Neural Networks in Bankruptcy Prediction - A Comparative Study on the Basis of the First Hungarian Bankruptcy Model

The article attempts to answer the question whether or not the latest bankruptcy prediction techniques are more reliable than traditional mathematical–statistical ones in Hungary. Simulation experiments carried out on the database of the first Hungarian bankruptcy prediction model clearly prove that bankruptcy models built using artificial neural networks have higher classification accuracy than models created in the 1990s based on discriminant analysis and logistic regression analysis. The article presents the main results, analyses the reasons for the differences and presents constructive proposals concerning the further development of Hungarian bankruptcy prediction.

Application of support vector machines on the basis of the first Hungarian bankruptcy model

In our study we rely on a data mining procedure known as support vector machine (SVM) on the database of the first Hungarian bankruptcy model. The models constructed are then contrasted with the results of earlier bankruptcy models with the use of classification accuracy and the area under the ROC curve. In using the SVM technique, in addition to conventional kernel functions, we also examine the possibilities of applying the ANOVA kernel function and take a detailed look at data preparation tasks recommended in using the SVM method (handling of outliers). The results of the models assembled suggest that a significant improvement of classification accuracy can be achieved on the database of the first Hungarian bankruptcy model when using the SVM method as opposed to neural networks.

Signal processing of physiological signals for automated assessment of stress in computer users

The need to provide computers with the ability to distinguish the affective state of their users is a major requirement for the practical implementation of affective computing concepts. This dissertation proposes the application of signal processing methods on physiological signals to extract from them features that can be processed by learning pattern recognition systems to provide cues about a person's affective state. In particular, combining physiological information sensed from a user's left hand in a non-invasive way with the pupil diameter information from an eye-tracking system may provide a computer with an awareness of its user's affective responses in the course of human-computer interactions. In this study an integrated hardware-software setup was developed to achieve automatic assessment of the affective status of a computer user. A computer-based "Paced Stroop Test" was designed as a stimulus to elicit emotional stress in the subject during the experiment. Four signals: the Galvanic Skin Response (GSR), the Blood Volume Pulse (BVP), the Skin Temperature (ST) and the Pupil Diameter (PD), were monitored and analyzed to differentiate affective states in the user. Several signal processing techniques were applied on the collected signals to extract their most relevant features. These features were analyzed with learning classification systems, to accomplish the affective state identification. Three learning algorithms: Naïve Bayes, Decision Tree and Support Vector Machine were applied to this identification process and their levels of classification accuracy were compared. The results achieved indicate that the physiological signals monitored do, in fact, have a strong correlation with the changes in the emotional states of the experimental subjects. These results also revealed that the inclusion of pupil diameter information significantly improved the performance of the emotion recognition system. ^

Behavioral Ecology of the Western Atlantic Short-finned Pilot Whale (Globicephala macrorhynchus)

Social structure is a key determinant of population biology and is central to the way animals exploit their environment. The risk of predation is often invoked as an important factor influencing the evolution of social structure in cetaceans and other mammals, but little direct information is available about how cetaceans actually respond to predators or other perceived threats. The playback of sounds to an animal is a powerful tool for assessing behavioral responses to predators, but quantifying behavioral responses to playback experiments requires baseline knowledge of normal behavioral patterns and variation. The central goal of my dissertation is to describe baseline foraging behavior for the western Atlantic short-finnned pilot whales (Globicephala macrohynchus) and examine the role of social organization in their response to predators. To accomplish this I used multi-sensor digital acoustic tags (DTAGs), satellite-linked time-depth recorders (SLTDR), and playback experiments to study foraging behavior and behavioral response to predators in pilot whales. Fine scale foraging strategies and population level patterns were identified by estimating the body size and examining the location and movement around feeding events using data collected with DTAGs deployed on 40 pilot whales in summers of 2008-2014 off the coast of Cape Hatteras, North Carolina. Pilot whales were found to forage throughout the water column and performed feeding buzzes at depths ranging from 29-1176 meters. The results indicated potential habitat segregation in foraging depth in short-finned pilot whales with larger individuals foraging on average at deeper depths. Calculated aerobic dive limit for large adult males was approximately 6 minutes longer than that of females and likely facilitated the difference in foraging depth. Furthermore, the buzz frequency and speed around feeding attempts indicate this population pilot whales are likely targeting multiple small prey items. Using these results, I built decision trees to inform foraging dive classification in coarse, long-term dive data collected with SLTDRs deployed on 6 pilot whales in the summers of 2014 and 2015 in the same area off the coast of North Carolina. I used these long term foraging records to compare diurnal foraging rates and depths, as well as classify bouts with a maximum likelihood method, and evaluate behavioral aerobic dive limits (ADLB) through examination of dive durations and inter-dive intervals. Dive duration was the best predictor of foraging, with dives >400.6 seconds classified as foraging, and a 96% classification accuracy. There were no diurnal patterns in foraging depth or rates and average duration of bouts was 2.94 hours with maximum bout durations lasting up to 14 hours. The results indicated that pilot whales forage in relatively long bouts and the ADLB indicate that pilot whales rarely, if ever exceed their aerobic limits. To evaluate the response to predators I used controlled playback experiments to examine the behavioral responses of 10 of the tagged short-finned pilot whales off Cape Hatteras, North Carolina and 4 Risso’s dolphins (Grampus griseus) off Southern California to the calls of mammal-eating killer whales (MEK). Both species responded to a subset of MEK calls with increased movement, swim speed and increased cohesion of the focal groups, but the two species exhibited different directional movement and vocal responses. Pilot whales increased their call rate and approached the sound source, but Risso’s dolphins exhibited no change in their vocal behavior and moved in a rapid, directed manner away from the source. Thus, at least to a sub-set of mammal-eating killer whale calls, these two study species reacted in a manner that is consistent with their patterns of social organization. Pilot whales, which live in relatively permanent groups bound by strong social bonds, responded in a manner that built on their high levels of social cohesion. In contrast, Risso’s dolphins exhibited an exaggerated flight response and moved rapidly away from the sound source. The fact that both species responded strongly to a select number of MEK calls, suggests that structural features of signals play critical contextual roles in the probability of response to potential threats in odontocete cetaceans.

Mapping of Submerged Aquatic Vegetation in Rivers From Very High Resolution Image Data, Using Object Based Image Analysis Combined with Expert Knowledge

The use of remote sensing for monitoring of submerged aquatic vegetation (SAV) in fluvial environments has been limited by the spatial and spectral resolution of available image data. The absorption of light in water also complicates the use of common image analysis methods. This paper presents the results of a study that uses very high resolution (VHR) image data, collected with a Near Infrared sensitive DSLR camera, to map the distribution of SAV species for three sites along the Desselse Nete, a lowland river in Flanders, Belgium. Plant species, including Ranunculus aquatilis L., Callitriche obtusangula Le Gall, Potamogeton natans L., Sparganium emersum L. and Potamogeton crispus L., were classified from the data using Object-Based Image Analysis (OBIA) and expert knowledge. A classification rule set based on a combination of both spectral and structural image variation (e.g. texture and shape) was developed for images from two sites. A comparison of the classifications with manually delineated ground truth maps resulted for both sites in 61% overall accuracy. Application of the rule set to a third validation image, resulted in 53% overall accuracy. These consistent results show promise for species level mapping in such biodiverse environments, but also prompt a discussion on assessment of classification accuracy.

Requirement Risk Level Forecast Using Bayesian Networks Classifiers

Requirement engineering is a key issue in the development of a software project. Like any other development activity it is not without risks. This work is about the empirical study of risks of requirements by applying machine learning techniques, specifically Bayesian networks classifiers. We have defined several models to predict the risk level for a given requirement using three dataset that collect metrics taken from the requirement specifications of different projects. The classification accuracy of the Bayesian models obtained is evaluated and compared using several classification performance measures. The results of the experiments show that the Bayesians networks allow obtaining valid predictors. Specifically, a tree augmented network structure shows a competitive experimental performance in all datasets. Besides, the relations established between the variables collected to determine the level of risk in a requirement, match with those set by requirement engineers. We show that Bayesian networks are valid tools for the automation of risks assessment in requirement engineering.

Montreal Cognitive Assessment (MoCA): validation study for vascular dementia

The Montreal Cognitive Assessment (MoCA) is a brief instrument developed for the screening of milder forms of cognitive impairment, having surpassed the well-known limitations of the MMSE. The aim of the present study was to validate the MoCA as well as its short version, which was proposed by the NINDS-CSN VCI Harmonization Standards for screening Vascular Dementia (VaD) patients. The results, based on a homogeneous sample of 34 VaD patients, indicate that the MoCA is a psychometrically valid and reliable instrument for cognitive screening in VaD patients, showing excellent discriminant validity. Both the full and short versions of the MoCA had excellent diagnostic accuracy in discriminating VaD patients, exhibiting an area under curve (AUC) higher than the MMSE [AUC(MoCA full version) = .950; 95% IC = .868-.988; AUC(MoCA short version) = .936; 95% IC = .849-.981; AUC(MMSE) = .860; 95% IC = .754-.932]. With a cutoff below 17 on the MoCA full version and 8 on the short version, the results for sensitivity, specificity, positive and negative predictive values, and classification accuracy were superior compared to the MMSE. In conclusion, both versions of the MoCA are valid, reliable, sensitive and accurate screening instruments for VaD patients.

Understanding and validating accelerometry as a measure of physical activity in children with intellectual disabilities

Background Physical activity in children with intellectual disabilities is a neglected area of study, which is most apparent in relation to physical activity measurement research. Although objective measures, specifically accelerometers, are widely used in research involving children with intellectual disabilities, existing research is based on measurement methods and data interpretation techniques generalised from typically developing children. However, due to physiological and biomechanical differences between these populations, questions have been raised in the existing literature on the validity of generalising data interpretation techniques from typically developing children to children with intellectual disabilities. Therefore, there is a need to conduct population-specific measurement research for children with intellectual disabilities and develop valid methods to interpret accelerometer data, which will increase our understanding of physical activity in this population. Methods Study 1: A systematic review was initially conducted to increase the knowledge base on how accelerometers were used within existing physical activity research involving children with intellectual disabilities and to identify important areas for future research. A systematic search strategy was used to identify relevant articles which used accelerometry-based monitors to quantify activity levels in ambulatory children with intellectual disabilities. Based on best practice guidelines, a novel form was developed to extract data based on 17 research components of accelerometer use. Accelerometer use in relation to best practice guidelines was calculated using percentage scores on a study-by-study and component-by-component basis. Study 2: To investigate the effect of data interpretation methods on the estimation of physical activity intensity in children with intellectual disabilities, a secondary data analysis was conducted. Nine existing sets of child-specific ActiGraph intensity cut points were applied to accelerometer data collected from 10 children with intellectual disabilities during an activity session. Four one-way repeated measures ANOVAs were used to examine differences in estimated time spent in sedentary, moderate, vigorous, and moderate to vigorous intensity activity. Post-hoc pairwise comparisons with Bonferroni adjustments were additionally used to identify where significant differences occurred. Study 3: The feasibility on a laboratory-based calibration protocol developed for typically developing children was investigated in children with intellectual disabilities. Specifically, the feasibility of activities, measurements, and recruitment was investigated. Five children with intellectual disabilities and five typically developing children participated in 14 treadmill-based and free-living activities. In addition, resting energy expenditure was measured and a treadmill-based graded exercise test was used to assess cardiorespiratory fitness. Breath-by-breath respiratory gas exchange and accelerometry were continually measured during all activities. Feasibility was assessed using observations, activity completion rates, and respiratory data. Study 4: Thirty-six children with intellectual disabilities participated in a semi-structured school-based physical activity session to calibrate accelerometry for the estimation of physical activity intensity. Participants wore a hip-mounted ActiGraph wGT3X+ accelerometer, with direct observation (SOFIT) used as the criterion measure. Receiver operating characteristic curve analyses were conducted to determine the optimal accelerometer cut points for sedentary, moderate, and vigorous intensity physical activity. Study 5: To cross-validate the calibrated cut points and compare classification accuracy with existing cut points developed in typically developing children, a sub-sample of 14 children with intellectual disabilities who participated in the school-based sessions, as described in Study 4, were included in this study. To examine the validity, classification agreement was investigated between the criterion measure of SOFIT and each set of cut points using sensitivity, specificity, total agreement, and Cohen’s kappa scores. Results Study 1: Ten full text articles were included in this review. The percentage of review criteria met ranged from 12%−47%. Various methods of accelerometer use were reported, with most use decisions not based on population-specific research. A lack of measurement research, specifically the calibration/validation of accelerometers for children with intellectual disabilities, is limiting the ability of researchers to make appropriate and valid accelerometer use decisions. Study 2: The choice of cut points had significant and clinically meaningful effects on the estimation of physical activity intensity and sedentary behaviour. For the 71-minute session, estimations for time spent in each intensity between cut points ranged from: sedentary = 9.50 (± 4.97) to 31.90 (± 6.77) minutes; moderate = 8.10 (± 4.07) to 40.40 (± 5.74) minutes; vigorous = 0.00 (± .00) to 17.40 (± 6.54) minutes; and moderate to vigorous = 8.80 (± 4.64) to 46.50 (± 6.02) minutes. Study 3: All typically developing participants and one participant with intellectual disabilities completed the protocol. No participant met the maximal criteria for the graded exercise test or attained a steady state during the resting measurements. Limitations were identified with the usability of respiratory gas exchange equipment and the validity of measurements. The school-based recruitment strategy was not effective, with a participation rate of 6%. Therefore, a laboratory-based calibration protocol was not feasible for children with intellectual disabilities. Study 4: The optimal vertical axis cut points (cpm) were ≤ 507 (sedentary), 1008−2300 (moderate), and ≥ 2301 (vigorous). Sensitivity scores ranged from 81−88%, specificity 81−85%, and AUC .87−.94. The optimal vector magnitude cut points (cpm) were ≤ 1863 (sedentary), ≥ 2610 (moderate) and ≥ 4215 (vigorous). Sensitivity scores ranged from 80−86%, specificity 77−82%, and AUC .86−.92. Therefore, the vertical axis cut points provide a higher level of accuracy in comparison to the vector magnitude cut points. Study 5: Substantial to excellent classification agreement was found for the calibrated cut points. The calibrated sedentary cut point (ĸ =.66) provided comparable classification agreement with existing cut points (ĸ =.55−.67). However, the existing moderate and vigorous cut points demonstrated low sensitivity (0.33−33.33% and 1.33−53.00%, respectively) and disproportionately high specificity (75.44−.98.12% and 94.61−100.00%, respectively), indicating that cut points developed in typically developing children are too high to accurately classify physical activity intensity in children with intellectual disabilities. Conclusions The studies reported in this thesis are the first to calibrate and validate accelerometry for the estimation of physical activity intensity in children with intellectual disabilities. In comparison with typically developing children, children with intellectual disabilities require lower cut points for the classification of moderate and vigorous intensity activity. Therefore, generalising existing cut points to children with intellectual disabilities will underestimate physical activity and introduce systematic measurement error, which could be a contributing factor to the low levels of physical activity reported for children with intellectual disabilities in previous research.

Integrated visual perception architecture for robotic clothes perception and manipulation

This thesis proposes a generic visual perception architecture for robotic clothes perception and manipulation. This proposed architecture is fully integrated with a stereo vision system and a dual-arm robot and is able to perform a number of autonomous laundering tasks. Clothes perception and manipulation is a novel research topic in robotics and has experienced rapid development in recent years. Compared to the task of perceiving and manipulating rigid objects, clothes perception and manipulation poses a greater challenge. This can be attributed to two reasons: firstly, deformable clothing requires precise (high-acuity) visual perception and dexterous manipulation; secondly, as clothing approximates a non-rigid 2-manifold in 3-space, that can adopt a quasi-infinite configuration space, the potential variability in the appearance of clothing items makes them difficult to understand, identify uniquely, and interact with by machine. From an applications perspective, and as part of EU CloPeMa project, the integrated visual perception architecture refines a pre-existing clothing manipulation pipeline by completing pre-wash clothes (category) sorting (using single-shot or interactive perception for garment categorisation and manipulation) and post-wash dual-arm flattening. To the best of the author’s knowledge, as investigated in this thesis, the autonomous clothing perception and manipulation solutions presented here were first proposed and reported by the author. All of the reported robot demonstrations in this work follow a perception-manipulation method- ology where visual and tactile feedback (in the form of surface wrinkledness captured by the high accuracy depth sensor i.e. CloPeMa stereo head or the predictive confidence modelled by Gaussian Processing) serve as the halting criteria in the flattening and sorting tasks, respectively. From scientific perspective, the proposed visual perception architecture addresses the above challenges by parsing and grouping 3D clothing configurations hierarchically from low-level curvatures, through mid-level surface shape representations (providing topological descriptions and 3D texture representations), to high-level semantic structures and statistical descriptions. A range of visual features such as Shape Index, Surface Topologies Analysis and Local Binary Patterns have been adapted within this work to parse clothing surfaces and textures and several novel features have been devised, including B-Spline Patches with Locality-Constrained Linear coding, and Topology Spatial Distance to describe and quantify generic landmarks (wrinkles and folds). The essence of this proposed architecture comprises 3D generic surface parsing and interpretation, which is critical to underpinning a number of laundering tasks and has the potential to be extended to other rigid and non-rigid object perception and manipulation tasks. The experimental results presented in this thesis demonstrate that: firstly, the proposed grasp- ing approach achieves on-average 84.7% accuracy; secondly, the proposed flattening approach is able to flatten towels, t-shirts and pants (shorts) within 9 iterations on-average; thirdly, the proposed clothes recognition pipeline can recognise clothes categories from highly wrinkled configurations and advances the state-of-the-art by 36% in terms of classification accuracy, achieving an 83.2% true-positive classification rate when discriminating between five categories of clothes; finally the Gaussian Process based interactive perception approach exhibits a substantial improvement over single-shot perception. Accordingly, this thesis has advanced the state-of-the-art of robot clothes perception and manipulation.

A Fuzzy Decision Strategy for Topic Identification and Dynamic Selection of Language Models

International audience

A GEOBIA methodology for fragmented agricultural landscapes

Very high resolution remotely sensed images are an important tool for monitoring fragmented agricultural landscapes, which allows farmers and policy makers to make better decisions regarding management practices. An object-based methodology is proposed for automatic generation of thematic maps of the available classes in the scene, which combines edge-based and superpixel processing for small agricultural parcels. The methodology employs superpixels instead of pixels as minimal processing units, and provides a link between them and meaningful objects (obtained by the edge-based method) in order to facilitate the analysis of parcels. Performance analysis on a scene dominated by agricultural small parcels indicates that the combination of both superpixel and edge-based methods achieves a classification accuracy slightly better than when those methods are performed separately and comparable to the accuracy of traditional object-based analysis, with automatic approach.

Pattern recognition methods for EMG prosthetic control

In this work we focus on pattern recognition methods related to EMG upper-limb prosthetic control. After giving a detailed review of the most widely used classification methods, we propose a new classification approach. It comes as a result of comparison in the Fourier analysis between able-bodied and trans-radial amputee subjects. We thus suggest a different classification method which considers each surface electrodes contribute separately, together with five time domain features, obtaining an average classification accuracy equals to 75% on a sample of trans-radial amputees. We propose an automatic feature selection procedure as a minimization problem in order to improve the method and its robustness.

The sensitivity of mapping methods to reference data quality:training supervised image classifications with imperfect reference data

The accuracy of a map is dependent on the reference dataset used in its construction. Classification analyses used in thematic mapping can, for example, be sensitive to a range of sampling and data quality concerns. With particular focus on the latter, the effects of reference data quality on land cover classifications from airborne thematic mapper data are explored. Variations in sampling intensity and effort are highlighted in a dataset that is widely used in mapping and modelling studies; these may need accounting for in analyses. The quality of the labelling in the reference dataset was also a key variable influencing mapping accuracy. Accuracy varied with the amount and nature of mislabelled training cases with the nature of the effects varying between classifiers. The largest impacts on accuracy occurred when mislabelling involved confusion between similar classes. Accuracy was also typically negatively related to the magnitude of mislabelled cases and the support vector machine (SVM), which has been claimed to be relatively insensitive to training data error, was the most sensitive of the set of classifiers investigated, with overall classification accuracy declining by 8% (significant at 95% level of confidence) with the use of a training set containing 20% mislabelled cases.

Towards person-centered neuroimaging markers for resilience and vulnerability in Bipolar Disorder

Improved clinical care for Bipolar Disorder (BD) relies on the identification of diagnostic markers that can reliably detect disease-related signals in clinically heterogeneous populations. At the very least, diagnostic markers should be able to differentiate patients with BD from healthy individuals and from individuals at familial risk for BD who either remain well or develop other psychopathology, most commonly Major Depressive Disorder (MDD). These issues are particularly pertinent to the development of translational applications of neuroimaging as they represent challenges for which clinical observation alone is insufficient. We therefore applied pattern classification to task-based functional magnetic resonance imaging (fMRI) data of the n-back working memory task, to test their predictive value in differentiating patients with BD (n=30) from healthy individuals (n=30) and from patients' relatives who were either diagnosed with MDD (n=30) or were free of any personal lifetime history of psychopathology (n=30). Diagnostic stability in these groups was confirmed with 4-year prospective follow-up. Task-based activation patterns from the fMRI data were analyzed with Gaussian Process Classifiers (GPC), a machine learning approach to detecting multivariate patterns in neuroimaging datasets. Consistent significant classification results were only obtained using data from the 3-back versus 0-back contrast. Using contrast, patients with BD were correctly classified compared to unrelated healthy individuals with an accuracy of 83.5%, sensitivity of 84.6% and specificity of 92.3%. Classification accuracy, sensitivity and specificity when comparing patients with BD to their relatives with MDD, were respectively 73.1%, 53.9% and 94.5%. Classification accuracy, sensitivity and specificity when comparing patients with BD to their healthy relatives were respectively 81.8%, 72.7% and 90.9%. We show that significant individual classification can be achieved using whole brain pattern analysis of task-based working memory fMRI data. The high accuracy and specificity achieved by all three classifiers suggest that multivariate pattern recognition analyses can aid clinicians in the clinical care of BD in situations of true clinical uncertainty regarding the diagnosis and prognosis.

Insights into population structure of Diplodus vulgaris along the SW Portuguese coast from otolith elemental signatures

The common two-banded sea bream (Diplodus vulgaris) is an important fish in the marine ecosystems of the NW Atlantic and Mediterranean. In southern Portugal it is a major fishery resource being targeted mainly by the artisanal fleets. Although there is some knowledge of the age, growth and reproductive biology of the species, information about its population structure is scarce and somewhat limited to the Mediterranean Sea. In this study the otolith elemental signatures of 90 specimens of D. vulgaris of the same age group (2+ years) and cohort collected from the important fishery regions of SW Portugal (Sesimbra, Sagres and Faro) have been analysed by inductively coupled plasma mass spectrometry (ICP-MS). Two different methodologies have been applied: solution based analysis of the whole otoliths; representative of the entire life-history prior to capture, and laser ablation analysis of otolith cores; representative of the larval and early post-settlement phase. Whole otolith comparisons utilised Sr/Ca, Ba/Ca, Mn/Ca, Li/Ca and Ni/Ca to demonstrate regional population structure. Classification accuracy rates from linear discriminant function analyses (LDFA) of whole otolith chemistry data were high for each region; Faro - 93%, Sagres - 90% and Sesimbra - 80%. Comparison of the otolith core chemistry utilised Sr/Ca, Ba/Ca, Mn/Ca and Mg/Ca and Zn/Ca. LDFA for the otolith core chemistry also achieved accurate classification for samples from Sesimbra (73%), but there was high overlap of otolith chemistry between samples from Faro and Sagres (47 and 43% classification accuracy respectively). The whole otolith results suggest that D. vulgaris are resident in the regional fishing areas during the juvenile phase. Both the core and whole otolith chemistry data supported separation of the Sesimbra fishery region from the more southern and closely associated Faro and Sagres regions for management purposes. However, while the whole otolith data indicated that the populations at Faro and Sagres likely remained separated in the juvenile stage, the otolith core chemistry data was inconclusive as to whether recruitment to these two areas was derived, or not, from different spawning areas.