Nikolaos Nodarakis

Efficient Multidimensional AkNN Query Processing in the Cloud

A k-nearest neighbor (kNN) query determines the k nearest points, using distance metrics, from a given location. An all k-nearest neighbor (AkNN) query constitutes a variation of a kNN query and retrieves the k nearest points for each point inside a database. Their main usage resonates in spatial databases and they consist the backbone of many location-based applications and not only. In this work, we propose a novel method for classifying multidimensional data using an AkNN algorithm in the MapReduce framework. Our approach exploits space decomposition techniques for processing the classification procedure in a parallel and distributed manner. To our knowledge, we are the first to study the kNN classification of multidimensional objects under this perspective. Through an extensive experimental evaluation we prove that our solution is efficient, robust and scalable in processing the given queries.

Large Scale Implementations for Twitter Sentiment Classification

Sentiment Analysis on Twitter Data is indeed a challenging problem due to the nature, diversity and volume of the data. People tend to express their feelings freely, which makes Twitter an ideal source for accumulating a vast amount of opinions towards a wide spectrum of topics. This amount of information offers huge potential and can be harnessed to receive the sentiment tendency towards these topics. However, since no one can invest an infinite amount of time to read through these tweets, an automated decision making approach is necessary. Nevertheless, most existing solutions are limited in centralized environments only. Thus, they can only process at most a few thousand tweets. Such a sample is not representative in order to define the sentiment polarity towards a topic due to the massive number of tweets published daily. In this work, we develop two systems: the first in the MapReduce and the second in the Apache Spark framework for programming with Big Data. The algorithm exploits all hashtags and emoticons inside a tweet, as sentiment labels, and proceeds to a classification method of diverse sentiment types in a parallel and distributed manner. Moreover, the sentiment analysis tool is based on Machine Learning methodologies alongside Natural Language Processing techniques and utilizes Apache Spark’s Machine learning library, MLlib. In order to address the nature of Big Data, we introduce some pre-processing steps for achieving better results in Sentiment Analysis as well as Bloom filters to compact the storage size of intermediate data and boost the performance of our algorithm. Finally, the proposed system was trained and validated with real data crawled by Twitter, and, through an extensive experimental evaluation, we prove that our solution is efficient, robust and scalable while confirming the quality of our sentiment identification.

kdANN+: A Rapid AkNN Classifier for Big Data

A k-nearest neighbor kNN query determines the k nearest points, using distance metrics, from a given location. An all k-nearest neighbor AkNN query constitutes a variation of a kNN query and retrieves the k nearest points for each point inside a database. Their main usage resonates in spatial databases and they consist the backbone of many location-based applications and not only. In this work, we propose a novel method for classifying multidimensional data using an AkNN algorithm in the MapReduce framework. Our approach exploits space decomposition techniques for processing the classification procedure in a parallel and distributed manner. To our knowledge, we are the first to study the kNN classification of multidimensional objects under this perspective. Through an extensive experimental evaluation we prove that our solution is efficient, robust and scalable in processing the given queries.

MR-SAT: A MapReduce Algorithm for Big Data Sentiment Analysis on Twitter

Sentiment analysis on Twitter data has attracted much attention recently. People tend to express their feelings freely, which makes Twitter an ideal source for accumulating a vast amount of opinions towards a wide diversity of topics. In this paper, we develop a novel method to harvest sentiment knowledge in the MapReduce framework. Our algorithm exploits the hashtags and emoticons inside a tweet, as sentiment labels, and proceeds to a classification procedure of diverse sentiment types in a parallel and distributed manner. Moreover, we utilize Bloom filters to compact the storage size of intermediate data and boost the performance of our algorithm. Through an extensive experimental evaluation, we prove that our solution is efficient, robust and scalable and confirm the quality of our sentiment identification.

(A)kNN Query Processing on the Cloud: A Survey

A k-nearest neighbor (kNN) query determines the k nearest points, using distance metrics, from a given location. An all k-nearest neighbor (AkNN) query constitutes a variation of a kNN query and retrieves the k nearest points for each point inside a database. Their main usage resonates in spatial databases and they consist the backbone of many location-based applications and not only. Although (A)kNN is a fundamental query type, it is computationally very expensive. During the last years a multiplicity of research papers has focused around the distributed (A)kNN query processing on the cloud. This work constitutes a survey of research efforts towards this direction. The main contribution of this work is an up-to-date review of the latest (A)kNN query processing approaches. Finally, we discuss various research challenges and directions of further research around this domain.

Mining Biological Data on the Cloud – A MapReduce Approach

During last decades, bioinformatics has proven to be an emerging field of research leading to the development of a wide variety of applications. The primary goal of bioinformatics is to detect useful knowledge hidden under large volumes biological and biomedical data, gain a greater insight into their relationships and, therefore, enhance the discovery and the comprehension of biological processes. To achieve this, a great number of text mining techniques have been developed that efficiently manage and disclose meaningful patterns and correlations from biological and biomedical data repositories. However, as the volume of data grows rapidly these techniques cannot cope with the computational burden that is produced since they apply only in centralized environments. Consequently, a turn into distributed and parallel solutions is indispensable. In the context of this work, we propose an efficient and scalable solution, in the MapReduce framework, for mining and analyzing biological and biomedical data.

Intelligent system for cultural objects identification, damage assessment and restoration

Cultural objects and art works need ongoing conservation interventions in order to be available for the next generations. The most object-friendly analysis approaches are based on non-destructive techniques (NDTs) that allow both the materials characterisation as well as the decay detection of cultural artefacts. Non-destructive testing and evaluation includes the employment of several methods such as the well-established technique of diffuse reflectance spectroscopy with fibre optics (FORS). Such techniques produce output with multiple series of data for multiple different pigment used in objects. In this work, we present a data management solution that contributes with: 1) a library of known reference pigments/colours; 2) a proposed pattern matching technique that allows the automatic classification of any new pigment. The experimental evaluation results show that the data processing proposed is effective. Feedback is particularly encouraging as it allows automation and therefore radically decreased time for pigment/colour matching and identification.

NSM-Tree: Efficient Indexing on Top of NoSQL Databases

During the last years, there is a huge proliferation in the usage of location-based services (LBSs), mostly through a multitude of mobile devices (GPS, smartphones, mapping devices, etc.). The volume of the data derived by such services, grows exponentially and conventional databases tend to be ineffective in storing and indexing them efficiently. Ultimately, we need to turn to scalable solutions and methods using the NoSQL database model. Quite a few indexing methods exist in literature that work on top of NoSQL database. In this spirit, we deploy a new distributed indexing structure based on M-tree and perform a thorough experimental analysis to display its benefits.

Standardizing NDT& E Techniques and Conservation Metadata for Cultural Artifacts

Conservation activities, before and after decay detection, are considered as a prerequisite for maintaining cultural artifacts in their initial/original form. Taking into account the strict regulations where sampling from art works of great historical value is restricted or in many cases prohibited, the application of Non-Destructive Testing techniques (NDTs) during the conservation or even decay detection is highly appreciated by conservators. Non-destructive examination include the employment of multiple analysis approaches and techniques namely Infrared Thermography (IRT), Ultrasonics (US), Ground Penetrating Radar (GPR), VIS–NIR Fiber Optics Diffuse Reflectance Spectroscopy (FORS), portable X-Ray Fluorescence (XRF), Environmental Scanning Electron Microscopy with Energy Dispersive X-Ray Analysis (ESEM-EDX), Attenuated Total Reflectance-Fourier Transform Infrared Spectroscopy (ATR-FTIR) and micro-Raman Spectroscopy. These produce a huge amount of data, in different formats, such as text, numerical sets and visual objects (i.e. images, thermograms, radargrams, spectral data, graphs, etc). Moreover, conservation documentation presents major drawbacks, as fragmentation and incomplete description of the related information is usually the case. Assigning conservation data to the objects’ metadata collection is very rare and not yet standardized. The Doc-Culture Project aims to provide solutions for the NDT application methodologies, analysis and process along with their output data and all related conservation documentation. The preliminary results are discussed in this paper.

Convex Polygon Planar Range Queries on the Cloud: Grid vs. Angle-Based Partitioning

The polygon retrieval problem is, in essence, the problem of preprocessing a set of n 2-dimensional points, so than given a special ContainedIn spatial query, the subset of points falling inside the polygon can be reported efficiently. Such queries find great applicability in areas such as computer graphics, spatial databases and GIS applications. However, as the size of spatial data grows rapidly existing centralized solutions fail to retrieve the results in reasonable response time. In this paper, we propose a novel MapReduce algorithm for efficiently processing convex polygon planar range queries in a distributed manner. We apply a grid-based and an angle-based partitioning scheme on the data space and perform a comparative analysis. Through our experimental evaluation we prove that our system is efficient, robust and scalable.