Search

Machine learning and deep learning have become the most useful and powerful tools in the last years to mine information from large datasets. Despite the successful application to many research fields, it is widely known that some of these solutions based on artificial intelligence are considered black-box models, meaning that most experts find difficult to explain and interpret the models and why they generate such outputs. In this context, explainable artificial intelligence is emerging with the aim of providing black-box models with sufficient interpretability. Thus, models could be easily understood and further applied. This work proposes a novel method to explain black-box models, by using numeric association rules to explain and interpret multi-step time series forecasting models. Thus, a multi-objective algorithm is used to discover quantitative association rules from the target model. Then, visual explanation techniques are applied to make the rules more interpretable. Data from Spanish electricity energy consumption has been used to assess the suitability of the proposal.

Deep learning has become one of the most useful tools in the last years to mine information from large datasets. Despite the successful application to many research fields, deep learning is known as a black box approach and most experts experience difficulties to explain and interpret deep learning results. In this context, explainable artificial intelligence (XAI) is emerging with the aim of providing black box models with sufficient interpretability so that models can be easily understood by humans. The use of an evolutionary-based association rules extraction algorithm to explain deep learning models for multi-step time series forecasting is addressed in this work. This evolutionary application is proposed to be used with the predictions obtained by long-short term memory (LSTM) deep learning network. Data from Spanish electricity energy consumption has been used to assess the suitability of the

Renewable energies are currently experiencing promising growth as an alternative solution to minimize the emission of pollutant gases from the use of fossil fuels, which contribute to global warming. To integrate these renewable energies safely with the grid system and make the electric grid system more stable, it is vitally important to accurately forecast the amount of wind power generated at specific wind power generation sites and the timing of this generation. Deep learning approaches have shown good forecasting performance for complex and nonlinear problems, such as time series wind power data. However, further study is needed to optimize deep learning models by integrating multiple models with hyperparameter optimization, to attain optimal performance from these individual models. In this paper, we propose a hybrid CNN-LSTM model for wind power forecasting in Ethiopia. Bayesian optimization is applied to tune the hyperparameters of the individual learners, including 1D-CNN and LSTM models, before building the hybrid CNN-LSTM model. The proposed model is tested on three case study wind power datasets obtained from the Ethiopian Electric Power Corporation. According to the MAE, RMSE, and MAPE evaluation metrics, the hybrid model performs significantly better than benchmark models, including ANN, RNN, BiLSTM, CNN, and LSTM models, for all case study data.

This paper explores the use of deep learning techniques for detecting sleep apnea. Sleep apnea is a common sleep disorder characterized by abnormal breathing pauses or infrequent breathing during sleep. The current standard for diagnosing sleep apnea involves overnight polysomnography, which is expensive and requires specialized equipment and personnel. The proposed method utilizes a neural network to analyze physiological signals, such as heart rate and respiratory patterns, that are recorded during sleep to authomatic sleep apnea detection. The neural network is trained on a dataset of polysomnography recordings to identify patterns that are indicative of sleep apnea. The results compare the use of different physiological signals to detect sleep apnea. Nasal airflow seems to have the most accurate results and higher specificity, whereas EEG and ECG have higher levels of sensitivity. The best model concerning accuracy is compared to bias models previously applied to sleep apnea detection in literature, achieving greater results. This approach has the potential to provide automatic sleep apnea detection, being an accessible solution for diagnosing sleep apnea.

Machine and deep learning has become one of the most useful tools in the last years as a diagnosis-decision-support tool in the health area. However, it is widely known that artificial intelligence models are considered a black box and most experts experience difficulties explaining and interpreting the models and their results. In this context, explainable artificial intelligence is emerging with the aim of providing black-box models with sufficient interpretability so that models can be easily understood and further applied. Obstructive sleep apnea is a common chronic respiratory disease related to sleep. Its diagnosis nowadays is done by processing different data signals, such as electrocardiogram or respiratory rate. The waveform of the respiratory signal is of importance too. Machine learning models could be applied to the signal's analysis. Data from a polysomnography study for automatic sleep apnea detection have been used to evaluate the use of the Local Interpretable Model-Agnostic (LIME) library for explaining the health data models. Results obtained help to understand how several features have been used in the model and their influence in the quality of sleep.

Precision agriculture focuses on the development of site-specific harvest considering the variability of each crop area. Vegetation indices allow the study and delineation of different characteristics of each field zone, generally invisible to the naked-eye. This paper introduces a new big data triclustering approach based on evolutionary algorithms. The algorithm shows its capability to discover three-dimensional pat-terns on the basis of vegetation indices from vine crops. Different vegetation indices have been tested to find different patterns in the crops. The results reported using a vineyard crop located in Portugal depicts four areas with different moisture stress particularities that can lead to changes in the management of the vineyard. Furthermore, scalability studies have been performed, showing that the proposed algorithm is suitable for dealing with big datasets.

The disease caused by the SARS-CoV-2 (COVID-19) has affected millions of people around the world since its detection in 2019. This pandemic inspired the development of the Coronavirus Optimization Algorithm (CVOA), a bio-inspired metaheuristic that was originally used to adjust deep learning models for time series forecasting, by means of a binary codification. In this paper, a integer codification for the CVOA individual is introduced and used for optimizing a novel approach for numerical association rules mining. As an application case, the prediction of earthquakes of large magnitude has been addressed. This kind of events are rare and, therefore, they can be characterized by rules with very high interest or lift and low support. Thus, the algorithm has been applied to the extraction of rules meeting specific criteria in an earthquake data set, provided by the National Geographic Institute of Spain. The results show CVOA as a promising tool for numerical association rules mining, obtaining rules with useful and meaningful information for predicting the occurrence of large earthquakes.

This paper presents a new forecasting algorithm for time series in streaming named StreamWNN. The methodology has two well-differentiated stages: the algorithm searches for the nearest neighbors to generate an initial prediction model in the batch phase. Then, an online phase is carried out when the time series arrives in streaming. In par-ticular, the nearest neighbor of the streaming data from the training set is computed and the nearest neighbors, previously computed in the batch phase, of this nearest neighbor are used to obtain the predictions. Results using the electricity consumption time series are reported, show-ing a remarkable performance of the proposed algorithm in terms of fore-casting errors when compared to a nearest neighbors-based benchmark algorithm. The running times for the predictions are also remarkable

Triclustering algorithms group sets of coordinates of 3-dimensional datasets. In this paper, a new triclustering approach for data streams is introduced. It follows a streaming scheme of learning in two steps: offline and online phases. First, the offline phase provides a sum mary model with the components of the triclusters. Then, the second stage is the online phase to deal with data in streaming. This online phase consists in using the summary model obtained in the offline stage to update the triclusters as fast as possible with genetic operators. Results using three types of synthetic datasets and a real-world environmental sensor dataset are reported. The performance of the proposed triclustering streaming algo rithm is compared to a batch triclustering algorithm, showing an accurate performance both in terms of quality and running times

One of the techniques that provides systematic insights into biolog ical processes is High-Content Screening (HCS). It measures cells phenotypes simultaneously. When analysing these images, features like fluorescent colour, shape, spatial distribution and interaction between components can be found. STriGen, which works in the real-time environment, leads to the possibility of studying time evolution of these features in real-time. In addition, data stream ing algorithms are able to process flows of data in a fast way. In this article, STriGen (Streaming Triclustering Genetic) algorithm is presented and applied to HCS images. Results have proved that STriGen finds quality triclusters in HCS images, adapts correctly throughout time and is faster than re-computing the triclustering algorithm each time a new data stream image arrives.

This study proposes a novel bioinspired metaheuristic simulating how the coronavirus spreads and infects healthy people. From a primary infected individual (patient zero), the coronavirus rapidly infects new victims, creating large populations of infected people who will either die or spread infection. Relevant terms such as reinfection probability, super-spreading rate, social distancing measures, or traveling rate are introduced into the model to simulate the coronavirus activity as accurately as possible. The infected population initially grows exponentially over time, but taking into consideration social isolation measures, the mortality rate, and number of recoveries, the infected population gradually decreases. The coronavirus optimization algorithm has two major advantages when compared with other similar strategies. First, the input parameters are already set according to the disease statistics, preventing researchers from initializing them with arbitrary values. Second, the approach has the ability to end after several iterations, without setting this value either. Furthermore, a parallel multivirus version is proposed, where several coronavirus strains evolve over time and explore wider search space areas in less iterations. Finally, the metaheuristic has been combined with deep learning models, to find optimal hyperparameters during the training phase. As application case, the problem of electricity load time series forecasting has been addressed, showing quite remarkable performance.

Agriculture has undergone some very important changes over the last few decades. The emergence and evolution of precision agri culture has allowed to move from the uniform site management to the site-specific management, with both economic and environmental advan tages. However, to be implemented effectively, site-specific management requires within-field spatial variability to be well-known and character ized. In this paper, an algorithm that delineates within-field management zones in a maize plantation is introduced. The algorithm, based on tri clustering, mines clusters from temporal remote sensing data. Data from maize crops in Alentejo, Portugal, have been used to assess the suit ability of applying triclustering to discover patterns over time, that may eventually help farmers to improve their harvests.

The vast amount of data stored nowadays has turned big data analytics into a very trendy research field. The Spark distributed computing platform has emerged as a dominant and widely used paradigm for cluster deployment and big data analytics. However, to get started up is still a task that may take much time when manually done, due to the requisites that all nodes must fulfill. This work introduces LadonSpark, an open-source and non-commercial solution to configure and deploy a Spark cluster automatically. It has been specially designed for easy and efficient management of a Spark cluster with a friendly graphical user interface to automate the deployment of a cluster and to start up the distributed file system of Hadoop quickly. Moreover, LadonSpark includes the functionality of integrating any algorithm into the system. That is, the user only needs to provide the executable file and the number of required inputs for proper parametrization. Source codes developed in Scala, R, Python, or Java can be supported on LadonSpark. Besides, clustering, regression, classification, and association rules algorithms are already integrated so that users can test its usability from its initial installation.

The use of electric vehicle across the world has become one of the most challenging issues for environmental policies. The galloping climate change and the expected running out of fossil fuels turns the use of such non-polluting cars into a priority for most developed countries. However, such a use has led to major concerns to power companies, since they must adapt their generation to a new scenario, in which electric vehicles will dramatically modify the curve of generation. In this paper, a novel approach based on ensemble learning is proposed. In particular, ARIMA, GARCH and PSF algorithms' performances are used to forecast the electric vehicle power consumption in Spain. It is worth noting that the studied time series of consumption is non-stationary and adds difficulties to the forecasting process. Thus, an ensemble is proposed by dynamically weighting all algorithms over time. The proposal presented has been implemented for a real case, in particular, at the Spanish Control Centre for the Electric Vehicle. The performance of the approach is assessed by means of WAPE, showing robust and promising results for this research field.

In this paper, we introduce a deep learning approach, based on feed-forward neural networks, for big data time series forecasting with arbitrary prediction horizons. We firstly propose a random search to tune the multiple hyper-parameters involved in the method perfor-mance. There is a twofold objective for this search: firstly, to improve the forecasts and, secondly, to decrease the learning time. Next, we pro-pose a procedure based on moving averages to smooth the predictions obtained by the different models considered for each value of the pre-diction horizon. We conduct a comprehensive evaluation using a real-world dataset composed of electricity consumption in Spain, evaluating accuracy and comparing the performance of the proposed deep learning with a grid search and a random search without applying smoothing. Reported results show that a random search produces competitive accu-racy results generating a smaller number of models, and the smoothing process reduces the forecasting error.

The currently burst of the Internet of Things (IoT) tech-nologies implies the emergence of new lines of investigation regarding not only to hardware and protocols but also to new methods of pro-duced data analysis satisfying the IoT environment constraints: a real-time and a big data approach. The Real-time restriction is about the continuous generation of data provided by the endpoints connected to an IoT network; due to the connection and scaling capabilities of an IoT network, the amount of data to process is so high that Big data tech-niques become essential. In this article, we present a system consisting of two main modules. In one hand, the infrastructure, a complete LoRa based network designed, tested and deployment in the Pablo de Olavide University and, on the other side, the analytics, a big data streaming sys-tem that processes the inputs produced by the network to obtain useful, valid and hidden information.

This work describes how an internal quality assurance sys tem is deployed at Pablo de Olavide University of Seville, Spain, in order to follow up all the existing degrees among the faculties and schools, seven centers in total, and how the teaching-learning process is improved. In the first place, the quality management structure existing in all the centers and degrees of the university is described. Additionally, all the actions related to the quality and improvement of the degrees of a cen ter are reported. Unlike in other Universities, in the Pablo de Olavide University there is no specific procedure for monitoring degrees, but the strategic procedure PE04: Measurement, analysis and improvement of the Internal Quality Assurance System is used to carry out such a proce dure. Therefore, the procedure is detailed specifying the different phases it consists of and those responsible for each of them. Once this procedure has been implemented, the centers have a follow-up report for each of their degrees, which also includes an improvement plan to be developed during the next course. The case of the degree of Computer Science in Information Systems, included in the School of Engineering, is analyzed over time in order to show how the implementation of such a system improves the overall performance of students

The main objective of this paper is the application of big data analytics to a real case in the field of smart electric networks. Smart meters are not only elements to measure consumption, but they also con stitute a network of millions of sensors in the electricity network. These sensors provide a huge amount of data that, once analyzed, can lead to significant advances for the society. In this way, tools are being developed in order to reach certain goals, such as obtaining a better consumption estimation (which would imply a better production planning), finding better rates based on the time discrimination or the contracted power, or minimizing the non-technical losses in the network, whose actual costs are eventually paid by end-consumers, among others. In this work, real data from Spanish consumers have been analyzed to detect fraud in con sumption. First, 1 TB of raw data was preprocessed in a HDFS-Spark infrastructure. Second, data duplication and outliers were removed, and missing values handled with specific big data algorithms. Third, cus tomers were characterized by means of clustering techniques in different scenarios. Finally, several key factors in fraud consumption were found. Very promising results were achieved, verging on 80% accuracy

New technologies such as sensor networks have been incorporated into the management of buildings for organizations and cities. Sensor networks have led to an exponential increase in the volume of data available in recent years, which can be used to extract consumption patterns for the purposes of energy and monetary savings. For this reason, new approaches and strategies are needed to analyze information in big data environments. This paper proposes a methodology to extract electric energy consumption patterns in big data time series, so that very valuable conclusions can be made for managers and governments. The methodology is based on the study of four clustering validity indices in their parallelized versions along with the application of a clustering technique. In particular, this work uses a voting system to choose an optimal number of clusters from the results of the indices, as well as the application of the distributed version of the k-means algorithm included in Apache Spark’s Machine Learning Library. The results, using electricity consumption for the years 2011–2017 for eight buildings of a public university, are presented and discussed. In addition, the performance of the proposed methodology is evaluated using synthetic big data, which cab represent thousands of buildings in a smart city. Finally, policies derived from the patterns discovered are proposed to optimize energy usage across the university campus.

This editorial summarizes the performance of the special issue entitled Data Science and Big Data in Energy Forecasting, which was published at MDPI’s Energies journal. The special issue took place in 2017 and accepted a total of 13 papers from 7 different countries. Electrical, solar and wind energy forecasting were the most analyzed topics, introducing new methods with applications of utmost relevance.

Background: Biclustering algorithms search for groups of genes that share the same behavior under a subset of samples in gene expression data. Nowadays, the biological knowledge available in public repositories can be used to drive these algorithms to find biclusters composed of groups of genes functionally coherent. On the other hand, a distance among genes can be defined according to their information stored in Gene Ontology (GO). Gene pairwise GO semantic similarity measures report a value for each pair of genes which establishes their functional similarity. A scatter search-based algorithm that optimizes a merit function that integrates GO information is studied in this paper. This merit function uses a term that addresses the information through a GO measure. Results: The effect of two possible different gene pairwise GO measures on the performance of the algorithm is analyzed. Firstly, three well known yeast datasets with approximately one thousand of genes are studied. Secondly, a group of human datasets related to clinical data of cancer is also explored by the algorithm. Most of these data are high-dimensional datasets composed of a huge number of genes. The resultant biclusters reveal groups of genes linked by a same functionality when the search procedure is driven by one of the proposed GO measures. Furthermore, a qualitative biological study of a group of biclusters show their relevance from a cancer disease perspective. Conclusions: It can be concluded that the integration of biological information improves the performance of the biclustering process. The two different GO measures studied show an improvement in the results obtained for the yeast dataset. However, if datasets are composed of a huge number of genes, only one of them really improves the algorithm performance. This second case constitutes a clear option to explore interesting datasets from a clinical point of view.

The continuous evaluation allows for the assessment of the progressive assimilation of concepts and the competences that must be achieved in a course. There are several ways to implement such contin uous evaluation system. We propose auto-evaluation tests as a valuable tool for the student to judge his level of knowledge. Furthermore, these tests are also used as a small part of the continuous evaluation process, encouraging students to learn the concepts seen in the course, as they have the feeling that the time dedicated to this study will have an assured reward, binge able to answer correctly the questions in the continuous evaluation exams. New technologies are a great aid to improve the auto evaluation experience both for the students and the teachers. In this research work we have compared the results obtained in courses where auto-evaluation tests were provided against courses where they were not provided, showing how the tests improve a set of quality metrics in the results of the course

This editorial summarizes the performance of the special issue entitled Energy Time Series Forecasting, which was published in MDPI’s Energies journal. The special issue took place in 2016 and accepted a total of 21 papers from twelve different countries. Electrical, solar, or wind energy forecasting were the most analyzed topics, introducing brand new methods with very sound results.

There exist several works in the literature in which fitness functions based on a combination of weighted measures for the discovery of association rules have been proposed. Nevertheless, some differences in the measures used to assess the quality of association rules could be obtained according to the values of the weights of the measures included in the fitness function. Therefore, user's decision is very important in order to specify the weights of the measures involved in the optimization process. This paper presents a study of well-known quality measures with regard to the weights of the measures that appear in a fitness function. In particular, the fitness function of an existing evolutionary algorithm called QARGA has been considered with the purpose of suggesting the values that should be assigned to the weights, depending on the set of measures to be optimized. As initial step, several experiments have been carried out from 35 public datasets in order to show how the weights for confidence, support, amplitude and number of attributes measures included in the fitness function have an influence on different quality measures according to several minimum support thresholds. Second, statistical tests have been conducted for evaluating when the differences in measures of the rules obtained by QARGA are significative, and thus, to provide the best weights to be considered depending on the group of measures to be optimized. Finally, the results obtained when using the recommended weights for two real-world applications related to ozone and earthquakes are reported.

Biclustering is an unsupervised machine learning technique that simultaneously clusters genes and conditions in gene expression data. Gene Ontology (GO) is usually used in this context to validate the biological relevance of the results. However, although the integration of biological information from different sources is one of the research directions in Bioinformatics, GO is not used in biclustering as an input data. A scatter search-based algorithm that integrates GO information during the biclustering search process is presented in this paper. SimUI is a GO semantic similarity measure that defines a distance between two genes. The algorithm optimizes a fitness function that uses SimUI to integrate the biological information stored in GO. Experimental results analyze the effect of integration of the biological information through this measure. A SimUI fitness function configuration is experimentally studied in a scatter search-based biclustering algorithm

A forecasting algorithm for big data time series is presented in this work. A nearest neighbours-based strategy is adopted as the main core of the algorithm. A detailed explanation on how to adapt and imple ment the algorithm to handle big data is provided. Although some parts remain iterative, and consequently requires an enhanced implementation, execution times are considered as satisfactory. The performance of the proposed approach has been tested on real-world data related to elec tricity consumption from a public Spanish university, by using a Spark cluster.

Los datos de expresión génica, y su particular naturaleza e importancia, motivan no sólo el desarrollo de nuevas técnicas sino la formulación de nuevos problemas como el problema del biclustering. El biclustering es una técnica de aprendizaje no supervisado que agrupa tanto genes como condiciones. Este doble agrupamiento lo diferencia del clustering tradicional sobre este tipo de datos ya que éste sólo agrupa o bien genes o condiciones. La presente tesis presenta un nuevo algoritmo de biclustering que permite el estudio de distintos criterios de búsqueda. Dicho algoritmo utiliza esquema de búsqueda dispersa, o scatter search, que independiza el mecanismo de búsqueda del criterio empleado. Se han estudiado tres criterios de búsqueda diferentes que motivan las tres principales aportaciones de la tesis. En primer lugar se estudia la correlación lineal entre los genes, que se integra como parte de la función objetivo empleada por el algoritmo de biclustering. La correlación lineal permite encontrar biclusters con patrones de desplazamiento y escalado, lo que mejora propuestas anteriores. En segundo lugar, y motivado por el significado biológico de los patrones de activación-inhibición entre genes, se modifica la correlación lineal de manera que se contemplen estos patrones. Por último, se ha tenido en cuenta la información disponible sobre genes en repositorios públicos, como la ontología de genes GO, y se incorpora dicha información como parte del criterio de búsqueda. Se añade un término extra que refleja, por cada bicluster que se evalúe, la calidad de ese grupo de genes según su información almacenada en GO. Se estudian dos posibilidades para dicho término de integración de información biológica, se comparan entre sí y se comprueba que los resultados son mejores cuando se usa información biológica en el algoritmo de biclustering. Las tres aportaciones descritas, junto con una serie de pasos intermedios, han dado lugar a resultados publicados tanto en revistas como e