Showing papers by "Wrocław University of Technology published in 2016"

Learning from imbalanced data: open challenges and future directions

Abstract: Despite more than two decades of continuous development learning from imbalanced data is still a focus of intense research. Starting as a problem of skewed distributions of binary tasks, this topic evolved way beyond this conception. With the expansion of machine learning and data mining, combined with the arrival of big data era, we have gained a deeper insight into the nature of imbalanced learning, while at the same time facing new emerging challenges. Data-level and algorithm-level methods are constantly being improved and hybrid approaches gain increasing popularity. Recent trends focus on analyzing not only the disproportion between classes, but also other difficulties embedded in the nature of data. New real-life problems motivate researchers to focus on computationally efficient, adaptive and real-time methods. This paper aims at discussing open issues and challenges that need to be addressed to further develop the field of imbalanced learning. Seven vital areas of research in this topic are identified, covering the full spectrum of learning from imbalanced data: classification, regression, clustering, data streams, big data analytics and applications, e.g., in social media and computer vision. This paper provides a discussion and suggestions concerning lines of future research for each of them.

Ensemble boosted trees with synthetic features generation in application to bankruptcy prediction

Abstract: We propose a novel ensemble model for bankruptcy prediction.We use Extreme Gradient Boosting as an ensemble of decision trees.We propose a new approach for generating synthetic features to improve prediction.The presented method is evaluated on real-life data of Polish companies. Bankruptcy prediction has been a subject of interests for almost a century and it still ranks high among hottest topics in economics. The aim of predicting financial distress is to develop a predictive model that combines various econometric measures and allows to foresee a financial condition of a firm. In this domain various methods were proposed that were based on statistical hypothesis testing, statistical modeling (e.g., generalized linear models), and recently artificial intelligence (e.g., neural networks, Support Vector Machines, decision tress). In this paper, we propose a novel approach for bankruptcy prediction that utilizes Extreme Gradient Boosting for learning an ensemble of decision trees. Additionally, in order to reflect higher-order statistics in data and impose a prior knowledge about data representation, we introduce a new concept that we refer as to synthetic features. A synthetic feature is a combination of the econometric measures using arithmetic operations (addition, subtraction, multiplication, division). Each synthetic feature can be seen as a single regression model that is developed in an evolutionary manner. We evaluate our solution using the collected data about Polish companies in five tasks corresponding to the bankruptcy prediction in the 1st, 2nd, 3rd, 4th, and 5th year. We compare our approach with the reference methods.

Evolutionary undersampling boosting for imbalanced classification of breast cancer malignancy

Abstract: Graphical abstractDisplay Omitted HighlightsAutomatic clinical decision support system for breast cancer malignancy grading.Different methodologies for segmentation and feature extraction from FNA slides.An efficient classifier ensemble for imbalanced problems with difficult data.Ensemble combines boosting with evolutionary undersampling.Extensive computational experiments on a large database collected by authors. In this paper, we propose a complete, fully automatic and efficient clinical decision support system for breast cancer malignancy grading. The estimation of the level of a cancer malignancy is important to assess the degree of its progress and to elaborate a personalized therapy. Our system makes use of both Image Processing and Machine Learning techniques to perform the analysis of biopsy slides. Three different image segmentation methods (fuzzy c-means color segmentation, level set active contours technique and grey-level quantization method) are considered to extract the features used by the proposed classification system. In this classification problem, the highest malignancy grade is the most important to be detected early even though it occurs in the lowest number of cases, and hence the malignancy grading is an imbalanced classification problem. In order to overcome this difficulty, we propose the usage of an efficient ensemble classifier named EUSBoost, which combines a boosting scheme with evolutionary undersampling for producing balanced training sets for each one of the base classifiers in the final ensemble. The usage of the evolutionary approach allows us to select the most significant samples for the classifier learning step (in terms of accuracy and a new diversity term included in the fitness function), thus alleviating the problems produced by the imbalanced scenario in a guided and effective way. Experiments, carried on a large dataset collected by the authors, confirm the high efficiency of the proposed system, shows that level set active contours technique leads to an extraction of features with the highest discriminative power, and prove that EUSBoost is able to outperform state-of-the-art ensemble classifiers in a real-life imbalanced medical problem.

A review of potential innovations for production, conditioning and utilization of biogas with multiple criteria assessment

Abstract: Biogas is a relatively mature renewable energy technology but still most commercial biogas power plants require significant financial incentives. Additionally, local shortages of very cheap digestible feedstocks limit biogas productivity, especially for larger biogas power plants (>1 MWe). Innovations that could improve cost-effectiveness and resource efficiency of biogas energy technology are therefore required. Over the last few years a number of potential process innovations for biogas technology have been proposed and investigated. However, the majority of these novel concepts has minimal or no impact on technology development. Disruptive innovations are very rare, but only they really matter for the economy. Therefore review reports that systematically compare, analyze and evaluate the suitability of these emerging methods with emphasis on technological excellence and realistic commercial potential are needed. This study presents potential process innovations from most recent patent and academic literature proposed for biogas (i) production, (ii) conditioning, (iii) utilization and (iv) industrial symbiosis. Within these four highly interdisciplinary categories the review attempts to provide short practical comments on selected methods and briefly analyzes their perspectives and constraints. Further, relevant biogas process innovation criteria are designed and multiple-criteria assessment of pre-selected potential process innovations is made. The paper concludes with the characterization of innovativeness of selected solutions and suggests future research needs for biogas energy technology that could bring new innovations in near term.

Pee power urinal – microbial fuel cell technology field trials in the context of sanitation

Abstract: This paper reports on the pee power urinal field trials, which are using microbial fuel cells for internal lighting. The first trial was conducted on Frenchay Campus (UWE, Bristol) from February–May 2015 and demonstrated the feasibility of modular MFCs for lighting, with University staff and students as the users; the next phase of this trial is ongoing. The second trial was carried out during the Glastonbury Music Festival at Worthy Farm, Pilton in June 2015, and demonstrated the capability of the MFCs to reliably generate power for internal lighting, from a large festival audience (∼1000 users per day). The power output recorded for individual MFCs is 1–2 mW, and the power output of one 36-MFC-module, was commensurate of this level of power. Similarly, the real-time electrical output of both the pee power urinals was proportional to the number of MFCs used, subject to temperature and flow rate: the campus urinal consisted of 288 MFCs, generating 75 mW (mean), 160 mW (max) with 400 mW when the lights were connected directly (no supercapacitors); the Glastonbury urinal consisted of 432 MFCs, generating 300 mW (mean), 400 mW (max) with 800 mW when the lights were connected directly (no supercapacitors). The COD removal was >95% for the campus urinal and on average 30% for the Glastonbury urinal. The variance in both power and urine treatment was due to environmental conditions such as temperature and number of users. This is the first time that urinal field trials have demonstrated the feasibility of MFCs for both electricity generation and direct urine treatment. In the context of sanitation and public health, an independent power source utilising waste is essential in terms of both developing and developed world.

Physical meaning of the Sauter mean diameter of spherical particulate matter

Abstract: Diameter of a collection of different-size spherical objects such as particles, droplets, and bubbles can be represented by a single value called the mean diameter. There are many ways of calculating and expressing the mean diameter of particulate matter. One of them is the Sauter mean diameter also called the surface-volume mean diameter. This diameter has been extensively used in numerous publications without paying attention to its physical meaning. In this article, the physical meaning of the Sauter mean diameter of spherical objects was derived and presented. The Sauter mean diameter of a collection of spherical objects of different sizes is equal to the diameter of equisized spherical objects forming a collection. The polysized and equisized systems have different numbers of spherical objects, identical total surface area, and identical total volume. If the surface energy of all spheres of both systems is the same, they can be called equienergetic.

Comprehensive Study on Ceramic Membranes for Low‐Cost Microbial Fuel Cells

Abstract: © 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim. Microbial fuel cells (MFCs) made with different types of ceramic membranes were investigated to find a low-cost alternative to commercially available proton exchange membranes. The MFCs operated with fresh human urine as the fuel. Pyrophyllite and earthenware produced the best performance to reach power densities of 6.93 and 6.85 W m-3, respectively, whereas mullite and alumina achieved power densities of 4.98 and 2.60 Wm-3, respectively. The results indicate the dependence of bio-film growth and activity on the type of ceramic membrane applied. The most favourable conditions were created in earthenware MFCs. The performance of the ceramic membranes was related to their physical and chemical properties determined by environmental scanning electron microscopy and energy dispersive X-ray spectroscopy. The cost of mullite, earthenware, pyrophyllite and alumina was estimated to be 13.61, 4.14, 387.96 and 177.03 GBP m-2, respectively. The results indicate that earthenware and mullite are good substitutes for commercially available proton exchange membranes, which makes the MFC technology accessible in developing countries.

Flowing Liquid Anode Atmospheric Pressure Glow Discharge as an Excitation Source for Optical Emission Spectrometry with the Improved Detectability of Ag, Cd, Hg, Pb, Tl, and Zn.

Abstract: A novel atmospheric pressure glow discharge generated in contact with a flowing liquid anode (FLA-APGD) was developed as the efficient excitation source for the optical emission spectrometry (OES) detection. Differences in the appearance and the electrical characteristic of the FLA-APGD and a conventional system operated with a flowing liquid cathode (FLC-APGD) were studied in detail and discussed. Under the optimal operating conditions for the FLA-APGD, the emission from the analytes (Ag, Cd, Hg, Pb, Tl, and Zn) was from 20 to 120 times higher as compared to the FLC-APGD. Limits of detections (LODs) established with a novel FLA-APGD system were on average 20 times better than those obtained for the FLC-APGD. A further improvement of the LODs was achieved by reducing the background shift interferences and, as a result, the LODs for Ag, Cd, Hg, Pb, Tl, and Zn were 0.004, 0.040, 0.70, 1.7, 0.035, and 0.45 μg L–1, respectively. The precision of the FLA-APGD-OES method was evaluated to be within 2–5% (as the ...

Experimental and theoretical studies of structural phase transition in a novel polar perovskite-like [C2H5NH3][Na0.5Fe0.5(HCOO)3] formate.

Abstract: We report the synthesis, single crystal X-ray diffraction, and thermal, dielectric, Raman and infrared studies of a novel heterometallic formate [C2H5NH3][Na0.5Fe0.5(HCOO)3] (EtANaFe). The thermal studies show that EtANaFe undergoes a second-order phase transition at about 360 K. X-ray diffraction data revealed that the high-temperature structure is monoclinic, space group P2(1)/n, with dynamically disordered ethylammonium (EtA(+)) cations. EtANaFe possesses a polar low-temperature structure with the space group Pn and, in principle, is ferroelectric below 360 K. Dielectric data show that the reciprocal of the real part of dielectric permittivity above and below the phase transition temperature follows the Curie-Weiss, as expected for a ferroelectric phase transition. Based on theoretical calculations, we estimated the polarization as (0.2, 0, 0.8) μC cm(-2), i.e., lying within the ac plane. The obtained data also indicate that the driving force of the phase transition is ordering of EtA(+) cations. However, this ordering is accompanied by significant distortion of the metal formate framework.

Robust Discrete Optimization Under Discrete and Interval Uncertainty: A Survey

Abstract: In this chapter a review of recent results on robust discrete optimization is presented. The most popular discrete and interval uncertainty representations are discussed. Various robust concepts are presented, namely the traditional minmax (regret) approach with some of its recent extensions, and several two-stage concepts. A special attention is paid to the computational properties of the robust problems considered.

Application of Polymer Based Composite Materials in Transportation

Abstract: The purpose of this paper is to present the latest applications of fibre-reinforced polymer matrix composites described on exemplary applications. It contains the description of various examples, w...

Empowering one-vs-one decomposition with ensemble learning for multi-class imbalanced data

Abstract: Extending binary ensemble techniques to multi-class imbalanced data.OVO scheme enhancement for multi-class imbalanced data by ensemble learning.A complete experimental study of comparison of the ensemble learning techniques with OVO.Study of the impact of base classifiers used in the proposed scenario. Multi-class imbalance classification problems occur in many real-world applications, which suffer from the quite different distribution of classes. Decomposition strategies are well-known techniques to address the classification problems involving multiple classes. Among them binary approaches using one-vs-one and one-vs-all has gained a significant attention from the research community. They allow to divide multi-class problems into several easier-to-solve two-class sub-problems. In this study we develop an exhaustive empirical analysis to explore the possibility of empowering the one-vs-one scheme for multi-class imbalance classification problems with applying binary ensemble learning approaches. We examine several state-of-the-art ensemble learning methods proposed for addressing the imbalance problems to solve the pairwise tasks derived from the multi-class data set. Then the aggregation strategy is employed to combine the binary ensemble outputs to reconstruct the original multi-class task. We present a detailed experimental study of the proposed approach, supported by the statistical analysis. The results indicate the high effectiveness of ensemble learning with one-vs-one scheme in dealing with the multi-class imbalance classification problems.

Recent advances in electricity price forecasting: A review of probabilistic forecasting

Abstract: Since the inception of competitive power markets two decades ago, electricity price forecasting (EPF) has gradually become a fundamental process for energy companies' decision making mechanisms. Over the years, the bulk of research has concerned point predictions. However, the recent introduction of smart grids and renewable integration requirements has had the effect of increasing the uncertainty of future supply, demand and prices. Academics and practitioners alike have come to understand that probabilistic electricity price (and load) forecasting is now more important for energy systems planning and operations than ever before. With this paper we offer a tutorial review of probabilistic EPF and present much needed guidelines for the rigorous use of methods, measures and tests, in line with the paradigm of 'maximizing sharpness subject to reliability'. The paper can be treated as an update and a further extension of the otherwise comprehensive EPF review of Weron (2014, IJF) or as a standalone treatment of a fascinating and underdeveloped topic, that has a much broader reach than EPF itself.

Influence of PM1 and PM2.5 on lung function parameters in healthy schoolchildren-a panel study.

Abstract: To evaluate lung function responses to short-term indoor PM1 and PM2.5 concentrations, we conducted a panel study of healthy schoolchildren aged 13–14 years. The following lung function parameters FVC, FEV1, PEF, and mid expiratory flows MEF25, MEF50, and MEF75 were measured in 141 schoolchildren of the secondary school in Wroclaw, Poland in years 2009–2010. On days when spirometry tests were conducted, simultaneously, PM1 and PM2.5 samples were collected inside the school premises. Information about differentiating factors for children including smoking parents, sex, living close to busy streets, dust, mold, and pollen allergies were collected by means of questionnaires. To account for repeated measurements, the method of generalized estimating equations (GEE) was used. The GEE models for the entire group of children revealed the adverse effects (p < 0.05) of PM1 and PM2.5. Small differences in effects estimates per interquartile range (IQR) of PM1 and PM2.5 on MEF25 (5.1 and 4.8 %), MEF50 (3.7 and 3.9 %), MEF75 (3.5 and 3.6 %) and FEV1 (1.3 and 1.0 %) imply that PM1 was likely the component of PM2.5 that might have a principal health effect on these lung function parameters. However, the reduction of FVC and PEF per IQR for PM2.5 (2.1 and 5.2 %, respectively) was higher than for PM1 (1.0 and 4.4 %, respectively). Adjustment for potential confounders did not change the unadjusted analysis.

Ebselen, a Small-Molecule Capsid Inhibitor of HIV-1 Replication

Abstract: The human immunodeficiency virus type 1 (HIV-1) capsid plays crucial roles in HIV-1 replication and thus represents an excellent drug target. We developed a high-throughput screening method based on a time-resolved fluorescence resonance energy transfer (HTS-TR-FRET) assay, using the C-terminal domain (CTD) of HIV-1 capsid to identify inhibitors of capsid dimerization. This assay was used to screen a library of pharmacologically active compounds, composed of 1,280 in vivo-active drugs, and identified ebselen [2-phenyl-1,2-benzisoselenazol-3(2H)-one], an organoselenium compound, as an inhibitor of HIV-1 capsid CTD dimerization. Nuclear magnetic resonance (NMR) spectroscopic analysis confirmed the direct interaction of ebselen with the HIV-1 capsid CTD and dimer dissociation when ebselen is in 2-fold molar excess. Electrospray ionization mass spectrometry revealed that ebselen covalently binds the HIV-1 capsid CTD, likely via a selenylsulfide linkage with Cys198 and Cys218. This compound presents anti-HIV activity in single and multiple rounds of infection in permissive cell lines as well as in primary peripheral blood mononuclear cells. Ebselen inhibits early viral postentry events of the HIV-1 life cycle by impairing the incoming capsid uncoating process. This compound also blocks infection of other retroviruses, such as Moloney murine leukemia virus and simian immunodeficiency virus, but displays no inhibitory activity against hepatitis C and influenza viruses. This study reports the use of TR-FRET screening to successfully identify a novel capsid inhibitor, ebselen, validating HIV-1 capsid as a promising target for drug development.

Relationship between patients' knowledge and medication adherence among patients with hypertension.

Abstract: Objective The purpose of this study was to investigate the relationship between knowledge on arterial hypertension (AH) and its management, and adherence to pharmaceutical treatment. Methods The study included 233 patients diagnosed with AH and treated with hypotensive drugs for at least 1 year. The 8-item © Morisky Medication Adherence Scale (MMAS-8) and the Hypertension Knowledge-Level Scale (HK-LS) were used. Results Sixty-three percent of the patients had a low level of knowledge on AH, with the smallest proportion of correct answers provided for items related to non-pharmaceutical treatment, diet, hypertension definition, and drug adherence. When compared to patients with a high level of knowledge, those with a low knowledge had lower scores in the MMAS (6.45±1.45 vs 7.08±1.04; P=0.038). Multiple-factor analysis showed that statistically significant independent determinants of good adherence included a high level of knowledge (β=0.208; P=0.001), non-pharmaceutical treatment (β=0.182; P=0.006), and frequent blood pressure measurements (β=0.183; P=0.004). The most significant factor in MMAS was knowledge in the "drug adherence" domain (ρ=0.303; P Conclusion Patients' knowledge on hypertension is a significant independent determinant of good adherence. Other independent determinants include non-pharmaceutical treatment and regular blood pressure measurements. Implication for practice The identification of knowledge deficits as a factor contributing to lack of adherence and poor hypertension control remains a key challenge for multidisciplinary team caring for patients with hypertension.

Compact all-fiber figure-9 dissipative soliton resonance mode-locked double-clad Er:Yb laser.

Abstract: The first demonstration of a compact all-fiber figure-9 double-clad erbium-ytterbium laser working in the dissipative soliton resonance (DSR) regime is presented. Mode-locking was achieved using a nonlinear amplifying loop (NALM) resonator configuration. The laser was assembled with an additional 475 m long spool of SMF28 fiber in the NALM loop in order to obtain large net-anomalous cavity dispersion (-10.4 ps2), and therefore ensure that DSR would be the dominant mode-locking mechanism. At maximum pump power (4.78 W) the laser generated rectangular-shaped pulses with 455 ns duration and an average power of 950 mW, which at a repetition frequency of 412 kHz corresponds to a record energy of 2.3 μJ per pulse.

The failure mechanisms of hot forging dies

Abstract: This paper describes the phenomena taking place on the surface of the dies used for hot forging. Because of this paper’s limited space only changes in the tool surface layer during the forging of a gear wheel, as most representative, are presented. Similar changes were observed in the case of the other two investigated closed die forging processes, i.e. the forging of a cover and a yoke, respectively. The research was aided by FEM, which supplied a lot of information about the forging conditions. The most intensive wear of the tools occurs in the place of their longest contact with the material being forged, regardless of the number of produced forgings. The research has shown that the one of the most adversely factor affecting the investigated forging process is thermomechanical fatigue which results in fine cracks quickly developing into a network of cracks extending over the entire tool/forged material contact surface. Also the abrasive wear of the investigated die is high due to the intensive flow of the material in the presence of abrasive oxide particles and tools bits created by thermomechanical fatigue. An attempt to model the abrasive wear using the Archard model is presented.

Co/ZIF-8 Heterometallic Nanoparticles: Control of Nanocrystal Size and Properties by a Mixed-Metal Approach

Abstract: A mixed-metal approach has been used to control the size and physicochemical properties of heterometallic Co/ZIF-8 nanomaterials. Intentional substitution of zinc with cobalt in a broad concentration range (from 0 to 100 molar percent with a 10% step) provided a series of Co/ZIF-8 nanoparticles, whose sizes could be tuned in the range from 20 to over 500 nm in diameter. Zinc ions from the ZIF-8 matrix were found to be uniformly substituted with the cobalt ions. The increase of nanoparticles size resulted in a change of their nitrogen sorption–desorption characteristics due to decreasing participation of the external surface area in the total surface area. Insights from UV–vis-NIR and IR spectroscopies, as well as remarks on nonlinear optical properties are also provided.