Financing probability prediction method for private equity right enterprise

Abstract

The invention provides a financing probability prediction method for a private stock right enterprise. The method comprises the following steps: preprocessing an underlying data source of an enterprise in a MySQL database; synchronizing an underlying data source to a Hadoop cluster, and performing secondary association processing to obtain a project advanced screening large-width table and a project investment and financing event large-width table; screening data in a current recent preset time period from the project advanced screening large-width table and the project investment and financing event large-width table, and performing deep processing on the data in a Hive number bin to generate a Hive index table which can be used by a model; establishing a model index table in the MySQL database according to years, and synchronizing the Hive index table into the MySQL database; and receiving a prediction instruction for one or more target enterprises, obtaining model entering index data from the model index table, inputting the model entering index data into the financing probability prediction model, and calculating the financing probability of the one or more target enterprises. The financing probability prediction model adopts related model entering indexes in the MySQL database to perform model training and optimization. Model training follows a complete machine learning model training process, strict processing is carried out in the aspects of labels and sampling, and unique characteristics of business requirements are combined in the aspects of feature engineering and model training. The method can perform quantitative prediction and dynamic evaluation, and is more suitable for mass data prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of private equity enterprise financing, and in particular to a method for predicting the financing probability of a private equity enterprise, a non-transient computer-readable storage medium, and a computer program product. Background Art

[0002] The refinancing capacity assessment of private equity companies is an important process for assessing their credit risk and production and operation health. Such companies usually face greater uncertainty and risk, making their evaluation more complex and challenging. However, the traditional credit assessment model of commercial banks and the marketing and risk control models of third-party big data vendors mostly focus on the assets and operating cash flow of enterprises. Many light-asset technology companies find it difficult to meet credit requirements. Banks need to innovate in the development of science and technology innovation financial business and upgrade the original science and technology innovation loan product model.

[0003] At present, the existing mainstream institutions and academic circles mainly use subjective judgment to evaluate private equity companies, and there is no quantitative analysis and prediction model. The main dimensions of the existing evaluation methods include: financial analysis, business model and market potential, management team capabilities, technological innovation and R&D capabilities, market and industry analysis, legal and compliance, risk assessment, etc. The evaluation method will select the importance factors in each dimension, and give each factor a subjective score or a simple calculation score; then artificially assign importance weights to the dimensions and factors, and finally obtain the overall evaluation results through weighted calculation.

[0004] Existing subjective evaluation methods are highly dependent on the evaluator's own quality and ability, and require high time costs or research costs, so they are difficult to promote or replicate when facing batch tasks. When facing a single sample, the subjective evaluation method may be able to achieve high accuracy; but when facing massive samples and dynamic data, the quantitative model can obtain the global optimal solution or minimum error.

[0005] There are not many models for quantitative prediction analysis in existing evaluation methods, mainly due to the following reasons:

[0006] 1) Lack of financial data: Many private equity firms are in the early stages of business development and lack long-term financial records;

[0007] 2) High market uncertainty: Private equity firms often face a rapidly changing market environment, and their business models and market demands may change rapidly;

[0008] 3) Unsystematic static qualitative evaluation: The status of each enterprise is constantly changing over time. The evaluation methods of existing evaluation agencies cannot reflect the immediate changes of the enterprises invested by the fund, which greatly reduces the reference significance for investors in selecting asset targets;

[0009] 4) The inherent properties of the evaluation factors: The inherent characteristics of the factors in the evaluation method also limit the evaluation method to subjective judgment. Except for the financial analysis and risk assessment factors in the evaluation factors, the remaining factors are mainly affected by the subjective evaluation criteria.

[0010] Therefore, based on these problems, it is of great practical significance to provide a financing probability prediction method based on machine learning models for quantitative prediction and dynamic evaluation, which is more suitable for massive data prediction. Summary of the invention

[0011] The purpose of the present invention is to provide a method for predicting the financing probability of a private equity enterprise, a non-transient computer-readable storage medium, and a computer program product. Compared with traditional subjective evaluation methods, the use of machine learning models for prediction can avoid subjective judgment, explore the value of underlying data, and perform quantitative prediction and dynamic evaluation. It is more accurate and efficient when facing massive data, and the prediction results are more stable and reproducible.

[0012] In order to achieve the above objectives, the present invention is implemented by the following technical solutions:

[0013] A first aspect of the present invention provides a method for predicting the financing probability of a private equity enterprise, comprising:

[0014] Preprocess the underlying data source of the enterprise in the MySQL database;

[0015] Synchronize the underlying data source to the Hadoop cluster, perform secondary association processing, and obtain a large wide table of advanced project screening and a large wide table of project investment and financing events;

[0016] Filter the data within the current preset time period from the project advanced screening large wide table and the project investment and financing event large wide table, perform in-depth processing on the data in the Hive data warehouse, and generate a Hive indicator table that can be used by the model;

[0017] Establish a model indicator table in the MySQL database by year, and synchronize the Hive indicator table to the MySQL database;

[0018] Receive prediction instructions for one or more target enterprises, obtain model input indicator data from the model indicator table, input the data into a pre-built financing probability prediction model, and calculate the financing probability of the one or more target enterprises.

[0019] Optionally, the underlying data sources include: a project basic information table, a project investment and financing data table, and a fund investment project table.

[0020] Optionally, the deep processing of data in the Hive data warehouse includes:

[0021] The values ​​of several original features are determined from the screened data, and the values ​​of several derived features are determined by data derivation according to the values ​​of the several original features.

[0022] Optionally, the derived features are obtained by cross-deriving the original features based on business logic and mathematical functions in advance.

[0023] Optionally, the model entry indicators involve the following three dimensions: the abundance of upstream capital in the track, the heat of track capital, and the company's own financing capabilities.

[0024] Optionally, the financing probability prediction model is an optimal model obtained by fusing the output results of multiple machine learning algorithms based on an integrated learning strategy.

[0025] Optionally, the multiple machine learning algorithms include a logistic regression model, a random forest model, and a gradient boosted decision tree.

[0026] Optionally, it also includes:

[0027] Regularly forecast the financing probability of the one or more target enterprises to achieve dynamic forecasting of the financing capabilities of the enterprises.

[0028] According to a second aspect of the present invention, there is provided a non-transitory computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method of the present invention.

[0029] A third aspect of the present invention provides a computer program product, comprising a computer program, which implements the method of the present invention when executed by a processor.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] (1) The product is exclusive and innovative

[0032] Financial institutions have a need for comprehensive evaluation of science and technology enterprises, but the financial information service institutions in the market are still at the level of only providing underlying data, neither deeply processing the data nor deriving models from the data. Therefore, the financing probability prediction model of the present invention is the first of its kind in the industry. In addition, in terms of meeting business needs, clarifying task objectives, screening underlying data, and selecting model algorithms, the present invention abandons the traditional subjective and manual evaluation methods and develops standardized model products that can meet batch and quantitative evaluation needs, which is innovative.

[0033] (2) The characteristic indicators are exclusive

[0034] The characteristic indicators are the input data of the model. The characteristic indicators of the financing probability prediction model of the present invention are the final results of the inventor's self-research and analysis based on in-depth exchanges with a number of leading banks and investment institutions and the applicant's exclusive private equity database.

[0035] (3) The model technology is scientific and advanced

[0036] Scientificity: Quantitative financial models are based on mathematical theories and systematic models, rather than relying on subjective judgment and emotional decision-making. By using a large amount of historical and real-time data, the financing probability prediction model of the present invention can avoid human emotional interference and provide more objective and accurate judgment results.

[0037] Advancement: Machine learning is a core component of artificial intelligence. Machine learning algorithms refer to the process of using certain algorithms to guide computers to autonomously build reasonable models using known data, and use this model to make judgments on new situations. In current financial scenarios, the most cutting-edge model technology is composed of machine learning algorithms. The financing probability prediction model of the present invention adopts machine learning algorithms as the core component of the model, as well as a series of excellent modeling techniques, such as resampling, feature engineering, Bayesian optimization, probability calibration and other technologies to ensure that the model is advanced and highly competitive. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flowchart of a method for predicting the financing probability of a private equity enterprise provided by one embodiment of the present invention;

[0039] Figure 2a A schematic diagram of the characteristics of the underlying data;

[0040] Figure 2b Schematic diagram of a large wide table for advanced filtering of items;

[0041] Figure 2c This is a schematic diagram of the large table of project investment and financing events;

[0042] Figure 3 A schematic diagram of the characteristics of the mold entry indicator;

[0043] Figure 4a and Figure 4b A schematic diagram showing the effect of the product;

[0044] Figure 5 Schematic diagram of the training process of the financing probability prediction model. DETAILED DESCRIPTION

[0045] The solution proposed by the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0046] Constructing a capital flow forecasting model for private equity financing companies to evaluate the refinancing capacity of companies has become one of the methods to further explore high-quality private equity companies. This type of model can help investment institutions find high-quality investment opportunities, provide comprehensive risk assessment, help investment institutions better understand the growth prospects of private equity companies, and reduce investment risks. By accurately evaluating the refinancing capacity of private equity companies, it can provide a more scientific basis for investment institutions' investment decisions. At the same time, it is also possible to help companies plan and adjust financing plans in advance, help private equity companies better adapt to the market environment, and improve the efficiency and effectiveness of their capital operations.

[0047] Based on the shortcomings of the subjective evaluation methods in the prior art, the present invention proposes a prediction model for the probability of sustainable financing of corporate capital flows, and evaluates scientific and technological innovation enterprises from the perspective of equity investment. The recognition of enterprises by the capital market is evaluated from top to bottom based on the industry capital sufficiency, industry capital heat, and the financing capacity of the enterprise itself. The machine learning prediction model based on the fusion of medium, high and low frequency data predicts the probability of the enterprise obtaining subsequent financing in the next two years, and realizes dynamic quantitative analysis of the enterprise's financing capacity, thereby improving the decision-making efficiency of financial institutions in risk assessment of start-ups and screening of high-quality enterprises.

[0048] Compared with traditional model building methods, machine learning models can avoid subjective judgment, explore the value of underlying data, perform quantitative predictions and dynamic evaluations, and are more accurate and efficient when faced with massive amounts of data. At the same time, the prediction results are more stable and reproducible.

[0049] like Figure 1 As shown, the present invention provides a method for predicting the financing probability of a private equity enterprise, comprising the following steps:

[0050] S110, preprocessing the underlying data source of the enterprise in the MySQL database.

[0051] Figure 2a The characteristics of the underlying data in the MySQL database are that the number of statistical subjects is sufficient, the data dimension is wide, and the dynamic update frequency is high. The underlying data source includes the project basic information table, the project investment and financing data table, and the fund investment project table, which contain all the underlying data of the model feature values. Specifically, the project basic information table includes: the establishment time of the project company, the unified social credit code, the track to which it belongs, etc.; the project investment and financing data table includes the time of occurrence of each financing obtained by the project company, the financing event ID, the financing amount, etc.; the fund investment project table includes the ID of each financing event, the investment fund, and the basic information of the investment fund.

[0052] Preprocessing mainly includes processing missing values, outliers, duplicate values ​​and text data.

[0053] S120, synchronizing the underlying data source to the Hadoop cluster, performing secondary association processing, and obtaining a large wide table for advanced project screening and a large wide table for project investment and financing events.

[0054] Since the data volume of the underlying data source is huge, it is impossible to process such a large amount of data directly in the MySQL database. Therefore, the underlying data source is synchronized to the Hadoop cluster. Hadoop can perform distributed processing on a large amount of data to improve data processing efficiency.

[0055] The obtained project advanced screening large wide table and project investment and financing event large wide table are as follows Figure 2b , Figure 2c shown.

[0056] S130, filter out data within the current most recent preset time period from the project advanced screening large wide table and the project investment and financing event large wide table, perform in-depth processing on the data in the Hive data warehouse, and generate a Hive indicator table that can be used by the model.

[0057] Specifically, the program dynamically obtains the current month as a filtering condition for screening. For example, if the current month is July 2024, July 2024 is used as the evaluation base date to filter out data from the current most recent preset time period (for example, the last three years, i.e. July 2021-July 2024).

[0058] The deep processing of data in the Hive data warehouse includes: determining the values ​​of several original features from the filtered data, and determining the values ​​of several derived features through data derivation based on the values ​​of several original features. The data derivation methods include: indicator filtering, indicator sliding window, aggregate calculation, etc.

[0059] The derived features are obtained by cross-deriving the original features based on business logic and mathematical functions.

[0060] In the present invention, the modeling index of the financing probability prediction model includes original features and derived features, involving the following three dimensions: the sufficiency of upstream capital of the track (or industry), the heat of track capital, and the financing capacity of the enterprise itself. The original features include but are not limited to: the company's registered capital, the track code, the company's historical financing times, the average financing years, the financing window period, the number of registered funds in the track in the past three years, the number of planned listed companies in the track in the past three years, etc.; the derived feature method is mainly cross-derivative, and the derived features include but are not limited to: the company's historical financing times / the company's track code, the number of registered funds in the track in the past three years / the financing window period, the logarithm of the amount of registered funds in the track in the past three years / the number of listed companies in the track in the past three years, the number of well-known investment institutions among the company's shareholders-the number of listed companies in the track in the past three years, etc.

[0061] Figure 3 The characteristics of the modeling indicators are that the modeling indicators are the result of in-depth communication between the applicant and a number of leading banks and investment institutions, based on the applicant's exclusive private equity database, and integrated with the applicant's company's self-developed indicator results.

[0062] S140, creating a model indicator table in the MySQL database by year, and synchronizing the Hive indicator table to the MySQL database.

[0063] Since more concise data that better meets the needs is obtained after calculating and processing massive data in the Hive data warehouse, namely the Hive indicator table, the Hive indicator table is synchronized to a new table in the MySQL database to facilitate subsequent use and updates (MySQL database is more commonly used and general).

[0064] S150, receiving a prediction instruction for one or more target enterprises, obtaining input index data from the model index table, inputting the data into a pre-built financing probability prediction model, and calculating the financing probability of the one or more target enterprises.

[0065] Before making predictions for the target enterprise, the financing probability prediction model can be trained and tuned according to the relevant model entry indicators in the MySQL database. The financing probability prediction model is an optimal model obtained by fusing the output results of multiple machine learning algorithms based on an integrated learning strategy. The multiple machine learning algorithms include a logistic regression model, a random forest model, and a gradient boosted decision tree. The financing probability prediction model is used to predict the probability of obtaining financing in the next period of time based on the model entry indicator data of the target enterprise in the most recent preset time period. The construction process of the financing probability prediction model will be described in detail below.

[0066] After the optimal model is trained, it is deployed in the production environment, so that the financing probability data can be dynamically calculated and output according to user needs. Specifically, the user can enter the name of the private equity company to trigger the prediction instruction for one or more target companies. The optimal model can calculate the classification result (can obtain the next round of financing / cannot obtain the next round of financing) and probability, helping users to quantify the financing ability of the company and meet the actual business needs of users. Furthermore, the financing probability prediction of the one or more target companies can be carried out regularly to realize the dynamic prediction of the financing ability of the company.

[0067] In addition, based on the prediction results, we can further calculate the ranking, quantile and other statistical indicators of each company in its own track, and generate various charts and other visual data. Finally, these contents are produced to the product-side SaaS platform as the final result delivered to users.

[0068] Therefore, in the SaaS product terminal, after the user (i.e. investment institution) enters a single company name or provides a list of batches of companies, and further sets the screening parameters and time range, the prediction results of the financing probability can be obtained. The user can obtain a series of information outputs including financing probability, quantile ranking, important feature statistics, etc. In addition, the user can also see a visualized line chart of the financing probability within the selected time range, and intuitively obtain the historical changes in the company's financing ability. The display effect of the product end is as follows Figure 4a and Figure 4b shown.

[0069] Combine the following Figure 5 The training process of the financing probability prediction model (hereinafter referred to as the "model") is described in detail. The training of this model follows the complete machine learning model training process, and has been strictly processed in terms of labeling and sampling, and combines the unique characteristics of business needs in terms of feature engineering and model training.

[0070] Step 1: Task processing

[0071] The goal is to predict the probability of a private equity company entering the next round of financing. For each time point (month) in the following two years, the companies with real financing records are marked as "1" and the companies that failed to obtain the next round of financing are marked as "0" for binary classification. The probability of a sample being defined as a "1" label is the probability of entering the next round of financing.

[0072] Step 2: Data Preprocessing

[0073] The data used in this model comes from the applicant's exclusive private equity database, and the feature values ​​(input data, hereinafter referred to as "features") are the applicant's self-developed data indicators. Preprocessing mainly includes processing missing values, outliers, duplicate values ​​and text data.

[0074] Step 3: Exploratory Data Analysis

[0075] Use statistical tools to understand the distribution of labels, the distribution of features, and the quantitative relationship between features. Exploratory analysis helps to discover potential data patterns and select appropriate feature engineering methods and machine learning algorithms. The distribution of most original features is skewed, island-shaped, or sawtooth-shaped, with a high degree of discreteness.

[0076] Step 4: Feature Engineering

[0077] Including feature derivation, feature screening or dimensionality reduction, data conversion, etc. Feature engineering is a mixture of professional domain knowledge, intuition and basic mathematical ability. After preprocessing the data through feature engineering, the algorithm model can reduce the interference of noise and better find the data pattern. This model cross-derives the 20 original features of the eigenvalues ​​based on business logic and mathematical functions to improve the information contained in the eigenvalues. The number of features increases to about 90, and the correlation is significantly improved; then based on indicators such as correlation, information ratio, and multicollinearity, feature screening is performed to reduce the time and resource usage of model operation, and finally 32 features are retained for modeling.

[0078] Step 5: Dataset segmentation

[0079] The main purpose of data set segmentation is to avoid overfitting and underfitting, stabilize model performance, and improve model generalization ability. The data set is divided into training set, validation set, and test set. The training set is used for model training, the validation set is used for model parameter tuning and performance evaluation, and the test set is used to evaluate the final performance and performance of the model. This model selects data before the end of 2021 as the training set for model training, uses K-fold cross-validation to select the validation set for parameter tuning, and selects data before the end of 2021 as the test set to evaluate the final performance of the model.

[0080] Step 6: Model training

[0081] Select appropriate machine learning algorithms for training. For binary classification problems, algorithms such as logistic regression, multi-layer perceptron, support vector machine, random forest, gradient boosted decision tree (Boost) often have good performance. This model trains the above machine learning algorithms and selects or integrates algorithms based on the performance of each algorithm.

[0082] Logistic regression: It is a statistical learning method for solving binary classification problems. It is based on a linear model and uses a logistic function (Sigmoid function) to map the linear output to a probability value between 0 and 1.

[0083] Multilayer Perceptron (MLP): is a basic feed-forward artificial neural network used to solve classification and regression problems. MLP consists of an input layer, one or more hidden layers, and an output layer, each containing multiple neurons.

[0084] Support vector machine: It is a binary classification model. Its basic model is a linear classifier with the largest interval defined in the feature space. It maximizes the boundary (i.e., "interval" or "margin") between two categories by finding an optimal hyperplane. This optimal hyperplane can separate samples of different categories as much as possible, and the sample point closest to the hyperplane is called the support vector.

[0085] Random Forest: is an ensemble learning method that can be used for classification and regression problems. It improves the accuracy and robustness of predictions by combining multiple decision tree models. A decision tree is a tree-like structure used to classify or regress input data, with a series of split nodes and leaf nodes representing different branches and decision paths of the data. Random Forest performs particularly well in processing high-dimensional data and large-scale data sets.

[0086] AdaBoost: It is also an ensemble learning method that creates a strong classifier by combining multiple weak classifiers (usually very simple classifiers). The core idea of ​​AdaBoost is to assign a weight to each training sample, which is updated in each iteration so that the algorithm can focus on samples that were misclassified in the previous iteration.

[0087] XGBoost: is an efficient and widely used ensemble learning algorithm, which performs well in various data science competitions and practical applications. XGBoost is an implementation of the gradient boosting algorithm, which optimizes the traditional gradient boosting method and provides faster performance and better scalability.

[0088] Step 7: Model evaluation and tuning

[0089] Use the validation set to evaluate the model and tune it based on performance indicators (such as accuracy, recall, etc.). This mainly includes hyperparameter adjustment, adding features, custom objective functions, algorithm fusion, etc. The main evaluation indicators for classification tasks include:

[0090] 1) Recall: The number of samples that the model predicts will receive financing and actually receive financing / the total number of samples that actually receive financing;

[0091] 2) Precision: The number of samples predicted by the model to have received financing and actually received financing / the total number of samples predicted by the model to have received financing;

[0092] 3) F1: F1=2×Recall×Precision / (Recall+Precision);

[0093] 4) AUC: The area under the ROC (Receiver Operating Characteristic Curve) curve is an important indicator for evaluating the performance of classification models in machine learning. It measures the accuracy of the model's classification results and reflects the model's ability to distinguish different categories.

[0094] From the final performance of each algorithm, the multilayer perceptron model performs the worst in the training set and the test set, and the original data characteristics are difficult to meet the requirements of the neural network in the continuous space derivative; the logistic regression model focuses more on the precision and recall of the "1" class label, while AdaBoost and XGBoost focus more on the precision and recall of the "0" class label; the random forest model performs more balanced. Therefore, the present invention adopts an integrated learning strategy to integrate the output results of various algorithms as the optimal model.

[0095] In actual business, users pay more attention to the coverage of enterprises that have actually obtained financing, that is, the recall rate, so as to ensure that high-quality enterprises are not missed as much as possible. After repeated testing and optimization, the recall rate of this model for enterprises that have actually obtained financing has reached more than 80%, and the recall rate for enterprises that have not actually obtained financing has reached more than 70%, maintaining a high recognition accuracy for both types of labels.

[0096] On the other hand, the stability of the model is particularly important. One of the goals of this model is dynamic evaluation. A stable model can better guarantee the evaluation effect in dynamics and provide users with more reliable data information. During the test, the AUC value can reach more than 0.85, and the model reproducibility (or stability) can reach more than 90%.

[0097] According to another aspect of the present invention, a non-transitory computer-readable storage medium storing a computer program is provided, wherein the computer program implements the method according to the present invention when executed by a processor.

[0098] According to another aspect of the present invention, a computer program product is provided, comprising a computer program, wherein the computer program implements the method according to the present invention when executed by a processor.

[0099] Although the content of the present invention has been described in detail through the above preferred embodiments, it should be appreciated that the above description should not be considered as a limitation of the present invention. After reading the above content, it will be apparent to those skilled in the art that various modifications and substitutions of the present invention will occur. Therefore, the protection scope of the present invention should be limited by the appended claims.

Claims

1. A method for predicting the financing probability of a private equity enterprise, characterized in that: include: Preprocess the underlying data source of the enterprise in the MySQL database; Synchronize the underlying data source to the Hadoop cluster, perform secondary association processing, and obtain a large wide table of advanced project screening and a large wide table of project investment and financing events; Filter the data within the current preset time period from the project advanced screening large wide table and the project investment and financing event large wide table, perform in-depth processing on the data in the Hive data warehouse, and generate a Hive indicator table that can be used by the model; Establish a model indicator table in the MySQL database by year, and synchronize the Hive indicator table to the MySQL database; Receive prediction instructions for one or more target enterprises, obtain model input indicator data from the model indicator table, input the data into a pre-built financing probability prediction model, and calculate the financing probability of the one or more target enterprises.

2. The method for predicting the financing probability of a private equity enterprise according to claim 1, characterized in that: The underlying data sources include: a project basic information table, a project investment and financing data table, and a fund investment project table.

3. The method for predicting the financing probability of a private equity enterprise according to claim 1, characterized in that: The in-depth processing of data in the Hive data warehouse includes: The values ​​of several original features are determined from the screened data, and the values ​​of several derived features are determined by data derivation according to the values ​​of the several original features.

4. The method for predicting the financing probability of a private equity enterprise as claimed in claim 3, characterized in that: The derived features are obtained by cross-deriving the original features based on business logic and mathematical functions in advance.

5. The method for predicting the financing probability of a private equity enterprise according to claim 1, characterized in that: The model entry indicators involve the following three dimensions: the abundance of upstream capital in the track, the heat of track capital, and the company's own financing capabilities.

6. The method for predicting the financing probability of a private equity enterprise according to claim 1, characterized in that: The financing probability prediction model is based on an integrated learning strategy and is an optimal model obtained by fusing the output results of multiple machine learning algorithms.

7. The method for predicting the financing probability of a private equity enterprise according to claim 6, characterized in that: The multiple machine learning algorithms include logistic regression model, random forest model and gradient boosting decision tree.

8. The method for predicting the financing probability of a private equity enterprise according to claim 1, characterized in that: Also includes: Regularly forecast the financing probability of the one or more target enterprises to achieve dynamic forecasting of the financing capabilities of the enterprises.

9. A non-transitory computer-readable storage medium storing a computer program, wherein: When the computer program is executed by a processor, the computer program implements the method according to any one of claims 1 to 8.

10. A computer program product comprising a computer program, wherein: When the computer program is executed by a processor, the computer program implements the method according to any one of claims 1 to 8.

Images