On the impact of disproportional samples in credit scoring models: An application to a Brazilian bank data

Statistical methods have been widely employed to assess the capabilities of credit scoring classification models in order to reduce the risk of wrong decisions when granting credit facilities to clients. The predictive quality of a classification model can be evaluated based on measures such as sensitivity, specificity, predictive values, accuracy, correlation coefficients and information theoretical measures, such as relative entropy and mutual information. In this paper we analyze the performance of a naive logistic regression model (Hosmer & Lemeshow, 1989) and a logistic regression with state-dependent sample selection model (Cramer, 2004) applied to simulated data. Also, as a case study, the methodology is illustrated on a data set extracted from a Brazilian bank portfolio. Our simulation results so far revealed that there is no statistically significant difference in terms of predictive capacity between the naive logistic regression models and the logistic regression with state-dependent sample selection models. However, there is strong difference between the distributions of the estimated default probabilities from these two statistical modeling techniques, with the naive logistic regression models always underestimating such probabilities, particularly in the presence of balanced samples. (C) 2012 Elsevier Ltd. All rights reserved.

Inflammatory events during murine squamous cell carcinoma development

Background: Squamous cell carcinoma (SCC) is one of the most common human cancers worldwide. In SCC, tumour development is accompanied by an immune response that leads to massive tumour infiltration by inflammatory cells, and consequently, local and systemic production of cytokines, chemokines and other mediators. Studies in both humans and animal models indicate that imbalances in these inflammatory mediators are associated with cancer development. Methods: We used a multistage model of SCC to examine the involvement of elastase (ELA), myeloperoxidase (MPO), nitric oxide (NO), cytokines (IL-6, IL-10, IL-13, IL-17, TGF-β and TNF-α), and neutrophils and macrophages in tumour development. ELA and MPO activity and NO, IL-10, IL −17, TNF-α and TGF-β levels were increased in the precancerous microenvironment. Results: ELA and MPO activity and NO, IL-10, IL −17, TNF-α and TGF-β levels were increased in the precancerous microenvironment. Significantly higher levels of IL-6 and lower levels of IL-10 were detected at 4 weeks following 7,12-Dimethylbenz(a)anthracene (DMBA) treatment. Similar levels of IL-13 were detected in the precancerous microenvironment compared with control tissue. We identified significant increases in the number of GR-1+ neutrophils and F4/80+/GR-1- infiltrating cells in tissues at 4 and 8 weeks following treatment and a higher percentage of tumour-associated macrophages (TAM) expressing both GR-1 and F4/80, an activated phenotype, at 16 weeks. We found a significant correlation between levels of IL-10, IL-17, ELA, and activated TAMs and the lesions. Additionally, neutrophil infiltrate was positively correlated with MPO and NO levels in the lesions. Conclusion: Our results indicate an imbalance of inflammatory mediators in precancerous SCC caused by neutrophils and macrophages and culminating in pro-tumour local tissue alterations.

Probabilistic Envelope Curves for Extreme Rainfall Events - Curve Inviluppo Probabilistiche per Precipitazioni Estreme

A regional envelope curve (REC) of flood flows summarises the current bound on our experience of extreme floods in a region. RECs are available for most regions of the world. Recent scientific papers introduced a probabilistic interpretation of these curves and formulated an empirical estimator of the recurrence interval T associated with a REC, which, in principle, enables us to use RECs for design purposes in ungauged basins. The main aim of this work is twofold. First, it extends the REC concept to extreme rainstorm events by introducing the Depth-Duration Envelope Curves (DDEC), which are defined as the regional upper bound on all the record rainfall depths at present for various rainfall duration. Second, it adapts the probabilistic interpretation proposed for RECs to DDECs and it assesses the suitability of these curves for estimating the T-year rainfall event associated with a given duration and large T values. Probabilistic DDECs are complementary to regional frequency analysis of rainstorms and their utilization in combination with a suitable rainfall-runoff model can provide useful indications on the magnitude of extreme floods for gauged and ungauged basins. The study focuses on two different national datasets, the peak over threshold (POT) series of rainfall depths with duration 30 min., 1, 3, 9 and 24 hrs. obtained for 700 Austrian raingauges and the Annual Maximum Series (AMS) of rainfall depths with duration spanning from 5 min. to 24 hrs. collected at 220 raingauges located in northern-central Italy. The estimation of the recurrence interval of DDEC requires the quantification of the equivalent number of independent data which, in turn, is a function of the cross-correlation among sequences. While the quantification and modelling of intersite dependence is a straightforward task for AMS series, it may be cumbersome for POT series. This paper proposes a possible approach to address this problem.

Holocene tsunami events in the Eastern Ionian Sea - Geoscientific evidence from Cefalonia and the western Peloponnese (Greece)

This study presents geo-scientific evidence for Holocene tsunami impact along the shores of the Eastern Ionian Sea. Cefalonia Island, the Gulf of Kyparissia and the Gialova Lagoon were subject of detailed geo-scientific investigations. It is well known that the coasts of the eastern Mediterranean were hit by the destructive influence of tsunamis in the past. The seismically highly active Hellenic Trench is considered as the most significant tsunami source in the Eastern Ionian Sea. This study focuses on the reconstruction and detection of sedimentary signatures of palaeotsunami events and their influence on the Holocene palaeogeographical evolution. The results of fine grained near coast geo-archives are discussed and interpreted in detail to differentiate between tsunami, storm and sea level highstands as sedimentation processes.rnA multi-method approach was applied using geomorphological, sedimentological, geochemical, geophysical and microfaunal analyses to detect Holocene tsunamigenic impact. Chronological data were based on radiocarbondatings and archaeological age estimations to reconstruct local geo-chronostratigraphies and to correlate them on supra-regional scales.rnDistinct sedimentary signatures of 5 generations of tsunami impact were found along the coasts of Cefalonia in the Livadi coastal plain. The results show that the overall coastal evolution was influenced by tsunamigenic impact that occured around 5700 cal BC (I), 4250 cal BC (II), at the beginning of the 2nd millennium cal BC (III), in the 1st millennium cal BC (IV) and posterior to 780 cal AD (V). Sea level reconstructions and the palaeogeographical evolution show that the local Holocene sea level has never been higher than at present.rnAt the former Mouria Lagoon along the Gulf of Kyparissia almost four allochtonous layers of tsunamigenic origin were identified. The stratigraphical record and palaeogeographical reconstructions show that major environmental coastal changes were linked to these extreme events. At the southern end of the Agoulenitsa Lagoon at modern Kato Samikon high-energy traces were found more than 2 km inland and upt ot 9 m above present sea level. The geo-chronological framework deciphered tsunami landfall for the 5th millennium cal BC (I), mid to late 2nd mill. BC (II), Roman times (1st cent. BC to early 4th cent. AD) (III) and most possible one of the historically well-known 365 AD or 521/551 AD tsunamis (IV).rnCoarse-grained allochthonous sediments of marine origin were found intersecting muddy deposits of the quisecent sediments of the Gialova Lagoon on the southwestern Peloponnese. Radiocarbondatings suggest 6 generations of major tsunami impact. Tsunami generations were dated to around 3300 cal BC (I), around the end of 4th and the beginning of 3rd millennium BC (II), after around 1100 cal BC (III), after the 4th to 2nd cent. BC (IV), between the 8th and early 15th cent. AD (V) and between the mid 14th to beginning of 15th cent. AD (VI). Palaeogeographical and morphological characteristics in the environs of the Gialova Lagoon were controlled by high-energy influence.rnSedimentary findings in all study areas are in good accordance to traces of tsunami events found all over the Ionian Sea. The correlation of geo-chronological data fits very well to coastal Akarnania, the western Peloponnese and finding along the coasts of southern Italy and the Aegean. Supra-regional influence of tsunamigenic impact significant for the investigated sites. The palaeogeographical evolution and palaeo-geomorphological setting of the each study area was strongly affected by tsunamigenic impact.rnThe selected geo-archives represent extraordinary sediment traps for the reconstruction of Holocene coastal evolution. Our result therefore give new insight to the exceptional high tsunami risk in the eastern Mediterranean and emphasize the underestimation of the overall tsunami hazard.

Spatio-temporal credit assignment in population learning

Learning by reinforcement is important in shaping animal behavior, and in particular in behavioral decision making. Such decision making is likely to involve the integration of many synaptic events in space and time. However, using a single reinforcement signal to modulate synaptic plasticity, as suggested in classical reinforcement learning algorithms, a twofold problem arises. Different synapses will have contributed differently to the behavioral decision, and even for one and the same synapse, releases at different times may have had different effects. Here we present a plasticity rule which solves this spatio-temporal credit assignment problem in a population of spiking neurons. The learning rule is spike-time dependent and maximizes the expected reward by following its stochastic gradient. Synaptic plasticity is modulated not only by the reward, but also by a population feedback signal. While this additional signal solves the spatial component of the problem, the temporal one is solved by means of synaptic eligibility traces. In contrast to temporal difference (TD) based approaches to reinforcement learning, our rule is explicit with regard to the assumed biophysical mechanisms. Neurotransmitter concentrations determine plasticity and learning occurs fully online. Further, it works even if the task to be learned is non-Markovian, i.e. when reinforcement is not determined by the current state of the system but may also depend on past events. The performance of the model is assessed by studying three non-Markovian tasks. In the first task, the reward is delayed beyond the last action with non-related stimuli and actions appearing in between. The second task involves an action sequence which is itself extended in time and reward is only delivered at the last action, as it is the case in any type of board-game. The third task is the inspection game that has been studied in neuroeconomics, where an inspector tries to prevent a worker from shirking. Applying our algorithm to this game yields a learning behavior which is consistent with behavioral data from humans and monkeys, revealing themselves properties of a mixed Nash equilibrium. The examples show that our neuronal implementation of reward based learning copes with delayed and stochastic reward delivery, and also with the learning of mixed strategies in two-opponent games.

Spatio-temporal credit assignment in population learning

Learning by reinforcement is important in shaping animal behavior. But behavioral decision making is likely to involve the integration of many synaptic events in space and time. So in using a single reinforcement signal to modulate synaptic plasticity a twofold problem arises. Different synapses will have contributed differently to the behavioral decision and, even for one and the same synapse, releases at different times may have had different effects. Here we present a plasticity rule which solves this spatio-temporal credit assignment problem in a population of spiking neurons. The learning rule is spike time dependent and maximizes the expected reward by following its stochastic gradient. Synaptic plasticity is modulated not only by the reward but by a population feedback signal as well. While this additional signal solves the spatial component of the problem, the temporal one is solved by means of synaptic eligibility traces. In contrast to temporal difference based approaches to reinforcement learning, our rule is explicit with regard to the assumed biophysical mechanisms. Neurotransmitter concentrations determine plasticity and learning occurs fully online. Further, it works even if the task to be learned is non-Markovian, i.e. when reinforcement is not determined by the current state of the system but may also depend on past events. The performance of the model is assessed by studying three non-Markovian tasks. In the first task the reward is delayed beyond the last action with non-related stimuli and actions appearing in between. The second one involves an action sequence which is itself extended in time and reward is only delivered at the last action, as is the case in any type of board-game. The third is the inspection game that has been studied in neuroeconomics. It only has a mixed Nash equilibrium and exemplifies that the model also copes with stochastic reward delivery and the learning of mixed strategies.

Spatio-temporal credit assignment in population learning

We present a model for plasticity induction in reinforcement learning which is based on a cascade of synaptic memory traces. In the cascade of these so called eligibility traces presynaptic input is first corre lated with postsynaptic events, next with the behavioral decisions and finally with the external reinforcement. A population of leaky integrate and fire neurons endowed with this plasticity scheme is studied by simulation on different tasks. For operant co nditioning with delayed reinforcement, learning succeeds even when the delay is so large that the delivered reward reflects the appropriateness, not of the immediately preceeding response, but of a decision made earlier on in the stimulus - decision sequence . So the proposed model does not rely on the temporal contiguity between decision and pertinent reward and thus provides a viable means of addressing the temporal credit assignment problem. In the same task, learning speeds up with increasing population si ze, showing that the plasticity cascade simultaneously addresses the spatial problem of assigning credit to the different population neurons. Simulations on other task such as sequential decision making serve to highlight the robustness of the proposed sch eme and, further, contrast its performance to that of temporal difference based approaches to reinforcement learning.

Spatio-Temporal Credit Assignment in Neuronal Population Learning

n learning from trial and error, animals need to relate behavioral decisions to environmental reinforcement even though it may be difficult to assign credit to a particular decision when outcomes are uncertain or subject to delays. When considering the biophysical basis of learning, the credit-assignment problem is compounded because the behavioral decisions themselves result from the spatio-temporal aggregation of many synaptic releases. We present a model of plasticity induction for reinforcement learning in a population of leaky integrate and fire neurons which is based on a cascade of synaptic memory traces. Each synaptic cascade correlates presynaptic input first with postsynaptic events, next with the behavioral decisions and finally with external reinforcement. For operant conditioning, learning succeeds even when reinforcement is delivered with a delay so large that temporal contiguity between decision and pertinent reward is lost due to intervening decisions which are themselves subject to delayed reinforcement. This shows that the model provides a viable mechanism for temporal credit assignment. Further, learning speeds up with increasing population size, so the plasticity cascade simultaneously addresses the spatial problem of assigning credit to synapses in different population neurons. Simulations on other tasks, such as sequential decision making, serve to contrast the performance of the proposed scheme to that of temporal difference-based learning. We argue that, due to their comparative robustness, synaptic plasticity cascades are attractive basic models of reinforcement learning in the brain.

High-resolution sequence stratigraphic correlation in the Upper Jurassic (Kimmeridgian)–Upper Cretaceous (Cenomanian) peritidal carbonate deposits (Western Taurides, Turkey)

Upper Jurassic (Kimmeridgian)±Upper Cretaceous (Cenomanian) inner platform carbonates in the Western Taurides are composed of metre-scale upward-shallowing cyclic deposits (parasequences) and important karstic surfaces capping some of the cycles. Peritidal cycles (shallow subtidal facies capped by tidal-¯at laminites or fenestrate limestones) are regressive- and transgressive-prone (upward-deepening followed by upward-shallowing facies trends). Subtidal cycles are of two types and indicate incomplete shallowing. Submerged subtidal cycles are composed of deeper subtidal facies overlain by shallow subtidal facies. Exposed subtidal cycles consist of deeper subtidal facies overlain by shallow subtidal facies that are capped by features indicative of prolonged subaerial exposure. Subtidal facies occur characteristically in the Jurassic, while peritidal cycles are typical for the Lower Cretaceous of the region. Within the foraminiferal and dasyclad algal biostratigraphic framework, four karst breccia levels are recognized as the boundaries of major second-order cycles, introduced for the ®rst time in this study. These levels correspond to the Kimmeridgian±Portlandian boundary, mid-Early Valanginian, mid-Early Aptian and mid-Cenomanian and represent important sea level falls which affected the distribution of foraminiferal fauna and dasyclad ¯ora of the Taurus carbonate platform. Within the Kimmeridgian±Cenomanian interval 26 third-order sequences (types 1 and 2) are recognized. These sequences are the records of eustatic sea level ¯uctuations rather than the records of local tectonic events because the boundaries of the sequences representing 1±4 Ma intervals are correlative with global sea level falls. Third-order sequences and metre-scale cyclic deposits are the major units used for long-distance, high-resolution sequence stratigraphic correlation in the Western Taurides. Metre-scale cyclic deposits (parasequences) in the Cretaceous show genetical stacking patterns within third-order sequences and correspond to fourth-order sequences representing 100±200 ka. These cycles are possibly the E2 signal (126 ka) of the orbital eccentricity cycles of the Milankovitch band. The slight deviation of values, calculated for parasequences, from the mean value of eccentricity cycles can be explained by the currently imprecise geochronology established in the Cretaceous and missed sea level oscillations when the platform lay above fluctuating sea level.

Multi-Model Data Fusion to Improve an Early Warning System for Hypo-/Hyperglycemic Events

Correct predictions of future blood glucose levels in individuals with Type 1 Diabetes (T1D) can be used to provide early warning of upcoming hypo-/hyperglycemic events and thus to improve the patient's safety. To increase prediction accuracy and efficiency, various approaches have been proposed which combine multiple predictors to produce superior results compared to single predictors. Three methods for model fusion are presented and comparatively assessed. Data from 23 T1D subjects under sensor-augmented pump (SAP) therapy were used in two adaptive data-driven models (an autoregressive model with output correction - cARX, and a recurrent neural network - RNN). Data fusion techniques based on i) Dempster-Shafer Evidential Theory (DST), ii) Genetic Algorithms (GA), and iii) Genetic Programming (GP) were used to merge the complimentary performances of the prediction models. The fused output is used in a warning algorithm to issue alarms of upcoming hypo-/hyperglycemic events. The fusion schemes showed improved performance with lower root mean square errors, lower time lags, and higher correlation. In the warning algorithm, median daily false alarms (DFA) of 0.25%, and 100% correct alarms (CA) were obtained for both event types. The detection times (DT) before occurrence of events were 13.0 and 12.1 min respectively for hypo-/hyperglycemic events. Compared to the cARX and RNN models, and a linear fusion of the two, the proposed fusion schemes represents a significant improvement.

Measurement of the underlying event in jet events from 7 TeV proton-proton collisions with the ATLAS detector

Distributions sensitive to the underlying event in QCD jet events have been measured with the ATLAS detector at the LHC, based on 37 pb−1 of proton–proton collision data collected at a centre-of-mass energy of 7 TeV. Chargedparticle mean pT and densities of all-particle ET and chargedparticle multiplicity and pT have been measured in regions azimuthally transverse to the hardest jet in each event. These are presented both as one-dimensional distributions and with their mean values as functions of the leading-jet transverse momentum from 20 to 800 GeV. The correlation of chargedparticle mean pT with charged-particle multiplicity is also studied, and the ET densities include the forward rapidity region; these features provide extra data constraints for Monte Carlo modelling of colour reconnection and beamremnant effects respectively. For the first time, underlying event observables have been computed separately for inclusive jet and exclusive dijet event selections, allowing more detailed study of the interplay of multiple partonic scattering and QCD radiation contributions to the underlying event. Comparisonsto the predictions of different Monte Carlo models show a need for further model tuning, but the standard approach is found to generally reproduce the features of the underlying event in both types of event selection.

Diatom events and Neogene nannofossils of ODP Hole 201-1225A

Shipboard investigation of magnetostratigraphy and shore-based investigation of diatoms and calcareous nannofossils were used to identify datum events in sedimentary successions collected at Ocean Drilling Program (ODP) Leg 201 Site 1225. The goal was to extend the magnetic record previously studied at the same site, ODP Leg 138 Site 851, and provide a comprehensive age model for Site 1225. Two high-magnetic intensity zones at 0-70 and 200-255 meters below seafloor (mbsf) were correlated with lithologic Subunits IA and IC in Hole 1225A. Subunit IA (0-70 mbsf) contains the magnetic reversal record until the Cochiti Subchronozone (3.8 Ma) and has a sedimentation rate of 1.7 cm/k.y. This agrees with previous work done at Site 851. Subunit IC (200-255 mbsf) was not sampled at Site 851. Diatom and nannofossil biostratigraphy constrained this subunit, and we found it to contain the magnetic reversal record between Subchrons C4n.2r and C5n.2n (8.6-9.7 Ma), yielding a sedimentation rate of 2.7 cm/k.y. Biostratigraphy was used to establish the sedimentation rates within Subunits IB and ID (70-200 mbsf and 255-300 mbsf, respectively). These subunits had higher sedimentation rates (~3.4 cm/k.y.) and coincide with the late Miocene-early Pliocene biogenic bloom event (4.5-7 Ma) and the Miocene global cooling trend (10-15 Ma). High biogenic productivity associated with these subunits resulted in the pyritization of the magnetic signal. In lithologic Subunit ID, basement flow is another factor that may be altering the magnetic signal; however, the good correlation between the biostratigraphy and magnetostratigraphy indicates that the magnetic record was locked-in near the seafloor and suggests the age model is robust.