Explaining Aggregated Recovery Rates

This study on explaining aggregated recovery rates (ARR) is based on the largest existing loss and recovery database for commercial loans provided by Global Credit Data, which includes defaults from 5 continents and over 120 countries. The dependence of monthly ARR from bank loans on various macroeconomic factors is examined and sources of their variability are stated. For the first time, an influence of stochastically estimated monthly growth of GDP USA and Europe is quantified. To extract monthly signals of GDP USA and Europe, dynamic factor models for panel data of different frequency information are employed. Then, the behavior of the ARR is investigated using several regression models with unshifted and shifted explanatory variables in time to improve their forecasting power by taking into account the economic situation after the default. An application of a Markov switching model shows that the distribution of the ARR differs between crisis and prosperity times. The best fit among the compared models is reached by the Markov switching model. Moreover, a significant influence of the estimated monthly growth of GDP in Europe is observed for both crises and prosperity times.

A federated learning based semi-supervised credit prediction approach enhanced by multi-layer label mean

Learning based credit prediction has attracted great interest from academia and industry. Different institutions hold a certain amount of credit data with limited users to build model. An institution has the requirement to obtain data from other institutions for improving model performance. However, due to the privacy protection and subject to legal restrictions, they encounter difficulties in data exchange. This affects the performance of credit prediction. In order to solve the above problem, this paper proposes a federated learning based semi-supervised credit prediction approach enhanced by multi-layer label mean, which can aggregate parameters of each institution via joint training while protecting the data privacy of each institution. Moreover, in actual production and life, there are usually more unlabeled credit data than labeled ones, and the distribution of their feature space presents multiple data-dense divisions. To deal with these, local meanNet model is proposed with a multi-layer label mean based semi-supervised deep learning network. In addition, this paper introduces a cost-sensitive loss function in the supervised part of the local mean model. Conducted on two public credit datasets, experimental results show that our proposed federated learning based approach has achieved promising credit prediction performance in terms of Accuracy and F1 measures. At the same time, the framework design mode that splits data aggregation and keys uniformly can improve the security of data privacy and enhance the flexibility of model training.

LITERASI INFORMASI LINTAS BUDAYA KUNCI KESUKSESAN BERBASIS DARING (KOLABORASI DOSEN DAN ALUMNI)

The purpose of this webinar with the theme Interkulturelle Landeskunde, which is conducted virtually or online, is to provide information on how important cross-cultural communication is in building a good relationship with individuals or communities with different backgrounds. This webinar activity was attended by 2nd semester students who took Interkulturelle Landeskunde, numbered 18 peoples and high school or vocational students, namely from SMAN 9 Ambon, SMAN 4 Maluku Tengah, SMAN 1 Seram Bagian barat, SMAN 44 Maluku Tengah, and SMKS Tourism Pamahanunusa in Masohi, a total of 62 peoples. The methods used in this webinar activity are the presentation of material, questions, and answers, filling out questionnaires distributed during the webinar, and playing videos about German and its language knowledge. The questionnaire distributed contained questions related to the material provided by the speakers. At the end of the competition, there is a quick and precise competition in the form of a quiz for students. Students who can answer quickly and accurately are given prizes in the form of credit data. The webinar was carried out by presenting material by three German speakers, and three alumni of German Language Education Study Program students who were already working and living in Germany. And one lecturer of the German Language Education Study Program FKIP Unpatti who is currently studying further in Dresden (Germany). The results of the questionnaire answers obtained showed that the activities carried out succeeded in making students learn to continue their studies or work in Germany as well as important cross-cultural understanding.

Using Model Performance to Assess the Representativeness of Data for Model Development and Calibration in Financial Institutions

This paper proposes a methodology that utilises model performance as a metric to assess the representativeness of external or pooled data when it is used by banks in regulatory model development and calibration. There is currently no formal methodology to assess representativeness. The paper provides a review of existing regulatory literature on the requirements of assessing representativeness and emphasises that both qualitative and quantitative aspects need to be considered. We present a novel methodology and apply it to two case studies. We compared our methodology with the Multivariate Prediction Accuracy Index. The first case study investigates whether a pooled data source from Global Credit Data (GCD) is representative when considering the enrichment of internal data with pooled data in the development of a regulatory loss given default (LGD) model. The second case study differs from the first by illustrating which other countries in the pooled data set could be representative when enriching internal data during the development of a LGD model. Using these case studies as examples, our proposed methodology provides users with a generalised framework to identify subsets of the external data that are representative of their Country’s or bank’s data, making the results general and universally applicable.

An Enhanced Cascading Model for E-Commerce Consumer Credit Default Prediction

As an important global policy guide to promote economic transformation and upgrading, the upsurge of E-Commerce has been continuously upgraded with continuous breakthroughs in information technology. In recent years, China’s e-commerce consumer credit has developed well, but due to its short time of production and insufficient experience for reference, credit risk, fraud risk, and regulatory risk continue to emerge. Aiming at the problem of E-Commerce Consumer Credit default analysis, this paper proposes a Fusion Enhanced Cascade Model (FECM). This model learns feature data of credit data by fusing multi-granularity modules, and incorporates random forest and GBDT trade-off variance and bias methods. The paper compares FECM and gcForest on multiple data sets, to prove the applicability of FECM in the field of E-commerce credit default prediction. The research results of this paper are helpful to the risk control of financial development, and to construct a relatively stable financial space for promoting the construction and development of E-Commerce.

An Enhanced Cascading Model for E-Commerce Consumer Credit Default Prediction

As an important global policy guide to promote economic transformation and upgrading, the upsurge of E-Commerce has been continuously upgraded with continuous breakthroughs in information technology. In recent years, China’s e-commerce consumer credit has developed well, but due to its short time of production and insufficient experience for reference, credit risk, fraud risk, and regulatory risk continue to emerge. Aiming at the problem of E-Commerce Consumer Credit default analysis, this paper proposes a Fusion Enhanced Cascade Model (FECM). This model learns feature data of credit data by fusing multi-granularity modules, and incorporates random forest and GBDT trade-off variance and bias methods. The paper compares FECM and gcForest on multiple data sets, to prove the applicability of FECM in the field of E-commerce credit default prediction. The research results of this paper are helpful to the risk control of financial development, and to construct a relatively stable financial space for promoting the construction and development of E-Commerce.

Facing the Challenges of Developing Fair Risk Scoring Models

Algorithmic scoring methods are widely used in the finance industry for several decades in order to prevent risk and to automate and optimize decisions. Regulatory requirements as given by the Basel Committee on Banking Supervision (BCBS) or the EU data protection regulations have led to an increasing interest and research activity on understanding black box machine learning models by means of explainable machine learning. Even though this is a step into a right direction, such methods are not able to guarantee for a fair scoring as machine learning models are not necessarily unbiased and may discriminate with respect to certain subpopulations such as a particular race, gender, or sexual orientation—even if the variable itself is not used for modeling. This is also true for white box methods like logistic regression. In this study, a framework is presented that allows analyzing and developing models with regard to fairness. The proposed methodology is based on techniques of causal inference and some of the methods can be linked to methods from explainable machine learning. A definition of counterfactual fairness is given together with an algorithm that results in a fair scoring model. The concepts are illustrated by means of a transparent simulation and a popular real-world example, the German Credit data using traditional scorecard models based on logistic regression and weight of evidence variable pre-transform. In contrast to previous studies in the field for our study, a corrected version of the data is presented and used. With the help of the simulation, the trade-off between fairness and predictive accuracy is analyzed. The results indicate that it is possible to remove unfairness without a strong performance decrease unless the correlation of the discriminative attributes on the other predictor variables in the model is not too strong. In addition, the challenge in explaining the resulting scoring model and the associated fairness implications to users is discussed.

autopsych: An R Shiny tool for the reproducible Rasch analysis, differential item functioning, equating, and examination of group effects

In this paper, we present autopsych, a novel online tool that allows school assessment experts, test developers, and researchers to perform routine psychometric analyses and equating of student test data and to examine the effect of student demographic and group conditions on student test performance. The app extends current open-source software by providing (1) extensive embedded result narration and summaries for written reports, (2) improved handling of partial credit data via customizable item-person Wright maps, (3) customizable item- and person-flagging systems, (4) item-response theory model constraints and controls, (5) many-facets Rasch analysis to examine item bias, (6) Rasch fixed item equating for mapping student ability across test forms, (7) tabbed spreadsheet outputs and immediate options for secondary data analysis, (8) customizable graphical color schemes, (9) extended ANOVA analysis for examining group differences, and (10) inter-rater reliability analyses for the verifying the consistency of rater scoring systems. We present the app’s architecture and functionalities and test its performance with simulated and real-world small-, medium-, and large-scale assessment data. Implications and planned future developments are also discussed.

Effects of partial quantity rationing of credit on technical efficiency of Boro rice growers in Bangladesh: Application of the stochastic frontier model

In this study, we analyzed the effects of the partial quantity rationing of credit on the technical efficiency of Boro rice growers in the Pabna district of Bangladesh. Before conducting the field survey, we designed a theoretical framework and identified farm households affected by the partial quantity rationing of credit. Data were collected from 174 Boro rice growers and analyzed in two stages, where the technical efficiency of Boro rice growers was assessed using stochastic frontier analysis, and the inefficiency effects model was then applied to evaluate determinants of the technical efficiency. The mean technical efficiency of Boro rice growers was 78%, which indicates that their technical efficiency was 22% beyond the production frontier curve. The variables comprising the household head’s age, education level, seed quality, formal training, access to the market, farm labor, tillage cost, fertilizer cost, irrigation cost, and price of seedlings significantly affected the technical efficiency of rice growers. The variables of interest comprising the rate and partial quantity rationing of credit had significant negative effects on the technical efficiency of rice growers. The findings obtained in this study will help to enhance the actual production level using the available resources and improve the food security situation in Bangladesh.