Javier Del Ser

Hybridizing Cartesian Genetic Programming and Harmony Search for adaptive feature construction in supervised learning problems

HighlightsWe present a new iterative feature construction approach for supervised learning model based on the meta-heuristic Harmony Search (HS) algorithm and Cartesian Genetic Programming.We propose a novel method to incorporate soft information about the relevance of the constructed features in the HS algorithm so as to enhance its convergence.The performance of the proposed scheme is assessed over datasets from the literature, with promising results that support its suitability to deal with legacy datasets. The advent of the so-called Big Data paradigm has motivated a flurry of research aimed at enhancing machine learning models by following very diverse approaches. In this context this work focuses on the automatic construction of features in supervised learning problems, which differs from the conventional selection of features in that new characteristics with enhanced predictive power are inferred from the original dataset. In particular this manuscript proposes a new iterative feature construction approach based on a self-learning meta-heuristic algorithm (Harmony Search) and a solution encoding strategy (correspondingly, Cartesian Genetic Programming) suited to represent combinations of features by means of constant-length solution vectors. The proposed feature construction algorithm, coined as Adaptive Cartesian Harmony Search (ACHS), incorporates modifications that allow exploiting the estimated predictive importance of intermediate solutions and, ultimately, attaining better convergence rate in its iterative learning procedure. The performance of the proposed ACHS scheme is assessed and compared to that rendered by the state of the art in a toy example and three practical use cases from the literature. The excellent performance figures obtained in these problems shed light on the widespread applicability of the proposed scheme to supervised learning with legacy datasets composed by already refined characteristics.

Semantic Information Fusion of Linked Open Data and Social Big Data for the Creation of an Extended Corporate CRM Database

The amount of on-line available open information from heterogeneous sources and domains is growing at an extremely fast pace, and constitutes an important knowledge base for the consideration of industries and companies. In this context, two relevant data providers can be highlighted: the “Linked Open Data” and “SocialMedia” paradigms. The fusion of these data sources – structured the former, and raw data the latter –, along with the information contained in structured corporate databases within the organizations themselves, may unveil significant business opportunities and competitive advantage to those who are able to understand and leverage their value. In this paper, we present a use case that represents the creation of an existing and potential customer knowledge base, exploiting social and linked open data based on which any given organization might infer valuable information as a support for decision making. In order to achieve this a solution based on the synergy of big data and semantic technologies will be designed and developed. The first will be used to implement the tasks of collection and initial data fusion based on natural language processing techniques, whereas the latter will perform semantic aggregation, persistence, reasoning and retrieval of information, as well as the triggering of alerts over the semantized information.

On a Machine Learning Approach for the Detection of Impersonation Attacks in Social Networks

Lately the proliferation of social networks has given rise to a myriad of fraudulent strategies aimed at getting some sort of benefit from the attacked individual. Despite most of them being exclusively driven by economic interests, the so called impersonation, masquerading attack or identity fraud hinges on stealing the credentials of the victim and assuming his/her identity to get access to resources (e.g. relationships or confidential information), credit and other benefits in that person’s name. While this problem is getting particularly frequent within the teenage community, the reality is that very scarce technological approaches have been proposed in the literature to address this issue which, if not detected in time, may catastrophically unchain other fatal consequences to the impersonated person such as bullying and intimidation. In this context, this paper delves into a machine learning approach that permits to efficiently detect this kind of attacks by solely relying on connection time information of the potential victim. The manuscript will demonstrate how these learning algorithms - in particular, support vector classifiers - can be of great help to understand and detect impersonation attacks without compromising the user privacy of social networks.

Chapter 5 – Big Data in Road Transport and Mobility Research

Abstract Ubiquitous computing has changed the acquisition of mobility data, with two aspects contributing: the high penetration rate and the ability to capture and share information on a continuous basis. This applies to geolocation information, operational mobile phone data, and also, social network crowdsourced information. Additionally, under the umbrella of the Internet of Things trend, the deployment of the Connected Vehicle (Car-as-a-sensor) concept, supported by advanced V2X communications, provides massive data volume. For all these cases, data are open to never before seen opportunities to analyze and predict individual and aggregated mobility patterns. Big Data refers to the processsing capabilities of such an explosion in the amount, quality, and heterogeneity of available data. This chapter will review the most relevant data sources, introduce the underlying techniques supporting the BigData paradigm and, finally, provide a list of some relevant applications in the transport and mobility domain.

Environmental Perception for Intelligent Vehicles

Abstract Environmental perception represents, because of its complexity, a challenge for Intelligent Transport Systems due to the great variety of situations and different elements that can happen in road environments and that must be faced by these systems. In connection with this, so far there are a variety of solutions as regards sensors and methods, so the results of precision, complexity, cost, or computational load obtained by these works are different. In this chapter some systems based on computer vision and laser techniques are presented. Fusion methods are also introduced in order to provide advanced and reliable perception systems.

Positioning and Digital Maps

A reliable positioning system is essential for the development of intelligent vehicles. This chapter provides an overview of different technologies and techniques that are crucial to understand modern positioning systems onboard road vehicles. It is written for the purpose of serving as a guide to students, engineers, and researchers in the field of vehicular technology or Intelligent Transportation Systems. In Section 4.1, after an introduction to the problem, a handy list of key definitions is provided, and some of the most relevant Location-Based Services mentioned. Section 4.2 presents the fundamentals of GNSS-based positioning. Aiding technologies, such as odometers and inertial sensors, and techniques for GNSS-based hybridized positioning are discussed in Section 4.3. Later, Section 4.4 analyzes the role of digital maps, map-matching, and map-aided positioning. Finally, Section 4.5 introduces alternatives to GNSS, such as visual odometry, with a brief mention of wireless networks and RFID.