CN114202432A - Method and device for evaluating risk of private fund raising and storage medium - Google Patents
Method and device for evaluating risk of private fund raising and storage medium Download PDFInfo
- Publication number
- CN114202432A CN114202432A CN202210148448.2A CN202210148448A CN114202432A CN 114202432 A CN114202432 A CN 114202432A CN 202210148448 A CN202210148448 A CN 202210148448A CN 114202432 A CN114202432 A CN 114202432A
- Authority
- CN
- China
- Prior art keywords
- historical
- information
- private
- product
- risk
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/06—Asset management; Financial planning or analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- General Engineering & Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Entrepreneurship & Innovation (AREA)
- Medical Informatics (AREA)
- Evolutionary Computation (AREA)
- Human Resources & Organizations (AREA)
- Operations Research (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- General Business, Economics & Management (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention discloses a method and a device for evaluating the risk of a private fund and a readable storage medium, wherein the method comprises the steps of obtaining historical product attribute information and historical risk grade information of various historical private products; performing data cleaning on historical product attribute information of each historical private product corresponding to each item, and acquiring historical initial effective parameter information of each historical private product; extracting a training data set and optimal GBDT model parameters from historical risk level information and all historical initial effective parameter information by adopting a GBDT algorithm; constructing a training GBDT model according to the training data set and the optimal GBDT model parameters; and applying the training GBDT model to the current private products, and predicting the current risk level of the current private products. According to the method for evaluating the risk of the private fund, the risk rating information of the private product to be evaluated is automatically generated, manual intervention is reduced, and the working efficiency is improved.
Description
Technical Field
The invention relates to the technical field of fund evaluation methods, in particular to a method and a device for evaluating risk of private fund and a storage medium.
Background
The risk rating of the private products is a relatively tedious work, and needs to comprehensively consider information in various aspects such as company performance, product strategies, product manager performance and the like.
At present, information such as company performance, product strategies, and performance of past products of a product manager under the strategies needs to be analyzed and compared manually, so that risks of private products to be invested are graded and classified. The following disadvantages exist when using manual analysis: in the process of analyzing and deciding the data, the time consumption is long, and the analysis efficiency is low; manual risk rating may have problems of misjudgment and poor timeliness; the flexibility is not enough, and if the rating rule is sent and changed, the risk rating information of the existing product is difficult to be synchronously updated in a large scale.
Disclosure of Invention
The invention provides a method for rating risk of private fund raising, which aims to solve the technical problems of low analysis efficiency and insufficient flexibility caused by the fact that the existing artificial analysis is adopted for rating and classifying the private fund raising products.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a method for rating risk of private fund raising comprises the following steps: acquiring historical product attribute information and historical risk level information of various historical private products; performing data cleaning on historical product attribute information of each historical private product corresponding to each item, and acquiring historical initial effective parameter information of each historical private product; extracting a training data set and optimal GBDT model parameters from historical risk level information and all historical initial effective parameter information by adopting a GBDT algorithm; constructing a training GBDT model according to the training data set and the optimal GBDT model parameters; and applying the training GBDT model to the current private products, and predicting the current risk level of the current private products.
Further, obtaining historical product attribute information for each historical item of the private recruit includes: acquiring historical product strategy attribute information corresponding to various historical private products; acquiring historical company performance information of each fund company under the same historical product strategy attribute information; and acquiring the historical expression information of the managers of each product manager under the same historical product strategy attribute information.
Further, the data cleaning of the historical product attribute information of the corresponding historical personal products includes the following steps: preprocessing historical product strategic attribute information corresponding to each historical private product to acquire initial effective strategic attribute information in the historical product strategic attribute information; preprocessing the historical company performance information corresponding to each historical private product, determining first abnormal data from all the historical company performance information corresponding to each historical private product, filtering the first abnormal data to obtain the initial effective historical performance information corresponding to each historical private product, or replacing the first abnormal data with the average value of the historical company performance information of fund companies under the same historical product strategy attribute information to obtain the initial effective historical performance information corresponding to each historical private product; preprocessing the historical performance information of the managers corresponding to the historical private products, determining second abnormal data from all the historical performance information of the managers corresponding to the historical private products, filtering the second abnormal data to obtain the initial effective information of the historical performance corresponding to the historical private products, or replacing the second abnormal data with the average value of the historical performance of the managers under the product strategy attribute to obtain the initial effective information of the historical performance of the historical private products.
Further, the first abnormal data is the historical performance information of the company which is not within the preset range under the same historical product strategy attribute information.
Further, the extracting of the training data set and the optimal GBDT model parameters from the historical risk level information and all historical initial effective parameter information by using the GBDT algorithm includes: taking historical risk grade information and all historical initial effective parameter information as a training set, and constructing an initial GBDT model by using a grid search method; analyzing and obtaining the importance of the weight of each historical initial effective parameter information by using an initial GBDT model; sequentially arranging all the historical initial effective parameter information according to the importance of the weight of each piece of historical initial effective parameter information from large to small; sequentially carrying out accumulative combination according to the importance of the weights from large to small to obtain a plurality of groups of combined weights and constructing a corresponding simulated GBDT model; obtaining model integrals of the simulated GBDT models corresponding to each group of combined weights; and obtaining historical initial effective parameter information corresponding to the simulated GBDT model with the highest model integral as a training data set, and obtaining model parameters corresponding to the simulated GBDT model with the highest model integral as optimal GBDT model parameters.
Further, applying the trained GBDT model to the current private products in step, comprising after predicting a current risk level of the current private products: judging whether the current risk level is greater than a preset risk level or not; and when the current risk level is greater than the preset risk level, outputting and displaying the training data set.
Further, the initial effective policy attribute information includes investment policy parameter information and investment term parameter information.
Further, the historical risk level information includes first-level risk information, second-level risk information, third-level risk information, fourth-level risk information and fifth-level risk information.
The invention also provides a device for rating the risk of the private fund, which comprises an acquisition unit, a data filtering unit, a model training unit, a model building unit and an application processing unit, wherein the acquisition unit is used for acquiring the historical product attribute information and the historical risk grade information of each historical private fund; the data filtering unit is used for performing data cleaning on the historical product attribute information of the historical private products corresponding to the various types of historical private products and acquiring historical initial effective parameter information of each type of historical private products; the model training unit is used for extracting a training data set and optimal GBDT model parameters from the historical risk level information and all historical initial effective parameter information by adopting a GBDT algorithm; the model building unit is used for building a training GBDT model according to the training data set and the optimal GBDT model parameters; the application processing unit is used for applying the training GBDT model to the current private products and predicting the current risk level of the current private products.
The invention also provides a readable storage medium storing a computer program for executing the method for privacy fund risk rating according to any one of the above.
The invention has the following beneficial effects:
according to the method for rating the risk of the private fund, historical product attribute information and historical risk grade information of each type of historical private fund are obtained, a machine learning method is adopted, the historical product attribute information is used as training data, the risk grade corresponding to the historical product attribute information is used as a training result to obtain a training GBDT model, then the training GBDT model is applied to the current private fund, the current risk grade of the current private fund is predicted, the risk grade information of the private fund to be evaluated is automatically generated, manual intervention is reduced, and the working efficiency is improved; the risk rating information of the private products can be updated in batches by using machine learning, and time delay and misjudgment do not exist; the used prediction model (the training GBDT model) can be dynamically adjusted, the rating information of the existing private products can be updated in batches, and the method is flexible and convenient.
In addition to the objects, features and advantages described above, other objects, features and advantages of the present invention are also provided. The present invention will be described in further detail below with reference to the drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:
FIG. 1 is a flow diagram of a method of privacy fund risk rating of an embodiment of the present invention;
FIG. 2 is a flow diagram of obtaining historical product attribute information for historical personal products of various types in accordance with one embodiment of the invention;
FIG. 3 is a schematic diagram of the principle of extracting training data sets and optimal GBDT model parameters according to one embodiment of the present invention;
FIG. 4 is a flow chart of extracting training data sets and optimal GBDT model parameters according to one embodiment of the present invention;
fig. 5 is a schematic structural diagram of a private fund risk rating apparatus according to an embodiment of the present invention.
Illustration of the drawings:
11. an acquisition unit; 12. a data filtering unit; 13. a model training unit; 14. a model construction unit; 15. an application processing unit.
Detailed Description
The embodiments of the invention will be described in detail below with reference to the accompanying drawings, but the invention can be embodied in many different forms, which are defined and covered by the following description.
As shown in fig. 1, the method for rating the risk of the personal fund in the embodiment includes the following steps:
s101, acquiring historical product attribute information and historical risk level information of each historical private product;
s102, performing data cleaning on historical product attribute information of each historical private product corresponding to each historical private product, and acquiring historical initial effective parameter information of each historical private product;
s103, extracting a training data set and optimal GBDT model parameters from the historical risk level information and all historical initial effective parameter information by adopting a GBDT algorithm;
s104, constructing a training GBDT model according to the training data set and the optimal GBDT model parameters;
and S105, applying the training GBDT model to the current private products, and predicting the current risk level of the current private products.
According to the method for grading the risks of the private fund, the historical initial effective parameter information and the historical risk grade information of each historical private fund are obtained, a machine learning method is adopted, a GBDT algorithm is adopted to extract a training data set and optimal GBDT model parameters from the historical risk grade information and all the historical initial effective parameter information, a training GBDT model is constructed according to the training data set and the optimal GBDT model parameters, then the training GBDT model is applied to the current private fund, the current risk grade of the current private fund is predicted, the risk grade of the private fund to be evaluated is automatically completed, manual intervention is reduced, and the working efficiency is improved; the risk rating information of the private products can be updated in batches by using machine learning, and time delay and misjudgment do not exist; the used GBDT model can be dynamically adjusted, the rating information of the existing private products can be updated in batches, and the method is flexible and convenient.
It can be understood that the machine learning method is adopted for carrying out the risk rating of the private fund, if the rating rule changes, the machine learning can be adaptive to the new rating rule to carry out synchronous updating so as to correctly evaluate the current risk level of the private fund or update the historical risk level of the private fund, and the flexibility is high.
Referring to fig. 2, further, since the risk ratings of the private products are generally related according to information such as product policies, historical performance of companies, and performance of past products of product managers under the policies, in order to ensure objectivity of risk assessment, step S101, acquiring historical product attribute information of each historical private product includes: s1011, obtaining historical product strategy attribute information corresponding to each historical private product; s1012, acquiring historical company performance information of each fund company under the same historical product strategy attribute information; and S1013, acquiring historical performance information of managers of each product manager under the same historical product strategy attribute information.
Optionally, the effective policy attribute information includes investment policy parameter information, investment term parameter information, and whether parameter information is structured, where the investment policy parameter information indicates a product policy selected by a current private product, such as a stock policy, a commodity-period goods policy, a bond policy, and the like, which indicates an investment style of the private product, and the investment term parameter information indicates an investment term selected by the current private product, and includes a 1 year period, a 2 year period, or a 3 year period.
Optionally, the historical company performance information includes average annual income parameter information of the fund company, average annual income rate parameter information of the fund company, average annual fluctuation rate parameter information of the fund company, average maximum withdrawal parameter information of the fund company, average descending risk parameter information of the fund company, average sharp rate parameter information of the fund company, average soyono rate parameter information of the fund company, and the historical company performance information of the fund company is evaluated by adopting the average annual income parameter information of the fund company, the average annual income rate parameter information of the fund company and the like, so that interference in the economic oscillation period is reduced, and objectivity in evaluating the risk rating of the current private fund is improved; meanwhile, historical performance information of the assessment company is acquired from multiple dimensions, so that the accuracy of assessing the risk rating of the current private fund is improved.
Optionally, the manager historical performance information includes manager average annual income parameter information, manager average annual fluctuation rate parameter information, manager average maximum withdrawal parameter information, manager average descending risk parameter information, manager average sharp ratio parameter information, manager average suggeno ratio parameter information, manager working age limit parameter information and manager working duration parameter information of the product manager under the same historical product strategy attribute information. The manager historical performance information of the product manager is evaluated by adopting manager average annual income parameter information, manager average annual income rate parameter information and the like, so that the interference of economic oscillation period is reduced, and the objectivity of evaluating the current private fund risk rating is improved; meanwhile, historical performance information of the assessment manager is acquired from multiple dimensions, so that the accuracy of assessing the risk rating of the current private fund is improved.
Understandably, in the present invention, historical withdrawal is used to gauge the risk-resistance of the private product, describing the maximum loss that an investor may face; the sharp ratio is used for measuring the excess return rate obtained by the fund assuming the unit risk, namely the total return rate of the fund is higher than the part of the risk-free return rate in the same period, and the higher the ratio is, the higher the excess return rate obtained by the fund assuming the unit risk is; the Sorano ratio is used for distinguishing the quality of fluctuation and measuring the excess return rate which can be obtained by the same unit downlink risk in the calculation period, and the higher the ratio is, the higher the excess return rate which is obtained by the fund bearing the unit downlink risk is; the GBDT fully-symmetric gradient descent tree is an algorithm for classifying or regressing data by adopting an additive model (namely linear combination of basis functions) and continuously reducing residual errors generated in a training process.
Further, since the history data acquired in step S101 is not completely accurate, dirty data may exist therein, and it is necessary to perform an abnormal value analysis on the data acquired in step S101 to reduce interference caused by the dirty data. In the invention, abnormal data can be cleaned and filtered by analyzing the average value, standard deviation, correlation and the like of the comparison data, so that useful data can be reserved; or substituting the average value for the outlier data to form useful data. In step S102, the data cleaning of the historical product attribute information of the corresponding historical personal products, and the obtaining of the historical initial effective parameter information of each historical personal product includes: preprocessing historical product strategic attribute information corresponding to each historical private product to acquire initial effective strategic attribute information in the historical product strategic attribute information; preprocessing the historical company performance information corresponding to each historical private product, determining first abnormal data from all the historical company performance information corresponding to each historical private product, filtering the first abnormal data to obtain the initial effective historical performance information corresponding to each historical private product, or replacing the first abnormal data with the average value of the historical company performance information of fund companies under the same historical product strategy attribute information to obtain the initial effective historical performance information corresponding to each historical private product; preprocessing the historical performance information of the managers corresponding to the historical private products, determining second abnormal data from all the historical performance information of the managers corresponding to the historical private products, filtering the second abnormal data to obtain the initial effective information of the historical performance corresponding to the historical private products, or replacing the second abnormal data with the average value of the historical performance of the managers under the product strategy attribute to obtain the initial effective information of the historical performance of the historical private products.
Further, the first abnormal data is the historical performance information of the company which is not within the preset range under the same historical product strategy attribute information. For example, the maximum company average annual income parameter and the minimum company average annual income parameter in the company average annual income parameter information are filtered out or replaced by the average value of the company average annual income parameter; for example, if the historical private fund is a bond strategy private fund, the fund yield is not particularly high, and the average annual income parameter of the company, which is higher than the average annual income parameter of the theoretical company in the average annual income parameter information of the company, is filtered out or replaced by the average value of the average annual income parameter of the company.
Illustratively, the second abnormal data is the historical performance information of the manager which is not within the preset range under the same historical product strategy attribute information. For example, the maximum manager average annual income parameter and the minimum manager average annual income parameter in the manager average annual income parameter information are filtered out or replaced by the average value of the manager average annual income parameter; for example, if the historical private fund is a bond strategy private fund, the fund yield is not particularly high, and the manager average annual income parameter higher than the theoretical manager average annual income parameter in the manager average annual income parameter information is filtered out or replaced by the manager average annual income parameter average.
Illustratively, the historical company performance information comprises the average annual income parameter information of the fund company under the same historical product strategy attribute information, the average annual income parameter information of a plurality of companies of the fund company under the same historical product strategy attribute information is obtained, if the average annual income parameter information of one company is not in the preset range, the average annual income parameter information of the company which is not in the preset range is filtered, or the average value of the historical company performance information of the fund company under the same historical product strategy attribute information is adopted to replace the average annual income parameter information of the company which is not in the preset range.
Illustratively, the manager average annual income parameter information of the product managers under the same historical product strategy attribute information is described by the manager historical performance information, the manager average annual income parameter information of the product managers under the same historical product strategy attribute information is obtained, if the manager average annual income parameter information is not in the preset range, the manager average annual income parameter information which is not in the preset range is filtered, or the manager average annual income parameter information which is not in the preset range is replaced by the average value of the manager average annual income parameter information of the product managers under the same historical product strategy attribute information.
Referring to fig. 3 and 4, further, the step S103 of extracting the training data set and the optimal GBDT model parameters from the historical risk level information and all historical initial effective parameter information by using the GBDT algorithm includes: s1031, taking the historical risk grade information and all historical initial effective parameter information as a training set, and constructing an initial GBDT model by using a grid search method; s1032, analyzing and obtaining the weight importance of each historical initial effective parameter information by using an initial GBDT model; s1033, sequentially arranging all the historical initial effective parameter information according to the importance of the weight of each piece of historical initial effective parameter information from big to small; s1034, performing accumulative combination in sequence according to the importance of the weight values from large to small to obtain a plurality of groups of combination weight values and constructing a corresponding simulated GBDT model; s1035, obtaining model integrals of the simulated GBDT models corresponding to each group of combination weights; s1036, obtaining historical initial effective parameter information corresponding to the simulated GBDT model with the highest model integral as a training data set, and obtaining model parameters corresponding to the simulated GBDT model with the highest model integral as optimal GBDT model parameters. In the invention, in order to ensure the accuracy of the subsequently established training model, effective characteristic data is extracted to obtain a training data set and optimal GBDT model parameters, thereby improving the accuracy of the training model.
Further, in step S105, applying the trained GBDT model to the current private recruits, after predicting the current risk level of the current private recruits, includes: step S106, judging whether the current risk level is greater than a preset risk level; and S107, outputting and displaying a training data set when the current risk level is greater than the preset risk level.
Understandably, when the current risk level is greater than the preset risk level, the training data set is output and displayed for the user to select and refer, so that the user can conveniently screen appropriate private fund, and the participation degree of the user is improved, thereby improving the experience of the user.
Further, the historical risk level information includes level one risk (R1) information, level two risk (R2) information, level three risk (R3) information, level four risk (R4) information, and level five risk (R5) information. The private products are divided into 5 grades which are respectively R1, R2, R3, R4 and R5, the represented risk degrees are respectively low, medium and low, medium and high, and investment crowds suitable for purchasing are respectively conservative, robust, balanced, advanced and aggressive.
It is understood that the risk level is preset to be R4, since the higher the risk level indicates the higher the risk of the private product, the higher the risk of the investor purchasing, and when the current risk level is greater than R4, a prompt message is sent out, and a training data set is output and displayed.
It is to be understood that the specific steps of the present invention include acquiring data; performing exception handling analysis on the acquired data, and filtering dirty data in the acquired data; extracting effective characteristic data from the residual data after filtering; constructing a training GBDT model by using the filtered data after the dirty data is taken out; and applying a training GBDT model to predict the risk of the current private fund.
The present invention provides an exemplary risk rating method as follows: wherein the historical product strategy attribute information comprises whether the product is structured, investment strategy and investment period; the historical performance information of the company comprises the average annual profitability of the company and the average maximum withdrawal rate of the company; product manager performance includes manager average annual profitability and manager average maximum withdrawal rate. The method comprises the following steps:
acquiring historical product attribute information and historical risk level information of each historical private product, and referring to table 1, wherein 0 represents unstructured and 1 represents structured in a structured data table;
TABLE 1
And performing data cleaning on the historical product attribute information of the corresponding historical private products, and acquiring historical initial effective parameter information of each historical private product. The invention adopts a plurality of methods to analyze and process abnormal data: if the investment strategy 1 is a bond strategy private fund, the average annual profitability of the company is not particularly high, a value with the particularly high average annual profitability of the company (a value larger than the theoretical average annual profitability of the company) is obtained in the data analysis process, dirty data are shown in the average annual profitability of the company, and the dirty data are filtered or replaced by the average value of the average annual profitability of the company; the second abnormal value analysis method is performed, for example, by filtering a value that cannot be 0 but is 0, data that cannot be negative but is negative, data in which the average value of the same policy is excessively different, and the like.
Referring to fig. 3 and 4, a GBDT algorithm is used to extract a training data set and optimal GBDT model parameters from historical risk level information and all historical initial valid parameter information. Firstly, taking historical risk grade information and all historical initial effective parameter information as a training set, and constructing an initial GBDT model by using a grid search method; secondly, analyzing and obtaining the importance of the weight of each historical initial effective parameter information by using an initial GBDT model, and obtaining the importance ratio of each characteristic value from the initial GBDT model; then, since the feature value with high importance greatly affects the score of the model, the features are arranged from high to low according to the importance as shown in table 2;
TABLE 2
Reorganizing the training data according to the importance of the eigenvalues from large to small to obtain 7 parts of training data: then, using 7 data to perform simulated GBDT training, obtaining a corresponding simulated GBDT model and a corresponding training score (model integral), if 7 GBDT models are respectively model 1, model 2, model 3, model 4, model 5, model 6, and model 7, their training scores are: 0.9, 0.95, 0.96, 0.94, 0.89, 0.92 and 0.93, obtaining a simulated GBDT model with the maximum score (the model integral is 0.96), and obtaining a corresponding training data set (shown in Table 3) and optimal GBDT model parameters;
TABLE 3
And constructing a training GBDT model according to the training data set and the optimal GBDT model parameters.
And finally, applying the training GBDT model to the current private products, and predicting the current risk level of the current private products.
Referring to fig. 5, the present invention further provides a device for rating risk of private funds, including an obtaining unit 11, a data filtering unit 12, a model training unit 13, a model constructing unit 14, and an application processing unit 15, where the obtaining unit 11 is configured to obtain historical product attribute information and historical risk level information of each type of historical private product; the data filtering unit 12 is configured to perform data cleaning on the historical product attribute information of each type of historical private recruitment product, and acquire historical initial effective parameter information of each type of historical private recruitment product; the model training unit 13 is used for extracting a training data set and optimal GBDT model parameters from the historical risk level information and all historical initial effective parameter information by adopting a GBDT algorithm; the model construction unit 14 is used for constructing a training GBDT model according to the training data set and the optimal GBDT model parameters; the application processing unit 15 is configured to apply the training GBDT model to the current private products, and predict the current risk level of the current private products.
According to the private fund risk rating device, the risk rating information of the private products to be evaluated is automatically generated, manual intervention is reduced, and the working efficiency is improved; the risk rating information of the private products can be updated in batches by using machine learning, and time delay and misjudgment do not exist; the used prediction model can be dynamically adjusted, and the rating information of the existing private products can be easily updated in batches, so that the method is flexible and convenient.
The present invention also provides a computer-readable storage medium storing a computer program for executing the above-described method of privacy fund risk rating.
The above is only an embodiment of the present invention, and is not intended to limit the present invention, and it is obvious to those skilled in the art that various modifications and variations can be made in the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method for rating risk of private fund raising is characterized by comprising the following steps:
acquiring historical product attribute information and historical risk level information of various historical private products;
performing data cleaning on the historical product attribute information of each historical private product corresponding to each item, and acquiring historical initial effective parameter information of each historical private product;
extracting a training data set and optimal GBDT model parameters from the historical risk level information and all the historical initial effective parameter information by adopting a GBDT algorithm;
constructing a training GBDT model according to the training data set and the optimal GBDT model parameters;
and applying the training GBDT model to the current private products, and predicting the current risk level of the current private products.
2. The privacy fund risk rating method of claim 1,
the acquiring of the historical product attribute information of each historical private product comprises:
acquiring historical product strategy attribute information corresponding to various historical private products;
acquiring historical company performance information of each fund company under the same historical product strategy attribute information;
and acquiring the historical expression information of the managers of each product manager under the same historical product strategy attribute information.
3. The privacy fund risk rating method of claim 2,
the data cleaning of the historical product attribute information of the corresponding historical private products and the obtaining of the historical initial effective parameter information of each historical private product comprise:
preprocessing the historical product strategic attribute information corresponding to each historical private product to obtain initial effective strategic attribute information in the historical product strategic attribute information;
preprocessing the historical company performance information corresponding to each historical private product, determining first abnormal data from all the historical company performance information corresponding to each historical private product, filtering the first abnormal data to obtain initial effective historical performance information corresponding to each historical private product, or replacing the first abnormal data with an average value of the historical company performance information of fund companies under the same historical product strategy attribute information to obtain the initial effective historical performance information corresponding to each historical private product;
preprocessing the historical performance information of the managers corresponding to the historical private products, determining second abnormal data from all the historical performance information of the managers corresponding to the historical private products, filtering the second abnormal data to obtain initial effective historical performance information corresponding to the historical private products, or replacing the second abnormal data with an average value of the historical performance of the managers under the product strategy attribute to obtain the initial effective historical performance information of the historical private products.
4. The privacy fund risk rating method of claim 3,
the first abnormal data is the historical performance information of the company which is not within a preset range under the same historical product strategy attribute information.
5. The privacy fund risk rating method of claim 3,
the extracting of the training data set and the optimal GBDT model parameters from the historical risk level information and all the historical initial effective parameter information by adopting the GBDT algorithm includes:
taking the historical risk grade information and all the historical initial effective parameter information as a training set, and constructing an initial GBDT model by using a grid search method;
analyzing and obtaining the weight importance of each historical initial effective parameter information by using the initial GBDT model;
sequentially arranging all the historical initial effective parameter information according to the weight importance of each piece of historical initial effective parameter information from large to small;
sequentially carrying out accumulative combination according to the importance of the weights from large to small to obtain a plurality of groups of combined weights and constructing a corresponding simulated GBDT model;
obtaining model integrals of the simulated GBDT models corresponding to each group of the combined weights;
and obtaining the historical initial effective parameter information corresponding to the simulated GBDT model with the highest model integral as a training data set, and obtaining the model parameter corresponding to the simulated GBDT model with the highest model integral as an optimal GBDT model parameter.
6. The privacy fund risk rating method of claim 5,
applying the trained GBDT model to a current personal listing at the step, comprising, after predicting a current risk level for the current personal listing:
judging whether the current risk level is greater than a preset risk level;
and when the current risk level is greater than a preset risk level, outputting and displaying the training data set.
7. The privacy fund risk rating method of claim 3,
the initial effective policy attribute information includes investment policy parameter information and investment deadline parameter information.
8. The privacy fund risk rating method of any one of claims 1-7, wherein the first set of one or more of the second set of one or more of the first set of one or more of the second set of one or more of the first set of one or more of the second set of one or more of the first one or more of the second one or more of the first one or more of the second set of claims,
the historical risk level information comprises first-level risk information, second-level risk information, third-level risk information, fourth-level risk information and fifth-level risk information.
9. A personal fund raising risk rating device is characterized by comprising an acquisition unit, a data filtering unit, a model training unit, a model constructing unit and an application processing unit,
the acquisition unit is used for acquiring historical product attribute information and historical risk level information of each historical private product;
the data filtering unit is used for performing data cleaning on the historical product attribute information of each historical private product, and acquiring historical initial effective parameter information of each historical private product;
the model training unit is used for extracting a training data set and optimal GBDT model parameters from the historical risk level information and all the historical initial effective parameter information by adopting a GBDT algorithm;
the model construction unit is used for constructing a training GBDT model according to the training data set and the optimal GBDT model parameters;
the application processing unit is used for applying the training GBDT model to the current private products and predicting the current risk level of the current private products.
10. A computer-readable storage medium storing a computer program for executing the privacy fund risk rating method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210148448.2A CN114202432A (en) | 2022-02-18 | 2022-02-18 | Method and device for evaluating risk of private fund raising and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210148448.2A CN114202432A (en) | 2022-02-18 | 2022-02-18 | Method and device for evaluating risk of private fund raising and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114202432A true CN114202432A (en) | 2022-03-18 |
Family
ID=80645683
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210148448.2A Pending CN114202432A (en) | 2022-02-18 | 2022-02-18 | Method and device for evaluating risk of private fund raising and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114202432A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108537671A (en) * | 2018-04-27 | 2018-09-14 | 广州品唯软件有限公司 | A kind of transaction risk appraisal procedure and system |
| CN109492783A (en) * | 2018-11-14 | 2019-03-19 | 中国电力科学研究院有限公司 | A kind of Application of Power Metering Instruments failure risk prediction technique based on GBDT |
-
2022
- 2022-02-18 CN CN202210148448.2A patent/CN114202432A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108537671A (en) * | 2018-04-27 | 2018-09-14 | 广州品唯软件有限公司 | A kind of transaction risk appraisal procedure and system |
| CN109492783A (en) * | 2018-11-14 | 2019-03-19 | 中国电力科学研究院有限公司 | A kind of Application of Power Metering Instruments failure risk prediction technique based on GBDT |
Non-Patent Citations (1)
| Title |
|---|
| 田文利: ""私募基金的选择"", 《中国优秀博硕士学位论文全文数据库(硕士)经济与管理科学辑》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110738564A (en) | Post-loan risk assessment method and device and storage medium | |
| CN108229828A (en) | A kind of analysis system based on industrial data | |
| CN105022825A (en) | Financial variety price prediction method capable of combining financial news mining and financial historical data | |
| CN104537433A (en) | Sold electricity quantity prediction method based on inventory capacities and business expansion characteristics | |
| CN103971023B (en) | R＆D process quality automatic evaluation system and method | |
| CN108830417B (en) | ARMA (autoregressive moving average) and regression analysis based life energy consumption prediction method and system | |
| CN114048436A (en) | Construction method and construction device for forecasting enterprise financial data model | |
| CN111651444A (en) | Self-adaptive time series data prediction method | |
| CN105184078A (en) | Technology maturity evaluation method based on patent relative-quantity analysis | |
| CN115952914A (en) | Big data-based electric power metering operation and maintenance work judgment planning method | |
| CN111626844A (en) | Enterprise credit assessment method and device based on big data analysis | |
| CN108122166A (en) | The method sorted out to stock and scored using intelligent quantization financial technology | |
| CN110738565A (en) | Real estate finance artificial intelligence composite wind control model based on data set | |
| CN117291655B (en) | Consumer life cycle operation analysis method based on entity and network collaborative mapping | |
| CN115660608B (en) | One-stop innovative entrepreneurship incubation method | |
| CN117592873A (en) | Data asset value assessment method based on data characteristics | |
| CN117592856A (en) | Algorithm flow based on petrochemical enterprise output prediction | |
| CN117746546A (en) | Service business handling method and system based on number calling device | |
| CN111915188A (en) | Enterprise system performance test method, device and equipment | |
| CN112070336A (en) | Manufacturing industry information quantitative analysis method and device based on analytic hierarchy process | |
| CN114202432A (en) | Method and device for evaluating risk of private fund raising and storage medium | |
| CN108388972A (en) | A kind of integrating skills appraisal procedure and device | |
| CN115186174A (en) | Crowdsourcing task personalized recommendation method and system based on machine learning | |
| CN113537759A (en) | User experience measurement model based on weight self-adaptation | |
| CN118152449A (en) | Method and device for generating analysis report, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220318 |