Data Mining in Time Series and Streaming Databases

The following sections are included:

• Dedication
• Preface
• About the Editors
• About the Contributors
• Contents

Fast Big Data is being produced at high-velocity in real-time. To effectively deal with this type of streaming data produced in real time, we need to be able to adapt to changes on the distribution of the data being produced, and we need to do it using the minimum amount of time and memory. The Internet of Things (IoT) is a good example and motivation of this type of streaming data produced in real time.

Massive Online Analytics (MOA) is a software environment for implementing algorithms and running experiments for online learning from evolving data streams. MOA is designed to deal with the challenging problem of scaling up the implementation of state of the art algorithms to real world dataset sizes. MOA includes classification and clustering methods. It contains collection of offline and online methods as well as tools for evaluation. MOA supports bi-directional interaction with WEKA, the Waikato Environment for Knowledge Analysis, and is released under the GNU GPL license.

Learning from data streams can only be realized by systems which are not only effective but also efficient. That is, knowledge discovery in this context is impossible without being aware of the computational resources available. Weightless artificial neural networks (WANNs) are based on an alternative principle to iterative optimization of weights employed by most mainstream artificial neural network models and related tools. WANNs explicitly manage knowledge pieces, which are stored by RAM nodes. Such foundational difference reflects on the adaptability of these models to streaming inputs: in such scenario, the application of weightless models can be considered more natural than the same for their weighted counterparts, with an ample control over learning capability as well as resources consumption. This chapter details a WANN-based approach for mining data streams, which allows the maintenance of an up-to-date data summary which can be used for several purposes. The insights and original ideas which power such model are explained as well, enabling novel applications and further development of them.

Stream classification is a challenging research field in which algorithms are required to process data online, use minimal resources, and react to concept changes. The task of mining incoming instances becomes even more demanding when the classifier is required to cope with imbalanced data — situations when one of the target classes is represented by much less instances than other classes. This chapter gives an overview of research on imbalanced stream classification. We present dedicated ensemble algorithms designed to cope with concept changes, discuss challenges posed by imbalanced class distributions along with assorted difficulty factors, and give an outlook on how class imbalance and concept changes can interact.

Learning a prototype from a set of given objects is a core problem in machine learning, data mining, and pattern recognition. A commonly used approach to consensus learning is to formulate it as an optimization problem in terms of generalized median computation. Sequential data can be effectively represented by strings. In this chapter we discuss how the median concept can be applied to strings. We introduce the notion of median string and provide related theoretical results. Then, we give a brief review of algorithmic procedures for computing median strings. Experimental results will be reported to compare some of the shown algorithms.

Automated categorization of textual information is becoming an increasingly important task in the digital world. However, most classification algorithms build upon manual labeling of text documents, which is a time-consuming and costly process. In this paper, we present a novel methodology for clustering-based classification of stationary document streams using active learning. The proposed active learning clustering-based classification algorithm (ACCA) obtains a continuous stream of unlabeled documents. The arriving documents are clustered incrementally so that each incoming document is inserted into an existing cluster or used to start a new cluster of its own. The number of possible clusters is unlimited. From time to time, an expert is called to label several clusters for the classification mechanism. With arrival of more documents, the expert can be called less frequently, since most of the incoming documents will eventually belong to existing labeled clusters. Our algorithm is aimed at finding the fastest way of reaching the point where most arriving documents can be classified automatically without the experts assistance. The evaluation experiments on two benchmark corpora show that active learning and clustering can increase the percentage of automatically and accurately categorized documents over time.

The selection and pre-processing of data are key activities in the interpretation and extraction of knowledge from data streams. Their role become even more crucial in the context of big data mining where huge amounts of (volatile) data from various types are normal, and simply storing all data is not only becoming infeasible but also undesired. This makes the exploitation of domain knowledge the key ingredient to properly reduce the data stream while simultaneously keep information loss to a minimum. Using several examples from the completely different fields of cyber security, astronomy (in particular the detection of exoplanets), criminal justice (in particular in the Netherlands) and cyber security the importance of domain knowledge is illustrated. For each of these examples time series analysis is the goal, but the challenges during the selection and pre-processing stages are completely different. However, despite the differences, two different, but related, trends can be distinguished. The first trend shows that steps that need to be emphasized during pre-processing depend on the consistency of the time series data, and the second trend shows a shift from optimizing the records included in the time series data to optimizing the approach taken to allow the data to be mined.

Over the past few years, a lot of devices and machines around us are becoming ‘smart’. Based on the idea of the Internet of Things (IoT), different devices and machines can connect to the internet and communicate with each other. Such internet enabled devices are continuously observing their environment and logging a lot of data in the back-end database. By applying data analytics on the gathered Big Data, smart decisions are made to facilitate the end user according to the current situation. This capability of adaptive decision making actually makes ordinary devices and machines ‘smart’. These devices and machines are becoming intelligent by learning about their surroundings from different sources, and develop the ability to avoid unforeseen situations by analyzing that data. In this chapter, we provide a comprehensive overview of how different industrial players are using data analytics to provide better services to their customers and improve their internal processes and workflows. We discuss how different industries use data analytics to gain vital insights for providing better healthcare to public, making homes more secure, increasing crop yield, delivering goods more quickly, reducing the downtime of a machine, avoiding a disease, etc. An overview of different analytics platforms and solutions used in different industries for time series and streaming data are also discussed in this chapter.

The following section is included:

• Index