台州公共数据企业评分模型

通过台州公开网站的数据采集渠道，获取企业的工商信息、年度报告、欠税信息、企业变更记录以及是否被列入经营异常名录等关键信息。基于上述企业数据，对数据进行清洗、比对、加工，深入分析企业的经营状况、正面、负面信息，从而精准计算出企业的信用评分，有助于金融机构有效地识别企业属性，并为不同属性的企业提供更灵活的贷款利率和贷款授信额度。
在企业信用评分体系中，将企业划分为四个等级：差（低于40分）、中（40-70分）、良（70-90分）及优（90-100分）。对于信用评分处于差等（即低于40分）的企业，金融机构可在风控审核阶段直接拒绝其贷款申请，降低金融风险。中等信用评分的企业（40-70分），则可通过基本的贷款审核流程。对于评分达到良好（70-90分）的企业，金融机构可适当增加其授信额度。信用评分达到优秀的企业（90-100分），企业可获得较低的贷款利率及较高额度的贷款支持，以激励企业持续健康发展。
此外，若企业存在经营异常状况或欠税金额超过100万元等严重问题，被视为一票否决。对于连续获得A级纳税评级的企业或科技创新型企业，金融机构将根据内部政策给予适当优惠，以支持企业快速成长与创新发展。1、数据采集：通过台州公开网站数据采集企业工商、年报、纳税、欠税、变更、异常名录信息。2、算法规则：(1)是否为A级纳税企业：被评为A级纳税企业记为1否则为2； (2)是否为科创企业: 根据《省级企业技术信息》是科创企业为1否为2; (3)是否欠税：根据《市企业欠税信息》欠费是为1否为2; (4)欠税税额：sum（欠税税额）; (5)否为异常企业: 根据《市企业异常名录信息》，异常是为1否为2； (6)企业变更次数:根据《企业变更信息》统计变更次数； (7)企业评分: 特征值设置：A级纳税企业，科创企业，是否欠税，欠税税额，异常企业，企业变更次数。数据处理：对特征值进行缺失值填充、异常值处理。模型开发：以欠税税额和企业变更次数进行分箱合并处理后的结果作为模型的Y标签，采用逻辑回归方法、机器学习算法进行建模，取模型效果最优者为最终评分，得到企业评分：TRANSFORM(数据集) , PREDICT（训练结果）*100% 。企业评分，低于40分的为差，可拒绝贷款申请；40-70为中，可通过贷款申请；70-90为良，可适当提升授信额度。90-100为优，可适当降低贷款利率，提高授信额度。

Data collection was conducted through Taizhou's public official website channels to obtain key enterprise information including industrial and commercial registration details, annual reports, tax arrears records, enterprise change history, and whether the enterprise is included in the business operation abnormal list.
Based on the aforementioned enterprise data, data cleaning, matching and processing were performed, followed by in-depth analysis of the enterprise's operating status, positive and negative information, to accurately calculate the enterprise's credit score. This helps financial institutions effectively identify enterprise attributes, and provide more flexible loan interest rates and loan credit lines for enterprises with different attributes.

In the enterprise credit scoring system, enterprises are divided into four levels: Poor (below 40 points), Medium (40-70 points), Good (70-90 points), and Excellent (90-100 points). For enterprises with a Poor credit score (i.e., below 40 points), financial institutions can directly reject their loan applications during the risk control review stage to reduce financial risks. For enterprises with a Medium credit score (40-70 points), their loan applications can pass the basic loan review procedures. For enterprises with a Good credit score (70-90 points), financial institutions can appropriately increase their credit lines. For enterprises with an Excellent credit score (90-100 points), they can obtain lower loan interest rates and higher loan amount support to incentivize their sustainable and healthy development.

In addition, enterprises with serious issues such as business operation abnormal status or tax arrears amount exceeding 1 million RMB are subject to one-vote veto. For enterprises that have continuously obtained Class A tax payment ratings or technological innovation-oriented enterprises, financial institutions will provide appropriate preferential treatments in accordance with internal policies to support their rapid growth and innovative development.

1. Data Collection: Collect enterprise industrial and commercial registration, annual report, tax payment, tax arrears, change history, and abnormal list information through Taizhou's public official websites.
2. Algorithm Rules:
(1) Whether it is a Class A tax payment enterprise: Record as 1 if rated as a Class A tax payment enterprise, otherwise 2;
(2) Whether it is a technological innovation enterprise: Record as 1 if it is a technological innovation enterprise based on "Provincial Enterprise Technical Information", otherwise 2;
(3) Whether there is tax arrears: Record as 1 if there is tax arrears based on "Municipal Enterprise Tax Arrears Information", otherwise 2;
(4) Tax arrears amount: Sum of tax arrears amounts;
(5) Whether it is an abnormal enterprise: Record as 1 if the enterprise is in the abnormal list based on "Municipal Enterprise Abnormal List Information", otherwise 2;
(6) Number of enterprise changes: Count the number of changes based on "Enterprise Change Information";
(7) Enterprise Credit Score: Feature settings: Class A tax payment enterprise, technological innovation enterprise, whether there is tax arrears, tax arrears amount, abnormal enterprise, number of enterprise changes.

Data Processing: Perform missing value imputation and outlier handling on the feature values.

Model Development: Use the results after binning and merging processing of tax arrears amount and number of enterprise changes as the Y label of the model. Adopt logistic regression and other machine learning algorithms for modeling, select the model with the best performance as the final scoring model, and obtain the enterprise credit score: TRANSFORM(Dataset), PREDICT(Training Result)*100%.

Enterprise Credit Score: Below 40 points is Poor, and the loan application can be rejected; 40-70 points is Medium, and the loan application can pass the review; 70-90 points is Good, and the credit line can be appropriately increased; 90-100 points is Excellent, and the loan interest rate can be appropriately reduced while increasing the credit line.

搜集汇总

数据集介绍

特点

台州公共数据企业评分模型数据集包含3071条企业信息，每月更新，用于计算企业信用评分。数据集涵盖企业工商、纳税、欠税等关键信息，通过机器学习算法评分，支持金融机构进行贷款决策。