关键词： 机器学习   企业贷款   违约风险  

摘要： 研究企业的贷款违约风险不仅对金融机构解决“惜贷”问题和防范信用风险具有重要的现实意义,而且能为企业规范自身经营和改善财务状况提出有针对性的建议及措施。本文根据某机构的企业贷款违约数据对贷款违约风险进行研究,首先对原始数据进行缺失值处理、特征选择和不平衡数据处理,然后利用逻辑回归、随机森林、XGBoost和LightGBM四种机器学习方法对数据进行建模和分析并比较模型优劣,最后利用GBDT模型计算特征重要性。结果表明:1) 三种集成模型的预测效果显著优于单一模型,2) 在集成模型中LightGBM模型表现出了最好的预测性能,3) 企业的纳税情况和曾经获得的授信情况可以作为判断该企业是否会发生贷款逾期现象的重要参考。

关键词： 感染数   科研团队   机器学习方法   持续时间   干预措施   肺炎死亡率   防护措施   新加坡南洋理工大学  

摘要： 近日,新加坡南洋理工大学的一个科研团队利用机器学习方法,通过输入大量2020年全球新冠肺炎死亡率和感染数据,提出了一种新冠预测模型NSGA-Ⅱ。该模型可用于就新冠感染和死亡率可能存在的激增情况提前向地方政府发出警报,让他们有时间更快地提出相关对策。通过及时、有针对性地为特定国家/地区给出新冠干预措施的最佳应用和持续时间(例如隔离、防护措施等)的建议,该模型可以减少新冠感染率和死亡率。

关键词： 产量预测   机器学习   神经网络   工程经验模型  

摘要： 页岩油气产量预测是确定其开发经济性的重要手段,目前的产量预测研究很少能在物理模型与数据挖掘方法之间达到统一.针对页岩油气的产量分析,本研究深入结合误差反向传递(BP)神经网络和长短期记忆(LSTM)神经网络的数学方法优势,综合考虑工程经验模型的约束,改善了模型预测精度,经过实例数据训练后可较好地预测油田产量,并研究了页岩储层深度、总有机碳含量(TOC)、脆性度等油田参数对产量预测的影响规律.这项工作可以为页岩油气规模化开发提供可靠的产量预测和经济评价.

关键词： Particle and Nuclear Physics  

摘要： This paper is devoted to the analysis of manifestations of clustering in rare multibody decays of heavy nuclei. A computer model of the fine structure was developed jointly with the physicists of FLNR JINR;it was found based on experiments with the transuranium element Californium. To test the hypothesis that the structure really exists and is not a noise artifact, it was proposed to use a deep convolution network as a binary classifier trained on a large sample of model and noise images. Preliminary results of using the developed neuroclassifier show the prospects for this approach.

关键词： Evapotranspiration   Machine learning   process-based ET   ET integration   China   Gaussian process regression  

摘要： Estimating terrestrial evapotranspiration (ET) accurately at various temporal and spatial scales is crucial for understanding the hydrological cycle and water resource management. The currently available ET estimates have some uncertainties and need to be further improved. In this study, six machine learning methods including the random forests, support vector machine, Gaussian process regression (GPR), ensemble trees, general regression neural network, and Bayesian Model Averaging, are applied and evaluated to improve China terrestrial ET estimation by integrating five process-based ET algorithms including SEMI-PM, RS-PM, RRS-PM, MOD16, and PMLv2. Then evaluations are conducted with the eddy covariance flux observations at 14 China flux tower sites distributing in forest, shrub, wetland, grassland, and cropland, as well as with water balance-based ET at basin scale. According to the multiple training, validation, and testing, the GPR method is superior to all the other methods. Compared with the individual algorithms, the GPR method can reduce the root mean square error (RMSE) by 0.45 mm day(-1) (for SEMI-PM) similar to 0.81 mm day(-1) (for PMLv2), improve the coefficient of determination (R-2) by 0.061 (for PMLv2) similar to 0.33 (for MOD16), and decrease the absolute relative percent error (RPE) by 8.32% (for RS-PM) similar to 42.47% (for PMLv2) for all the test data. At basin scale, the results demonstrate that the annual GPR-merged China terrestrial ET is reliable (R-2 = 0.88, RMSE = 57.18 mm year RPE = -0.26%) and has higher accuracy than the currently available eight high-resolution ET products and the estimates from the other five machine learning methods and the five single ET models. The annual average terrestrial ET across China for 2000-2018 estimated by the GPR method is approximately 397.65 mm year More ground-based observations of terrestrial ET covering various land types should be collected to update the integrating methods and improve ET estimates.

关键词： fuzzy C-means   GLCM   glioma grade   grasshopper optimization algorithm   local binary pattern  

摘要： Detection of brain tumor's grade is a very important task in treatment plan design which was done using invasive methods such as pathological examination. This examination needs resection procedure and resulted in pain, hemorrhage and infection. The aim of this study is to provide an automated non-invasive method for estimation of brain tumor's grade using Magnetic Resonance Images (MRI). After pre-processing, using Fuzzy C-Means (FCM) segmentation method, tumor region was extracted from post-processed images. In feature extraction, texture, Local Binary Pattern (LBP) and fractal-based features were extracted using Matlab software. Then using Grasshopper Optimization Algorithm (GOA), parameters of three different classification methods including Random Forest (RF), K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) were optimized. Finally, performance of three applied classifiers before and after optimization were compared. The results showed that the random forest with accuracy of 99.09% has achieved better performance comparing other classification methods.

关键词： susceptibility evaluation   machine-learning (ML)   particle swarm optimization (PSO)   support vector machines (SVM)   random forest (RF)  

摘要： Regarding the ever increasing and frequent occurrence of serious landslide disaster in eastern Guangxi, the current study was implemented to adopt support vector machines (SVM), particle swarm optimization support vector machines (PSO-SVM), random forest (RF), and particle swarm optimization random forest (PSO-RF) methods to assess landslide susceptibility in Zhaoping County. To this end, 10 landslide disaster-related variables including digital elevation model (DEM)-derived, meteorology-derived, Landsat8-derived, geology-derived, and human activities factors were provided. Of 345 landslide disaster locations found, 70% were used to train the models, and the rest of them were performed for model verification. The aforementioned four models were run, and landslide susceptibility evaluation maps were produced. Then, receiver operating characteristics (ROC) curves, statistical analysis, and field investigation were performed to test and verify the efficiency of these models. Analysis and comparison of the results denoted that all four landslide models performed well for the landslide susceptibility evaluation as indicated by the area under curve (AUC) values of ROC curves from 0.863 to 0.934. Among them, it has been shown that the PSO-RF model has the highest accuracy in comparison to other landslide models, followed by the PSO-SVM model, the RF model, and the SVM model. Moreover, the results also showed that the PSO algorithm has a good effect on SVM and RF models. Furthermore, the landslide models devolved in the present study are promising methods that could be transferred to other regions for landslide susceptibility evaluation. In addition, the evaluation results can provide suggestions for disaster reduction and prevention in Zhaoping County of eastern Guangxi.

关键词： Pyrolysis   Bio-oil yield   Machine learning   Genetic algorithm   Biomass to energy  

摘要： A novel genetic algorithm-based feature selection approach is incorporated and based on these features, four different ML methods were investigated. According to the findings, ML models could reliably predict bio-oil yield. The results showed that Random forest (RF) is preferred for bio-oil yield prediction (R2 - 0.98) and highly recommended when dealing with the complex correlation between variables and target. Multi-Linear regression model showed relatively poor generalization performance (R2 - 0.75). The partial dependence analysis was done for ML models to show the influence of each input variable on the target variable. Lastly, an easy-to-use software package was developed based on the RF model for the prediction of bio-oil yield. The current study offered new insights into the pyrolysis process of biomass and to improve bio-oil yield. It is an attempt to reduce the time-consuming and expensive experimental work for estimating the bio-oil yield of biomass during pyrolysis.

关键词： maize   yield prediction   fertilizer systems   machine learning  

摘要： Timely and reliable maize yield prediction is essential for the agricultural supply chain and food security. Previous studies using either climate or satellite data or both to build empirical or statistical models have prevailed for decades. However, to what extent climate and satellite data can improve yield prediction is still unknown. In addition, fertilizer information may also improve crop yield prediction, especially in regions with different fertilizer systems, such as cover crop, mineral fertilizer, or compost. Machine learning (ML) has been widely and successfully applied in crop yield prediction. Here, we attempted to predict maize yield from 1994 to 2007 at the plot scale by integrating multi-source data, including monthly climate data, satellite data (i.e., vegetation indices (VIs)), fertilizer data, and soil data to explore the accuracy of different inputs to yield prediction. The results show that incorporating all of the datasets using random forests (RF) and AB (adaptive boosting) can achieve better performances in yield prediction (R-2: 0.85 similar to 0.98). In addition, the combination of VIs, climate data, and soil data (VCS) can predict maize yield more effectively than other combinations (e.g., combinations of all data and combinations of VIs and soil data). Furthermore, we also found that including different fertilizer systems had different prediction accuracies. This paper aggregates data from multiple sources and distinguishes the effects of different fertilization scenarios on crop yield predictions. In addition, the effects of different data on crop yield were analyzed in this study. Our study provides a paradigm that can be used to improve yield predictions for other crops and is an important effort that combines multi-source remotely sensed and environmental data for maize yield prediction at the plot scale and develops timely and robust methods for maize yield prediction grown under different fertilizing systems.

关键词： >= 2 MeV electron fluxes   GOES satellites   fully connected neural network   relativistic electron enhancement event  

摘要： The energetic electrons in the Earth's radiation belt, known as "killer electrons", are one of the crucial factors for the safety of geostationary satellites. Geostationary satellites at different longitudes encounter different energetic electron environments. However, organizations of space weather prediction usually only display the real-time >= 2 MeV electron fluxes and the predictions of >= 2 MeV electron fluxes or daily fluences within the next 1-3 days by models at one location in GEO orbit. In this study, the relationship of >= 2 MeV electron fluxes at different longitudes is investigated based on observations from GOES satellites, and the relevant models are developed. Based on the observations from GOES-10 and GOES-12 after calibration verification, the ratios of the >= 2 MeV electron daily fluences at 135 degrees W to those at 75 degrees W are mainly in the range from 1.0 to 4.0, with an average of 1.92. The models with various combinations of two or three input parameters are developed by the fully connected neural network for the relationship between >= 2 MeV electron fluxes at 135 degrees W and 75 degrees W in GEO orbit. According to the prediction efficiency (PE), the model only using log10 (fluxes) and MLT from GOES-10 (135 degrees W), whose PE can reach 0.920, has the best performance to predict >= 2 MeV electron fluxes at the locations of GOES-12 (75 degrees W). Its PE is larger than that (0.882) of the linear model using log10 (fluxes four hours ahead) from GOES-10 (135 degrees W). We also develop models for the relationship between >= 2 MeV electron fluxes at 75 degrees W and at variable longitudes between 95.8 degrees W and 114.9 degrees W in GEO orbit by the fully connected neural network. The PE values of these models are larger than 0.90. These models realize the predictions of >= 2 MeV electron fluxes at arbitrary longitude between 95.8 degrees W and 114.9 degrees W in GEO orbit.