Does segmentation always improve model performance in credit scoring?

Credit scoring allows for the credit risk assessment of bank customers. A single scoring model (scorecard) can be developed for the entire customer population, e.g. using logistic regression. However, it is often expected that segmentation, i.e. dividing the population into several groups and building separate scorecards for them, will improve the model performance. The most common statistical methods for segmentation are the two-step approaches, where logistic regression follows Classification and Regression Trees (CART) or Chi-squared Automatic Interaction Detection (CHAID) trees etc. In this research, the two-step approaches are applied as well as a new, simultaneous method, in which both segmentation and scorecards are optimised at the same time: Logistic Trees with Unbiased Selection (LOTUS). For reference purposes, a single-scorecard model is used. The above-mentioned methods are applied to the data provided by two of the major UK banks and one of the European credit bureaus. The model performance measures are then compared to examine whether there is improvement due to the segmentation methods used. It is found that segmentation does not always improve model performance in credit scoring: for none of the analysed real-world datasets, the multi-scorecard models perform considerably better than the single-scorecard ones. Moreover, in this application, there is no difference in performance between the two-step and simultaneous approaches.     

NE–LP: Normalized entropy- and loss prediction-based sampling for active learning in Chinese word segmentation on EHRs

Neural Computing and Applications - Electronic health records (EHRs) in hospital information systems contain patients’ diagnoses and treatments, so EHRs are essential to clinical data mining....     

Does training in the cold improve cold performance?

Despite the well-described physiological response to cold, little is known about how to best train persons to perform motor tasks in cold conditions. It is unclear if principles of training specificity would apply to motor skill training when cold because cold exposure reduces tactile sensitivity and cognitive function, which may reduce rather than enhance training efficacy. The purpose of this study is to determine whether training in the cold facilitates performance in the cold. To do this, we investigated the effect of cold or thermoneutral training on performance of the Grooved Pegboard Task. Twenty persons (11M, 9F, mean age 31.2 ± 5.44 years) visited the lab on two separate days and were randomly assigned to either a cold (5M, 5F) or thermoneutral (6M, 4F) training group. On day one, participants were tested at baseline and then performed 20 repetitions of the task according to their respective group assignment. Prior to each task repetition during training, the cold training group immersed their hand in cold water (2 °C) while the thermoneutral group immersed their hand in thermoneutral water (34 °C). Following training, participants were tested once again (immediate retention test). During testing, participants performed the task twice; once after immersing their hands in cold water and once after immersing their hands in thermoneutral water. On day two, delayed retention tests were performed. Time to completion and number of errors were recorded during testing. There were no differences in time to completion at delayed retention tests between thermoneutral and cold-trained groups (p = 0.434). The incidence of errors was significantly less in the cold-trained group than the thermoneutral training group at delayed retention testing (p = 0.035). The main finding of this study was that dexterity but not speed was improved by cold training. Further research is required but the findings presented here suggest that industries that require motor tasks in the cold could benefit from training employees in cold conditions.     

Do background colors improve program comprehension in the #ifdef hell?

Software-product-line engineering aims at the development of variable and reusable software systems. In practice, software product lines are often implemented with preprocessors. Preprocessor directives are easy to use, and many mature tools are available for practitioners. However, preprocessor directives have been heavily criticized in academia and even referred to as “#ifdef hell”, because they introduce threats to program comprehension and correctness. There are many voices that suggest to use other implementation techniques instead, but these voices ignore the fact that a transition from preprocessors to other languages and tools is tedious, erroneous, and expensive in practice. Instead, we and others propose to increase the readability of preprocessor directives by using background colors to highlight source code annotated with ifdef directives. In three controlled experiments with over 70 subjects in total, we evaluate whether and how background colors improve program comprehension in preprocessor-based implementations. Our results demonstrate that background colors have the potential to improve program comprehension, independently of size and programming language of the underlying product. Additionally, we found that subjects generally favor background colors. We integrate these and other findings in a tool called FeatureCommander, which facilitates program comprehension in practice and which can serve as a basis for further research.     

Do simple cells in primary visual cortex form a tight frame?

Sets of neuronal tuning curves, which describe the responses of neurons as functions of a stimulus, can serve as a basis for approximating other functions of stimulus parameters. In a function-approximating network, synaptic weights determined by a correlation-based Hebbian rule are closely related to the coefficients that result when a function is expanded in an orthogonal basis. Although neuronal tuning curves typically are not orthogonal functions, the relationship between function approximation and correlation-based synaptic weights can be retained if the tuning curves satisfy the conditions of a tight frame. We examine whether the spatial receptive fields of simple cells in cat and monkey primary visual cortex (V1) form a tight frame, allowing them to serve as a basis for constructing more complicated extrastriate receptive fields using correlation-based synaptic weights. Our calculations show that the set of V1 simple cell receptive fields is not tight enough to account for the acuity observed psychophysically.     

Comparison of statistical models performance in case of?segmentation using a small amount of training datasets

Model-based image segmentation has been extensively used in medical imaging to learn both the shape and appearance of anatomical structures from training datasets. The more training datasets are used, the more accurate is the segmented model, as we account for more information about its variability. However, training datasets of large size with a proper sampling of the population may not always be available. In this paper, we compare the performance of statistical models in the context of lower limb bones segmentation using MR images when only a small number of datasets is available for training. For shape, both PCA-based priors and shape memory strategies are tested. For appearance, methods based on intensity profiles are tested, namely mean intensity profiles, multivariate Gaussian distributions of profiles and multimodal profiles from EM clustering. Segmentation results show that local and simple methods perform the best when a small number of datasets is available for training. Conversely, statistical methods feature the best segmentation results when the number of training datasets is increased.     

Hybrid Social Media Use and Guanxi Types: How Do Employees Use Social Media in the Chinese Workplace?

We investigate how work-related and personal-related social media (SM) use contributes to Chinese employees’ guanxi identification and extra-role behaviors (ERB). Our findings indicate that work-related SM use is positively associated with team member guanxi identification, which further contributes to employees’ ERB targeting the team. Personal-related SM use is positively related to insider guanxi identification and further contributes to ERB targeting the individual. When team communication is highly dependent on SM, the positive effect of personal-related SM use on guanxi identification is weakened. The results contribute to a nuanced understanding of the crucial role of SM use in the Chinese workplace.     

I spy with my AI: The effects of AI-based visual cueing on human operators’ performance and cognitive load in CCTV control rooms

The increased number of security cameras in modern cities has elevated the video-feed monitoring demands of closed-circuit television (CCTV) operators. As a result, new AI-driven support systems that leverage the power of computer vision algorithms have been deployed to facilitate the operators' work. However, to effectively design intuitive, AI-driven interfaces and validate their impact on the operators' performance, extensive user testing is required. To address this, we previously developed and tested a virtual reality (VR) control room that can be used to iteratively evaluate intelligent computer assistants and interfaces while operators are subjected to different cognitive load. In the present study, we use this VR environment and physiological markers (e.g., eye tracking measures) to investigate how AI-based visual cueing (i.e., pushing forward video streams on which detections are highlighted by rectangles drawn around targets) affects operator performance and cognitive load. Results suggest that support systems using such technology in a control room improve operators’ performance and decrease their cognitive load, as reflected by changes in pupil dilation and subjective reports irrespective of induced cognitive load.     

Do information systems improve the quality of clinical research? Results of a randomized trial in a cooperative multi-institutional cancer group

A facsimile communications network was established to serve eight randomly selected medical centers belonging to the Southeastern Cancer Study Group (SEG) to test whether clinical algorithms could augment protocol compliance and patient safety. An SEG protocol (75HD0103) for testing alternative chemotherapy regimens in Advanced Hodgkin's Disease was rewritten as a clinical algorithm. The algorithm generates treatment advice rules, emulating how the clinical researcher would accurately and appropriately apply the general protocol to an individual patient's specific visit, considering his prior response to therapy. The visit-specific advice rules were typed onto general encounter forms and sent to the oncologists over the facsimile equipment prior to each therapy visit. A facsimile of the completed form was transmitted back. Protocol compliance, as judged by explicit criteria on a visit-by-visit basis, was 94% in the algorithm user group and 64% in the nonuser group (p < 0.001). Two measures of hematologic toxicity, “low” counts and “toxic” nadirs, were associated more often with protocol noncompliance. We conclude that prospective information management can improve the “control” in cooperative controlled clinical trials.