Harald Scheule & Daniel Rösch 
深度信用风险 (Deep Credit Risk) – 使用Python进行机器学习 [EPUB ebook] 

– 了解流动性，房屋净值和许多其他关键银行业特征变量的作用；

– 选择并处理变量；

– 预测违约、偿付、损失率和风险敞口；

– 利用危机前特征预测经济衰退和危机后果；

– 理解COVID-19对信用风险带来的影响；

– 将创新的抽样技术应用于模型训练和验证；

– 从Logit分类器到随机森林和神经网络的深入学习;

– 进行无监督聚类、主成分和贝叶斯技术的应用；

– 为CECL、IFRS 9和CCAR建立多周期模型；

– 建立用于在险价值和期望损失的信贷组合相关模型；

– 使用更多真实的信用风险数据并运行超过1500行的代码…

– Understand the role of liquidity, equity and many other key banking features

– Engineer and select features

– Predict defaults, payoffs, loss rates and exposures

– Predict downturn and crisis outcomes using pre-crisis features

– Understand the implications of COVID-19

– Apply innovative sampling techniques for model training and validation

– Deep-learn from Logit Classifiers to Random Forests and Neural Networks

– Do unsupervised Clustering, Principal Components and Bayesian Techniques

– Build multi-period models for CECL, IFRS 9 and CCAR

– Build credit portfolio correlation models for Va R and Expected Shortfal

– Run over 1, 500 lines of pandas, statsmodels and scikit-learn Python code

– Access real credit data and much more …