Search

Credit scoring is an important topic in financial activities and bankruptcy prediction that has been extensively explored using deep neural network (DNN) methods. DNN-based credit scoring models rely heavily on a large amount of labeled data. The accuracy of DNN-based credit assessment models relies heavily on large amounts of labeled data. However, purely data-driven learning makes it difficult to encode human intent to guide the model to capture the desired patterns and leads to low transparency of the model. Therefore, the Probabilistic Soft Logic Posterior Regularization (PSLPR) framework is proposed for integrating prior knowledge of logic rule with neural network. First, the PSLPR framework calculates the rule satisfaction distance for each instance using a probabilistic soft logic formula. Second, the logic rules are integrated into the posterior distribution of the DNN output to form a logic output. Finally, a novel discrepancy loss which measures the difference between the real label and the logic output is used to incorporate logic rules into the parameters of the neural network. Extensive experiments were conducted on two datasets, the Australian credit dataset and the credit card customer default dataset. To evaluate the obtained systems, several performance metrics were used, including PCC, Recall, F1 and AUC. The results show that compared to the standard DNN model, the four evaluation metrics are increased by 7.14%, 14.29%, 8.15%, and 5.43% respectively on the Australian credit dataset.

Purpose Deficiency of adenosine deaminase 2 (DADA2), an autosomal recessive autoinflammatory disorder caused by biallelic loss-of-function variants in adenosine deaminase 2 (ADA2), has not been systemically investigated in Chinese population yet. We aim to further characterize DADA2 cases in China. Methods A retrospective analysis of patients with DADA2 identified through whole exome sequencing (WES) at seventeen rheumatology centers across China was conducted. Clinical characteristics, laboratory findings, genotype, and treatment response were analyzed. Results Thirty patients with DADA2 were enrolled between January 2015 and December 2021. Adenosine deaminase 2 enzymatic activity was low in all tested cases to confirm pathogenicity. Median age of disease presentation was 4.3 years and the median age at diagnosis was 7.8 years. All but one patient presented during childhood and two subjects died from complications of their disease. The patients most commonly presented with systemic inflammation (92.9%), vasculitis (86.7%), and hypogammaglobinemia (73.3%) while one patient presented with bone marrow failure (BMF) with variable cytopenia. Twenty-three (76.7%) patients were treated with TNF inhibitors (TNFi), while two (6.7%) underwent hematopoietic stem cell transplantation (HSCT). They all achieved clinical remission. A total of thirty-nine ADA2 causative variants were identified, six of which were novel. Conclusion To establish early diagnosis and improve clinical outcomes, genetic screening and/or testing of ADA2 enzymatic activity should be performed in patients with suspected clinical features. TNFi is considered as first line treatment for those with vascular phenotypes. HSCT may be beneficial for those with hematological disease or in those who are refractory to TNFi.

Meteorological drought is one of the most serious natural disasters, and its impact in arid and semi-arid areas is significant. In order to explore the temporal and spatial distribution of meteorological disasters in Gansu Province, we first calculated the standardized precipitation evapotranspiration index (SPEI) based on the monthly meteorological data from 1969 to 2018 and extracted the drought events through the theory of runs. Then, REOF rotation orthogonal decomposition was performed to divide the study area into five climatic subregions. With each subregion as the basic unit, the variation characteristics and evolution trends of drought events at different time scales were compared based on the B-G segmentation algorithm (BG-algorithm). Finally, a correlation analysis was conducted to explore the driving factors of drought events in each subregion. The main conclusions are as follows: (1) The cumulative duration of drought in the study area showed a slight increase trend (0.475 day/decade) and a 19-year main cycle. The drought intensity showed a trend of first easing and then intensifying, especially after 2000; the drought intensified significantly and showed a spatial trend of decreasing drought in the northwest and worsening drought in the southeast. (2) The cumulative contribution rate of the first five modes of REOF decomposition was 64.46%, and the study was divided into five arid subregions: the Hexi region, middle Hedong region, eastern Hedong region, Wushaoling region and western Hedong region. (3) The meteorological drought in the Hexi region has eased significantly since 1988. In the eastern, central and western parts of the Yellow River, drought intensification was observed to have occurred in different degrees (0.12/decade, 0.129/decade, and 0.072/decade). The meteorological drought in the Wuelyaling region has alleviated significantly with a watershed region formed between drought alleviation and drought intensification. (4) Seasonally, the eastern Hedong region showed a significant trend of drought in spring, but the opposite in autumn. The trend of climate drying was obvious in the spring and summer, rather than in autumn and winter. The spring drought trend is the most obvious in the middle of the Hedong region. (5) The meteorological drought in the study area was affected by local climatic factors and circulation factors, but there were significant differences in the responses of different arid subregions to these factors.

Credit is a part of external image of enterprises, and it directly affects interests of enterprises. Nowadays, most of researches on predictions of enterprises credit use a single algorithm model or optimize a single model to predict an enterprises credit score. The accuracy of each model is different, and the generalization ability is generally weak. In order to improve generalization ability of models and accuracy of prediction results, a parallel double-layer prediction model is proposed in this paper. The model is based on Stacking and Bagging methods, which can improve generalization ability with high accuracy. Through experiments, we compare three single algorithm models, four integrated learning models with other combination strategies and parallel double-layer prediction model. Average value of four evaluation indexes are increased by 4.2349%, 63.1464%, 34.11837%, 1.26104%, 15.7862%, 10.1457% and 25.6310% respectively. The results show that the parallel double-layer prediction model is accurate and feasible.

The evaluation performance of PDP (Policy Decision Point) plays a key role in the operation of a system. In order to solve bottlenecks of improving the PDP evaluation performance for large-scale policy sets, we propose an evaluation engine based on artificial neural networks, namely ANNPDP. We transform rules in a large-scale policy set described in the XACML (eXtensible Access Control Markup Language) into numerical rules. Evaluation networks are established and trained by the numerical rules. In order to ensure the accuracy, a misjudgment set is constructed for error corrections and stored by hash indexes. By simulating the arrival of requests, ANNPDP is compared with the Sun PDP, HPEngine, XEngine and SBA-XACML. The experiment results show that ANNPDP has: 1) high performance: if the number of requests reaches 10,000, the evaluation time of ANNPDP on the large-scale policy set with 100,000 rules is approximately 0.46%, 0.93%, 0.71% and 1.43% of that of the Sun PDP, HPEngine, XEngine and SBA-XACML, respectively, and 2) stability: as the size of the large-scale policy set and the number of requests increase, the evaluation time of ANNPDP grows linearly. ANNPDP can satisfy the requirements of an authorization system with large-scale policy sets