional classification technologies are no longer applicable. The most important criteria are to solve the real-time data streams mining problem with ‘concept drift’ efficiently. This paper presents the state-of-the art in this field with growing vitality and introduces the methods for detecting concept drift in data stream, then gives a significant summary of existing approaches to the problem of concept drift. The work is focused on applying these real time stream mining algorithms on vital signals of human body in Wireless Body Area Network( WBAN) based health care environment.

Discovery and diagnosis of behavioral transitions in patient event streams

Users with cognitive impairments use assistive technology (AT) as part of a clinical treatment plan. As the AT interface is manipulated, data stream mining techniques are used to monitor user goals. In this context, real-time data mining aids clinicians in tracking user behaviors as they attempt to achieve their goals. Quality metrics over stream-mined models identify potential changes in user goal attainment, as the user learns his or her personalized emailing system. When the quality of some data-mined models varies significantly from nearby models—as defined by quality metrics—the user's behavior is then flagged as a significant behavioral change. The specific changes in user behavior are then characterized by differencing the data-mined decision tree models. This article describes how model quality monitoring and decision tree differencing can aid in recognition and diagnoses of behavioral changes in a case study of cognitive rehabilitation via emailing. The technique may be more widely applicable to other real-time data-intensive analysis problems.

Anomalous Network Packet Detection Using Data Stream Mining

In recent years, significant research has been devoted to the development of Intrusion Detection Systems (IDS) able to detect anomalous computer network traffic indicative of malicious activity. While signature-based IDS have proven effective in discovering known attacks, anomaly-based IDS hold the even greater promise of being able to automatically detect previously undocumented threats. Traditional IDS are generally trained in batch mode, and therefore cannot adapt to evolving network data streams in real time. To resolve this limitation, data stream mining techniques can be utilized to create a new type of IDS able to dynamically model a stream of network traffic. In this paper, we present two methods for anomalous network packet detection based on the data stream mining paradigm. The first of these is an adapted version of the DenStream algorithm for stream clustering specifically tailored to evaluate network traffic. In this algorithm, individual packets are treated as points and are flagged as normal or abnormal based on their belonging to either normal or outlier clusters. The second algorithm utilizes a histogram to create a model of the evolving network traffic to which incoming traffic can be compared using Pearson correlation. Both of these algorithms were tested using the first week of data from the DARPA ’99 dataset with Generic HTTP, Shell-code and Polymorphic attacks inserted. We were able to achieve reasonably high detection rates with moderately low false positive percentages for different types of attacks, though detection rates varied between the two algorithms. Overall, the histogram-based detection algorithm achieved slightly superior results, but required more parameters than the clustering-based algorithm. As a result of its fewer parameter requirements, the clustering approach can be more easily generalized to different types of network traffic streams.