RISKAUTOMATIC PREDICTION FOR SOCIAL ECONOMY COMPANIES USING CAMELS

2025-04-15 0 0 287.18KB 9 页 10玖币

侵权投诉

RISK AUTOMATIC PREDICTION FOR SOCIAL ECONOMY

COMPANIES USING CAMELS ∗

Joseph A. Gallego-Mejia 1, Daniela Martin-Vega 2, Fabio A. González 3

MindLab

Universidad Nacional de Colombia1,3, Universidad Distrital Francisco Jose de Caldas2

Bogotá, Colombia

{jagallegom1,fagonzalezo3}@unal.edu.co dmartinv@correo.udistrital.edu.co2

ABSTRACT

Governments have to supervise and inspect social economy enterprises (SEEs). However, inspecting

all SEEs is not possible due to the large number of SEEs and the low number of inspectors in general.

We proposed a prediction model based on a machine learning approach. The method was trained with

the random forest algorithm with historical data provided by each SEE. Three consecutive periods of

data were concatenated. The proposed method uses these periods as input data and predicts the risk

of each SEE in the fourth period. The model achieved 76% overall accuracy. In addition, it obtained

good accuracy in predicting the high risk of a SEE. We found that the legal nature and the variation

of the past-due portfolio are good predictors of the future risk of a SEE. Thus, the risk of a SEE in a

future period can be predicted by a supervised machine learning method. Predicting the high risk of a

SEE improves the daily work of each inspector by focusing only on high-risk SEEs.

Keywords

Machine Learning

Random Forest

Forecasting

Regression

CAMELS

Risk

Prediction.

1 Introduction

Reviewing, monitoring and inspecting are important tasks for governments when evaluating social economy enterprises

(SEEs). Social economy enterprises are a group of private or semi-private enterprises in which proﬁt distribution and

decision making are not related to the asset ownership of each member [9]. Governments have to inspect these SEEs

and are responsible for categorizing and measuring companies that have medium-high risk and medium-high severity.

Risk relates to the likelihood of a company going into ﬁnancial failure, i.e., being unable to pay its debts. Severity

measures the relative impact of a potential SEE bankruptcy within a relative area, e.g., a town, city or state. It should be

noted that all SEEs cannot be inspected on-site due to limitations on the number of inspectors in a given time. Therefore,

a good analysis ﬁnding medium-high risk SEEs should be performed by supervisors without an on-site inspection. In

this paper, we present a model that measures the future risk of ﬁnancial failure of an SEE using: its ﬁnancial history;

operational reports; the CAMELS rating system, which is a system created in the U.S. to evaluate the overall condition

of its banks; and macroeconomic indicators, such as the consumer price index.

This paper presents a novel model to predict the risk of ﬁnancial failure of an SEE in the next period considering

historical ﬁnancial and operational data. This predictive model is built using machine learning models with the help of

historical ﬁnancial and operational data provided by each SEE. A preprocessing step was performed in order to clean

and transform the raw data. This new prototype model shows the opportunities offered by machine learning for the

decision making process. Our main hypothesis is that the new model will be a valuable tool for prioritizing which SEE

requires inspection by government inspectors.

∗Citation:Joseph et al., Risk Automatic Prediction for Social Economy Companies using Camels.

arXiv:2210.05052v1 [cs.LG] 10 Oct 2022

Risk Automatic Prediction for Social Economy Companies using Camels

The rest of this paper is organized as follows. Section 2 reviews related work on bankruptcy prediction and the CAMELS

system. Section 3 presents the description of the data. Section 4 presents the description of the proposed model. Section

5 presents the experimental evaluation of the proposed model. Section 6 presents the results and discussions achieved

with the proposed model. Finally, Section 7 presents the conclusions.

2 Background and Related Work

Social economy enterprises are really important for healthy societies; they provide several jobs and sustain part of the

economy. Assessing their performance and possible bankruptcy is a crucial task for governments. Therefore, several

machine learning methods have been applied to predict bankruptcy [13, 15, 18]. These algorithms are based on ﬁnancial

ﬁgures such as balance sheet and income statements, company-speciﬁc variables and stock market data. In addi-

tion, for a more general evaluation of a company, it is necessary to evaluate the qualitative attributes of the company [17].

Several types of methods for predicting bankruptcy have been proposed in the state-of-the-art. In [6], the authors used

decision trees to predict whether a given ﬁrm will go bankrupt in the following year. They showed that decision trees

are better than logit regression. In [12], the authors built a model to predict the distress of a given company. They used

logit, probit, multivariate discriminant analysis and artiﬁcial neural networks, showing that logit and probit are good, as

they have explainable and understandable properties. In contrast, artiﬁcial neural networks performed well but had

poor explainable and understandable properties. In [4], the authors evaluated 3,000 companies in Romania in eight

categories, from AAA to D. They tried to predict their probability of downward transition from one year to the next.

They used logit regression and artiﬁcial neural network and showed that artiﬁcial neural networks perform better than

logit regression. In [19], the authors supplemented traditional data, such as ﬁnancial ﬁgures, with quality data, such

as a company sharing director or senior managers. With these relationships, a neighborhood prediction model was

built. They showed that when quality data is combined with ﬁnancial data, the performance of the algorithm increases.

In [17], the authors evaluated four machine learning techniques: decision trees, neural networks, random forest, and

logistic regression. They showed that the random forest outperforms the other methods for bankruptcy prediction.

Financial datasets pose various problems, for example, the ratio of bankrupt to non-bankrupt companies is imbalanced,

i.e., there are more non-bankrupt companies. In this case, the algorithm cannot be evaluated with the accuracy metric

alone and more robust metrics such as f1-score, precision or recall can be used. Therefore, in [21], the authors propose

a comparison between three algorithms that deal with imbalance: probabilistic least-squares classiﬁcation for outlier

detection, an isolation forest, and a one-class support vector machine. They show that probabilistic least-squares

classiﬁcation (LSDA) outperforms the other two methods. The size of the company is also a really important factor to

take into account. Another problem is to evaluate small, medium and large companies with the same model. In [10], the

authors showed that ﬁrm size is a good predictor variable for the success or failure of a ﬁrm [6].

The CAMELS model was developed in the United States in 1991 to assess the overall performance of banks through

capital adequacy, assets, management capacity, earnings, liquidity and sensitivity to market risk. This model gives the

overall health and performance on a rating system between one and ﬁve, where one implies safe performance and ﬁve

implies unsatisfactory performance, of the banking system, but can be used to evaluate other types of companies such

as SEEs [1]. In [7], the authors developed an advanced CAMELs model with a Supervisory Risk Assessment and Early

Warning Systems to evaluate banks over a two-year period. In [20], the authors developed a new method to calculate the

probability of migrating from a low to a high level of risk based on CAMELS. In [3], the authors evaluate the possibility

of predicting changes in bank ratings in the following years by showing that they could predict some variables such as

capital adequacy, and leverage. In [5], the authors used the CAMELS system to evaluate the performance of Turkish

banks in the period from 2001 to 2008, showing that a strong liquidity ratio signiﬁes overall good health. In [14], the

authors analyzed the development of the public and private bank sector in India. They showed that CAMELS is a good

rating system for assessing bank performance. In [11], the authors evaluate several Indian banks using CAMELS and

showing that 95% of the change is given by debt-to-equity ratio, loan-to-deposit ratio, income per employee, capital

adequacy ratio, and total investment-to-total assets ratio.

3 Data Description

The Colombian government collects data from each SEE at regular intervals. Some companies have to send information

monthly, others every six months. The type of data collected are total assets, total savers, total employees, total

associates, ﬁnancial portfolio per debtor, total income in the period, among others. Each company receives a risk

label ranging from one to ﬁve, where one indicates low risk and ﬁve indicates high risk. The process of labeling each

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

RISKAUTOMATICPREDICTIONFORSOCIALECONOMYCOMPANIESUSINGCAMELSJosephA.Gallego-Mejia1,DanielaMartin-Vega2,FabioA.González3MindLabUniversidadNacionaldeColombia1;3,UniversidadDistritalFranciscoJosedeCaldas2Bogotá,Colombia{jagallegom1,fagonzalezo3}@unal.edu.codmartinv@correo.udistrital.edu.co2ABSTRACTGove...

展开>> 收起<<

RISKAUTOMATIC PREDICTION FOR SOCIAL ECONOMY COMPANIES USING CAMELS.pdf

共9页,预览2页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

RISKAUTOMATIC PREDICTION FOR SOCIAL ECONOMY COMPANIES USING CAMELS

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: