No abstract

No abstract

A note on versions:The version presented here may differ from the published version or from the version of record. If you wish to cite this item you are advised to consult the publisher's version. Please see the repository url above for details on accessing the published version and note that access may require a subscription. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.ACCEPTED MANUSCRIPT A C C E P T E D M A N U S C R I P T AbstractThe k-Nearest Neighbors classifier is a simple yet effective widely renowned method in data mining. The actual application of this model in the big data domain is not feasible due to time and memory restrictions. Several distributed alternatives based on MapReduce have been proposed to enable this method to handle large-scale data. However, their performance can be further improved with new designs that fit with newly arising technologies.In this work we provide a new solution to perform an exact k-nearest neighbor classification based on Spark. We take advantage of its in-memory operations to classify big amounts of unseen cases against a big training dataset. The map phase computes the k-nearest neighbors in different training data splits. Afterwards, multiple reducers process the definitive neighbors from the list obtained in the map phase. The key point of this proposal lies on the management of the test set, keeping it in memory when possible. Otherwise, it is split into a minimum number of pieces, applying a MapReduce per chunk, using the caching skills of Spark to reuse the previously partitioned * Corresponding author. ACCEPTED MANUSCRIPT A C C E P T E D M A N U S C R I P T training set. In our experiments we study the differences between Hadoop and Spark implementations with datasets up to 11 million instances, showing the scaling-up capabilities of the proposed approach. As a result of this work an open-source Spark package is available.

In the era of big data, analyzing and extracting knowledge from large-scale data sets is a very interesting and challenging task. The application of standard data mining tools in such data sets is not straightforward. Hence, a new class of scalable mining method that embraces the huge storage and processing capacity of cloud platforms is required. In this work, we propose a novel distributed partitioning methodology for prototype reduction techniques in nearest neighbor classification. These methods aim at representing original training data sets as a reduced number of instances. Their main purposes are to speed up the classification process and reduce the storage requirements and sensitivity to noise of the nearest neighbor rule. However, the standard prototype reduction methods cannot cope with very large data sets. To overcome this limitation, we develop a MapReduce-based framework to distribute the functioning of these algorithms through a cluster of computing elements, proposing several algorithmic strategies to integrate multiple partial solutions (reduced sets of prototypes) into a single one. The proposed model enables prototype reduction algorithms to be applied over big data classification problems without significant accuracy loss. We test the speeding up capabilities of our model with data sets up to 5.7 millions of Email addresses: triguero@decsai.ugr.es (Isaac Triguero), dperalta@decsai.ugr.es (Daniel Peralta), jaume.bacardit@newcastle.ac.uk (Jaume Bacardit), sglopez@ujaen.es (Salvador García), herrera@decsai.ugr.es (Francisco Herrera) March 3, 2014 instances. The results show that this model is a suitable tool to enhance the performance of the nearest neighbor classifier with big data. Preprint submitted to Neurocomputing

Detecting anomalies in time series data is becoming mainstream in a wide variety of industrial applications in which sensors monitor expensive machinery. The complexity of this task increases when multiple heterogeneous sensors provide information of different nature, scales and frequencies from the same machine. Traditionally, machine learning techniques require a separate data preprocessing before training, which tends to be very time-consuming and often requires domain knowledge. Recent deep learning approaches have shown to perform well on raw time series data, eliminating the need for pre-processing. In this work, we propose a deep learning based approach for supervised multitime series anomaly detection that combines a Convolutional Neural Network (CNN) and a Recurrent Neural Network (RNN) in different ways. Unlike other approaches, we use independent CNNs, so-called convolutional heads, to deal with anomaly detection in multi-sensor systems. We address each sensor individually avoiding the need for data pre-processing and allowing for a more tailored architecture for each type of sensor. We refer to this architecture as Multi-head CNN-RNN. The proposed architecture is assessed against a real industrial case study, provided by an industrial partner, where a service elevator is monitored. Within this case study, three type of anomalies are considered: point, context-specific, and collective. The experimental results show that the proposed architecture is suitable for multi-time series anomaly detection as it obtained promising results on the real industrial scenario.

The application of data mining and machine learning techniques to biological and biomedicine data continues to be an ubiquitous research theme in current bioinformatics. The rapid advances in biotechnology are allowing us to obtain and store large quantities of data about cells, proteins, genes, etc, that should be processed. Moreover, in many of these problems such as contact map prediction, the problem tackled in this paper, it is difficult to collect representative positive examples. Learning under these circumstances, known as imbalanced big data classification, may not be straightforward for most of the standard machine learning methods.In this work we describe the methodology that won the ECBDL'14 big data challenge for a bioinformatics big data problem. This algorithm, named as ROSEFW-RF, is based on several MapReduce approaches to (1) balance the classes distribution through random oversampling, (2) detect the most relevant features via an evolutionary feature weighting process and a thresh-

This paper introduces the 3 rd major release of the KEEL Software. KEEL is an open source Java framework (GPLv3 license) that provides a number of modules to perform a wide variety of data mining tasks. It includes tools to perform data management, design of multiple kind of experiments, statistical analyses, etc. This framework also contains KEEL-dataset, a data repository for multiple learning tasks featuring data partitions and algorithms' results over these problems. In this work, we describe the most recent components added to KEEL 3.0, including new modules for semi-supervised learning, multi-instance learning, imbalanced classification and subgroup discovery. In addition, a new interface in R has been incorporated to execute algorithms included in KEEL. These new features greatly improve the versatility of KEEL to deal with more modern data mining problems.

a b s t r a c tNearest neighbor classification is one of the most used and well known methods in data mining. Its simplest version has several drawbacks, such as low efficiency, high storage requirements and sensitivity to noise. Data reduction techniques have been used to alleviate these shortcomings. Among them, prototype selection and generation techniques have been shown to be very effective. Positioning adjustment of prototypes is a successful trend within the prototype generation methodology.Evolutionary algorithms are adaptive methods based on natural evolution that may be used for searching and optimization. Positioning adjustment of prototypes can be viewed as an optimization problem, thus it can be solved using evolutionary algorithms. This paper proposes a differential evolution based approach for optimizing the positioning of prototypes. Specifically, we provide a complete study of the performance of four recent advances in differential evolution. Furthermore, we show the good synergy obtained by the combination of a prototype selection stage with an optimization of the positioning of prototypes previous to nearest neighbor classification. The results are contrasted with non-parametrical statistical tests and show that our proposals outperform previously proposed methods.