Fund risk prediction method and device

Abstract

The invention discloses a fund risk prediction method and device, relates to the field of artificial intelligence and can also be used in the financial field, and the method comprises the steps: obtaining overseas study basic information filled in an overseas study fund risk assessment page by a customer; obtaining customer financial data of the customer from a bank system; generating overseas study feature data according to the overseas study basic information, generating customer financial feature data according to the customer financial data, and generating a multi-dimensional feature vector according to the overseas study feature data and the customer financial feature data; and inputting the multi-dimensional feature vector into a preset overseas study fund risk prediction model to obtain a risk prediction result output by the overseas study fund risk prediction model. According to the invention, the beneficial effect of accurately predicting the overseas study fund risk is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a fund risk prediction method and device. BACKGROUND

[0002] With the acceleration of globalization, studying abroad has become the choice of more and more families. However, the high cost of studying abroad and the complex international financial environment make the preparation of studying abroad funds face multiple risks. If the fund risk is not discovered in advance, the student's study and life will be affected during the period of studying abroad. In order to help families cope with these potential fund risks, banks have launched a studying abroad fund planning platform, on which the fund risk of a customer can be predicted when the customer consults studying abroad. However, the prediction of the current studying abroad fund planning platform is often based on expert experience to estimate a total expected cost, and then compare it with the customer's reserved fund for studying abroad to obtain the prediction result. The current risk prediction result is very inaccurate and too dependent on expert experience. The prior art lacks a more accurate studying abroad fund risk prediction scheme. SUMMARY

[0003] The present application proposes a fund risk prediction method and device to solve at least one of the technical problems in the background art.

[0004] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a fund risk prediction method is provided, which comprises:

[0005] obtaining the basic information of studying abroad filled by the customer in the studying abroad fund risk assessment page;

[0006] obtaining the customer financial data of the customer from the bank system;

[0007] generating studying abroad feature data according to the basic information of studying abroad, generating customer financial feature data according to the customer financial data, and generating a multi-dimensional feature vector according to the studying abroad feature data and the customer financial feature data;

[0008] inputting the multi-dimensional feature vector into a preset studying abroad fund risk prediction model to obtain a risk prediction result output by the studying abroad fund risk prediction model, wherein the studying abroad fund risk prediction model is obtained by training a preset classification model using training samples, the training samples are multi-dimensional feature vectors for model training generated based on historical studying abroad fund risk assessment data, and the risk prediction result labels are obtained on the multi-dimensional feature vectors for model training based on expert experience.

[0009] In some embodiments, the risk prediction result includes a first prediction result for indicating that there is a study abroad fund risk and a second prediction result for indicating that there is no study abroad fund risk.

[0010] The fund risk prediction method further includes:

[0011] If the risk prediction result is the first prediction result, a total study abroad cost is predicted according to study abroad target country information, study abroad target school information, study abroad start and end time information, study abroad education stage information, student personal consumption level information, and a preset study abroad cost prediction model.

[0012] A study abroad fund gap is determined according to the total study abroad cost and study abroad reserve fund information.

[0013] A study abroad financial product recommendation scheme is determined according to the study abroad fund gap and maximum historical drawdown data accepted by the customer, and the study abroad financial product recommendation scheme is sent to the customer.

[0014] In some embodiments, the total study abroad cost is predicted according to the study abroad target country information, the study abroad target school information, the study abroad start and end time information, the study abroad education stage information, the student personal consumption level information, and the preset study abroad cost prediction model, including:

[0015] A cost feature vector is generated according to the study abroad target country information, the study abroad target school information, the study abroad start and end time information, the study abroad education stage information, and the student personal consumption level information.

[0016] The cost feature vector is input into the study abroad cost prediction model to obtain a total study abroad cost prediction value output by the study abroad cost prediction model, wherein the study abroad cost prediction model is obtained by training a preset classification model using training samples, the training samples are cost feature vectors generated based on historical data for model training, and the total study abroad cost labels are marked on the cost feature vectors based on real study abroad costs.

[0017] In some embodiments, the study abroad financial product recommendation scheme is determined according to the study abroad fund gap and the maximum historical drawdown data accepted by the customer, including:

[0018] A study abroad financial product combination capable of meeting the study abroad fund gap is determined according to the study abroad reserve fund information and the study abroad start and end time information.

[0019] A historical average annual yield is determined according to the maximum historical drawdown data accepted by the customer.

[0020] According to the historical average annual yield and the average annual yield of each study abroad financial product in the study abroad financial product portfolio, each study abroad financial product in the study abroad financial product portfolio is screened to obtain the study abroad financial product recommendation scheme.

[0021] In some embodiments, the study abroad financial product portfolio includes fixed-income financial products and equity financial products, and the ratio between the fixed-income financial products and the equity financial products is determined according to historical customer risk assessment records.

[0022] In some embodiments, before obtaining the study abroad basic information filled in by the customer in the study abroad fund risk assessment page, the fund risk prediction method further comprises:

[0023] Obtaining popular study abroad country data and popular study abroad school data predicted by a preset popular study abroad destination prediction model;

[0024] Embedding the popular study abroad country data and the popular study abroad school data into the drop-down options of the input box of the study abroad fund risk assessment page for the customer to refer to and complete the input of the target country and the target school in the input box.

[0025] In some embodiments, the fund risk prediction method further comprises:

[0026] Obtaining historical study abroad fund risk assessment data and parsing historical data of each study abroad destination from the historical data, the historical data including click counts, share counts, save counts and timestamps, and the study abroad destinations including study abroad countries and study abroad schools;

[0027] Establishing a training sample set according to the historical data;

[0028] According to the training sample set and a preset popularity calculation formula, training on the basis of a preset machine learning model to obtain the popular study abroad destination prediction model, the popularity calculation formula containing a time decay factor and weight values of the click counts, the share counts and the save counts; during model training, making the popularity value calculated by the popularity calculation formula as a target variable of the model, making the prediction result of the model close to the target variable, and continuously optimizing the time decay factor and the weight values according to the feedback of the loss function until the prediction accuracy of the trained model meets the requirements.

[0029] In some embodiments, the study abroad basic information includes one or more of study abroad target country information, study abroad target school information, study abroad start and end time information, study abroad education stage information, study abroad reserve fund information and student personal consumption level information.

[0030] In some embodiments, the customer financial data includes one or more of maximum historical drawdown data accepted by the customer, customer household annual income data, loan information, historical credit record, historical customer risk assessment record, and historical in-line wealth management product purchase record.

[0031] To achieve the above object, according to another aspect of the present application, there is provided a fund risk prediction device, comprising:

[0032] Overseas study basic information acquisition unit, configured to acquire overseas study basic information filled in by a customer in an overseas study fund risk assessment page;

[0033] Customer financial data acquisition unit, configured to acquire customer financial data of the customer from a bank system;

[0034] Multi-dimensional feature vector generation unit, configured to generate overseas study feature data according to the overseas study basic information, generate customer financial feature data according to the customer financial data, and generate a multi-dimensional feature vector according to the overseas study feature data and the customer financial feature data;

[0035] Risk prediction unit, configured to input the multi-dimensional feature vector into a preset overseas study fund risk prediction model to obtain a risk prediction result output by the overseas study fund risk prediction model, wherein the overseas study fund risk prediction model is derived by training a preset classification model using training samples, the training samples are multi-dimensional feature vectors for model training generated based on historical overseas study fund risk assessment data, and the multi-dimensional feature vectors for model training are labeled with risk prediction results based on expert experience to obtain.

[0036] In some embodiments, the risk prediction result includes a first prediction result for indicating that there is an overseas study fund risk and a second prediction result for indicating that there is no overseas study fund risk;

[0037] The fund risk prediction device further comprises:

[0038] Overseas study total cost prediction unit, configured to, if the risk prediction result is the first prediction result, predict an overseas study total cost according to overseas study target country information, overseas study target school information, overseas study start and end time information, overseas study education stage information, student personal consumption level information, and a preset overseas study cost prediction model;

[0039] Overseas study fund gap determination unit, configured to determine an overseas study fund gap according to the overseas study total cost and overseas study reserve fund information;

[0040] The overseas study financial product recommendation unit is configured to determine an overseas study financial product recommendation scheme based on the overseas study fund gap and the maximum historical drawdown data accepted by the customer, and send the overseas study financial product recommendation scheme to the customer.

[0041] To achieve the above object, according to another aspect of the present application, there is further provided a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above fund risk prediction method when executing the computer program.

[0042] To achieve the above object, according to another aspect of the present application, there is further provided a computer readable storage medium having a computer program / instruction stored thereon, wherein the computer program / instruction implements the steps of the above fund risk prediction method when executed by a processor.

[0043] To achieve the above object, according to another aspect of the present application, there is further provided a computer program product comprising a computer program / instruction, wherein the computer program / instruction implements the steps of the above fund risk prediction method when executed by a processor.

[0044] The present application has the following beneficial effects:

[0045] The present application generates overseas study feature data according to the basic information of the customer, generates customer financial feature data according to the customer financial data of the customer, and generates a multi-dimensional feature vector according to the overseas study feature data and the customer financial feature data, and then predicts the overseas study risk through the trained overseas study fund risk prediction model, which improves the accuracy of the overseas study risk prediction by combining the multi-dimensional features for the overseas study risk prediction. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor. In the drawings:

[0047] Figure 1 is a flowchart of the fund risk prediction method of the embodiment of the present application;

[0048] Figure 2 is a flowchart of the financial product recommendation of the embodiment of the present application;

[0049] Figure 3 is a flowchart of the prediction of the total cost of overseas study of the embodiment of the present application;

[0050] Figure 4is a flowchart of determining a financial product recommendation scheme by an embodiment of the present application;

[0051] Figure 5 is a flowchart of processing a study fund risk assessment page by an embodiment of the present application;

[0052] Figure 6 is a flowchart of training a popular study destination prediction model by an embodiment of the present application;

[0053] Figure 7 is a first structural block diagram of a fund risk prediction device by an embodiment of the present application;

[0054] Figure 8 is a second structural block diagram of a fund risk prediction device by an embodiment of the present application;

[0055] Figure 9 is a schematic diagram of a computer device by an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to make the personnel in the art better understand the present application scheme, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by the person skilled in the art without creative labor should belong to the scope of protection of the present application.

[0057] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0058] It should be noted that the terms "include" and "have" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device containing a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0059] The information collected in the technical solutions in the application is information and data authorized by the user or authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data comply with relevant laws, regulations and standards of countries and regions, necessary security measures are taken, do not violate public order and good customs, and provide corresponding operation portal for user to choose authorization or refusal.

[0060] The acquisition, transmission, storage, use and processing of data in the technical solutions in the application comply with relevant provisions of national laws and regulations.

[0061] It should be noted that in the embodiments of the application, some existing industry solutions such as software, components, models, etc. may be mentioned, which should be considered as exemplary, and the purpose is only to illustrate the feasibility of the implementation of the technical solutions of the application, but it does not mean that the applicant has or will necessarily use the solution.

[0062] It should be noted that the fund risk prediction method and device can be used in the financial field, and can also be used in any field other than the financial field, and the application field of the fund risk prediction method and device is not limited.

[0063] Figure 1 The flowchart of the fund risk prediction method of the embodiment of the application is shown in FIG. Figure 1 In one embodiment of the application, the fund risk prediction method of the application comprises steps S101 to S104.

[0064] Step S101, obtaining the basic information of studying abroad filled in by the customer in the studying abroad fund risk assessment page.

[0065] The basic information of studying abroad includes one or more of studying abroad target country information, studying abroad target school information, studying abroad start and end time information, studying abroad education stage information, studying abroad reserve fund information and student personal consumption level information.

[0066] In one embodiment of the application, the basic information of studying abroad can include studying abroad target country information. In another embodiment of the application, the basic information of studying abroad can include studying abroad target country information and studying abroad target school information. In another embodiment of the application, the basic information of studying abroad includes studying abroad target country information, studying abroad target school information, studying abroad start and end time information, studying abroad education stage information, studying abroad reserve fund information and student personal consumption level information.

[0067] In an embodiment of the present application, a customer can use a bank-provided overseas study fund planning platform (hereinafter referred to as the platform) to make overseas study fund planning and overseas study fund risk prediction. When making overseas study fund risk prediction, the customer first fills in information in an overseas study fund risk assessment page provided by the overseas study fund planning platform, and then the platform makes multi-dimensional overseas study fund risk prediction in combination with the information filled in by the customer.

[0068] In step S101, the platform acquires overseas study related basic information filled in by the customer in the overseas study fund risk assessment page, which directly reflects the overseas study plan and fund demand of the customer. In some embodiments, the overseas study related basic information mainly includes overseas study target country information, overseas study target school information, overseas study start and end time information, overseas study education stage information, overseas study reserve fund information, and student personal consumption level information.

[0069] In some embodiments, the overseas study target country information can include any one or more of living costs, tuition fees, and economic environments in different countries. The living costs, tuition fees, and economic environments in different countries are different and will directly affect the overseas study fund risk.

[0070] In some embodiments, the overseas study target school information can include a specific school that the customer plans to study in. The tuition fees, scholarships, etc. of different schools can be different and will directly affect the overseas study fund demand.

[0071] In some embodiments, the overseas study start and end time information can include the length of time that the customer plans to study abroad. The length of time that the customer plans to study abroad will directly affect the overseas study fund demand.

[0072] In some embodiments, the overseas study education stage information can include the education stage (undergraduate, master, doctoral, etc.) that the customer is in. The fee structure, scholarship opportunities, etc. of different education stages are different and will directly affect the overseas study fund demand.

[0073] In some embodiments, the overseas study reserve fund information can include the amount of fund reserve that the customer currently has to support overseas study. The amount of fund reserve that the customer currently has to support overseas study directly determines the financial flexibility of the customer during overseas study.

[0074] In some embodiments, the student personal consumption level information can include the expected personal consumption level of the customer. The personal consumption of the customer includes living expenses, accommodation expenses, transportation expenses, etc., which is an important factor for predicting the fund risk of the customer.

[0075] These overseas study related basic information constitutes the overseas study characteristic data and is one of the important inputs of the entire risk assessment model.

[0076] Step S102, obtaining the customer financial data of the customer from the banking system.

[0077] In step S102, the financial data of the customer is obtained from the banking system, which reflects the financial status and historical behavior of the customer. The customer financial data includes one or more of the following: the maximum historical drawdown data accepted by the customer, the customer's annual household income data, loan information, historical credit records, historical customer risk assessment records, and historical in-house financial product purchase records.

[0078] In one embodiment of the present application, the customer financial data can include the maximum historical drawdown data accepted by the customer. In another embodiment of the present application, the customer financial data can include the maximum historical drawdown data accepted by the customer and the customer's annual household income data. In another embodiment of the present application, the customer financial data includes the maximum historical drawdown data accepted by the customer, the customer's annual household income data, loan information, historical credit records, historical customer risk assessment records, and historical in-house financial product purchase records.

[0079] In some embodiments, the above-mentioned maximum historical drawdown data accepted by the customer can include the maximum drawdown amplitude of the customer's past investment or account. The maximum drawdown amplitude of the customer's past investment or account represents the customer's ability to withstand market fluctuations.

[0080] In some embodiments, the above-mentioned annual household income data can include the annual income of the customer and his or her family. The annual income of the customer and his or her family reflects the basic source of support for the study abroad funds.

[0081] In some embodiments, the above-mentioned loan information can include any one or more of the following: whether the customer has a loan currently or historically, the loan amount, and the loan term. The specific loan situation of the customer has a direct impact on the customer's cash flow and debt situation, which will directly affect the demand for study abroad funds.

[0082] In some embodiments, the above-mentioned historical credit records can include any one or more of the following: the customer's credit score and default records. The customer's credit score and default records can measure the customer's creditworthiness and future ability to bear risks.

[0083] In some embodiments, the above-mentioned historical customer risk assessment records can include the customer's risk rating records in banks or other financial institutions. The customer's risk rating records can reflect the customer's overall risk tolerance.

[0084] In some embodiments, the above-mentioned historical financial product purchase records can include information about the financial products purchased by the customer. The information about the financial products purchased by the customer can reflect his or her investment style, risk preference, and capital allocation.

[0085] The customer financial data constitutes customer financial feature data, and the customer financial feature data and the study feature data constitute a multi-dimensional feature vector.

[0086] In step S103, the study feature data and the customer financial feature data are integrated to generate a multi-dimensional feature vector.

[0087] In step S103, the platform integrates the study feature data and the customer financial feature data to generate a multi-dimensional feature vector. The multi-dimensional feature vector integrates the study plan, financial status and consumption capacity of the customer, which helps to improve the prediction accuracy.

[0088] In an embodiment of the present application, the study basic information is standardized in step S103 to generate the study feature data, and the customer financial data is standardized to generate the customer financial feature data.

[0089] In an embodiment of the present application, the study feature data and the customer financial feature data can be combined to obtain a multi-dimensional feature vector. In the combination, the study feature data and the customer financial feature data can be converted into data of the same dimension, and then weighted calculation is performed based on the weight set by the expert to obtain the multi-dimensional feature vector.

[0090] In step S104, the multi-dimensional feature vector is input into a preset study fund risk prediction model to obtain a risk prediction result output by the study fund risk prediction model.

[0091] The study fund risk prediction model is obtained by training a preset classification model using training samples. The training samples are multi-dimensional feature vectors generated based on historical study fund risk assessment data for model training, and are labeled with risk prediction results based on expert experience.

[0092] In an embodiment of the present application, the present application can generate historical multi-dimensional feature vectors after preprocessing historical study fund risk assessment data, customer basic information and financial data collected by the platform. These historical data samples are labeled, and the present application can label the corresponding risk based on the actual study fund risk situation (such as historical records of running out of money during the study) or based on expert experience to generate training samples.

[0093] In an embodiment of the present application, the study abroad fund risk prediction model can be trained using Logistic Regression, Random Forest, Support Vector Machine (SVM), or Gradient Boosting Tree (such as XGBoost, LightGBM).

[0094] As can be seen, the fund risk prediction method of the present application improves the accuracy of study abroad fund risk prediction by integrating customer's study abroad information and financial data, using a multi-dimensional feature vector to generate and predicting risk through a classification model. The present application improves the accuracy of risk prediction through machine learning, and combines expert experience to achieve the technical effect of accurately and efficiently predicting study abroad fund risk.

[0095] In an embodiment of the present application, the risk prediction result includes a first prediction result indicating the existence of study abroad fund risk and a second prediction result indicating the non-existence of study abroad fund risk.

[0096] In the present application, the first prediction result indicates that the customer has study abroad fund risk, meaning that the customer may face a shortage of funds or other financial pressure during study abroad, and the second prediction result indicates that the customer has no study abroad fund risk, meaning that the customer's financial situation is good enough to cope with expenses during study abroad.

[0097] As shown in Figure 2 In an embodiment of the present application, the fund risk prediction method of the present application further includes steps S201 to S203.

[0098] Step S201, if the risk prediction result is the first prediction result, the total cost of study abroad is predicted according to the study abroad target country information, the study abroad target school information, the study abroad start and end time information, the study abroad education stage information, the student personal consumption level information, and the preset study abroad cost prediction model.

[0099] In an embodiment of the present application, the study abroad cost prediction model is pre-trained for the estimation of the total cost of study abroad, and the trained study abroad cost prediction model is set in the platform.

[0100] Step S202, determine the study abroad fund gap according to the total cost of study abroad and the study abroad reserve fund information.

[0101] Step S203, determine the study abroad financial product recommendation scheme according to the study abroad fund gap and the maximum historical drawdown data accepted by the customer, and send the study abroad financial product recommendation scheme to the customer.

[0102] In an optional embodiment of the present application, the present application can adopt a machine learning method to train a study abroad financial product recommendation model, and then input the study abroad funding gap and the maximum historical drawdown data accepted by the customer into the model as inputs, and output a recommended study abroad financial product recommendation scheme from the study abroad financial product recommendation model.

[0103] As can be seen, the present application first judges whether the customer has a funding risk through the risk prediction model, and then predicts the total cost of study abroad, calculates the funding gap according to the basic information and financial data of the customer, and recommends appropriate financial products according to the customer's funding risk tolerance. The present application not only helps the customer to identify potential funding risks, but also provides an effective solution to help the customer to prepare for the funding needs during the period of study abroad, and improves the customer's use experience.

[0104] As shown in Figure 3 In an embodiment of the present application, the step S201 of predicting the total cost of study abroad according to the information of the target country of study, the information of the target school of study, the information of the start and end time of study, the information of the education stage of study, the information of the personal consumption level of the student and the preset study abroad cost prediction model specifically comprises steps S301 and S302.

[0105] Step S301, generating a cost feature vector according to the information of the target country of study, the information of the target school of study, the information of the start and end time of study, the information of the education stage of study and the information of the personal consumption level of the student.

[0106] In an embodiment of the present application, the present application standardizes the information of the target country of study, the information of the target school of study, the information of the start and end time of study, the information of the education stage of study and the information of the personal consumption level of the student to generate a multi-dimensional feature vector, i.e. a cost feature vector, representing the study cost related features of the customer.

[0107] Step S302, inputting the cost feature vector into the study cost prediction model to obtain the total cost of study predicted by the study cost prediction model.

[0108] In an embodiment of the present application, the study cost prediction model is obtained by training a preset classification model using training samples, the training samples are cost feature vectors generated based on historical data for model training, and the total cost of study is labeled on the cost feature vectors for model training based on the real study cost.

[0109] In an embodiment of the present application, the study cost prediction model is trained using historical study cost related data. Each training sample includes a set of known cost feature vectors, and the corresponding real study total cost label. In this way, the model can capture the impact of different features on the study cost. The real study cost label in the training data is obtained by labeling the actual cost in a large number of historical study cases. These real cost labels help the model gradually adjust the parameters during the training process to achieve more accurate prediction.

[0110] In an embodiment of the present application, the study cost prediction model can be trained using decision tree model, random forest, neural network model, gradient boosting tree (GBDT).

[0111] In an embodiment of the present application, the training process of the study cost prediction model generally includes:

[0112] Process 1, the platform trains the model using a large amount of historical data, and each training sample includes a cost feature vector and a real study total cost;

[0113] Process 2, through the training process of the model, the system adjusts the model parameters to minimize the prediction error, for example, using mean square error (MSE) or root mean square error (RMSE) as the loss function to measure the deviation between the predicted value and the real cost;

[0114] Process 3, as the number of iterations of the model training increases, the model can gradually learn the impact weight of different features on the study cost, and achieve more accurate prediction.

[0115] As shown in Figure 4 In an embodiment of the present application, the step S203 of determining the study financial product recommendation scheme according to the study fund gap and the maximum historical drawdown accepted by the customer includes steps S401 to S403.

[0116] Step S401, according to the study reserve fund information and the study start and end time information, determine the study financial product combination that can meet the study fund gap.

[0117] The present application finds a financial product combination that can make up the fund gap within the customer specified study start and end time by screening the products in the financial product library. The present application can use optimization algorithms (such as linear programming, goal programming, etc.) to generate a financial product combination that can maximize the yield and make up the fund gap under the premise of meeting the time limit and risk requirements.

[0118] In the present application, the following conditions are met: 1, the term matching, the expiration date of the financial product should match the start and end time of studying abroad, to ensure that the customer can obtain sufficient funds during the study period; 2, risk matching, according to the risk preference of the customer (which can be inferred from historical drawdown data and other information), select the financial product suitable for its risk tolerance; 3, the gap of the funds is filled, select a group of financial products, the expected yield can fill the gap of the funds for studying abroad.

[0119] Step S402, according to the maximum historical drawdown data accepted by the customer to determine the historical average annual yield.

[0120] In the present application, the maximum historical drawdown refers to the maximum loss of the customer in the investment history, which reflects the risk tolerance of the customer. For example, if the maximum historical drawdown of the customer is 10%, it means that the risk that the customer can bear is relatively high. If the customer's maximum drawdown is high, it means that the customer may accept higher risk to pursue higher yield. Correspondingly, the historical average annual yield will be higher; on the contrary, the customer with lower drawdown tends to be conservative investment, and the annual yield will be lower. The present application can infer the average annual yield of the customer's historical investment and financing based on the maximum historical drawdown data accepted by the customer.

[0121] Step S403, according to the historical average annual yield and the average annual yield of each financial product in the combination of the financial products for studying abroad, each financial product in the combination of the financial products for studying abroad is screened, and the recommended scheme of the financial products for studying abroad is obtained.

[0122] In the present application, the financial product with risk matching can be screened according to the maximum historical drawdown that the customer can bear. The product with too high risk or too low yield will be excluded. Specifically, the annual yield of each product in the combination of the financial products should be close to or higher than the historical average annual yield of the customer, to ensure that the yield level can meet the customer's expectation.

[0123] In an optional embodiment of the present application, the present application can set a yield threshold according to the historical average annual yield of the customer. Only those financial products with annual yield higher than the threshold and consistent with the risk tolerance of the customer can enter the recommended list.

[0124] The present application will output the final recommended scheme of the financial products for studying abroad after screening and combination optimization, including the detailed information of each financial product (such as name, annual yield, term, investment amount, etc.). The recommended scheme of the financial products for studying abroad will show how to make up the gap of the funds for studying abroad through the yield of the combination, and ensure that the customer can obtain the required funds on time during the study period.

[0125] Therefore, the application can screen a suitable financial product portfolio based on the customer's fund demand, investment risk preference and income expectation, so as to meet the customer's demand for studying abroad funds and provide a financial product recommendation scheme.

[0126] In an embodiment of the application, the studying abroad financial product portfolio comprises fixed-income financial products and equity financial products, and the ratio between the fixed-income financial products and the equity financial products is determined according to the historical customer risk assessment records.

[0127] The fixed-income financial products generally include bonds, fixed deposits and other low-risk and stable-yield products. Such products have small yield fluctuations and provide stable cash flow, but have relatively low return rates. The equity financial products generally include stocks, funds, hybrid financial products and other high-risk and high-yield products. Such products have large yield fluctuations and may bring higher returns, but have greater risks.

[0128] In an embodiment of the application, the application can determine the risk bearing capacity of the customer according to the historical customer risk assessment records, divide the customer into low-risk, medium-risk and high-risk types, and then set the ratio between the fixed-income financial products and the equity financial products corresponding to the three different types. For example, for a low-risk customer, the ratio of the fixed-income financial products is 80%-90%, and the ratio of the equity financial products is 10%-20%; for a medium-risk customer, the ratio of the fixed-income financial products is 50%-70%, and the ratio of the equity financial products is 30%-50%; and for a high-risk customer, the ratio of the fixed-income financial products is 30%-40%, and the ratio of the equity financial products is 60%-70%.

[0129] As shown in FIG. Figure 5 In an embodiment of the application, before the step S101 of obtaining the studying abroad basic information filled in the studying abroad fund risk assessment page by the customer, the fund risk prediction method further comprises steps S501 and S502.

[0130] In step S501, hot studying abroad country data and hot studying abroad school data predicted by a preset hot studying abroad destination prediction model are obtained.

[0131] In step S502, the hot studying abroad country data and the hot studying abroad school data are embedded in the drop-down options of the input box of the studying abroad fund risk assessment page for the customer to refer to and complete the input of the target country and the target school in the input box.

[0132] In an embodiment of the present application, when a customer enters the study abroad fund risk assessment page, the system provides recommendations of popular study abroad countries and schools in the drop-down options of the input box. The recommended list in the drop-down box is sorted according to the degree of popularity, and the user can see the countries and schools that the prediction model considers most likely to be the destination of study abroad first. By embedding the data of popular study abroad countries and schools, the customer does not need to manually input the country and school name, greatly simplifying the data filling work in the evaluation process. The customer can understand which popular study abroad destinations and schools are currently available by viewing the recommended items in the drop-down options, to help them make better choices.

[0133] As shown in Figure 6 In an embodiment of the present application, the fund risk prediction method of the present application further includes steps S601 to S603.

[0134] Step S601, historical study abroad fund risk assessment data is obtained, and historical data of each study abroad destination is parsed therefrom, the historical data including the number of clicks, the number of shares, the number of saves, and the timestamp, and the study abroad destination including a study abroad country and a study abroad school.

[0135] In an embodiment of the present application, the number of clicks refers to the number of times a customer selects or inputs a study abroad destination (a study abroad country or a study abroad school) in the study abroad fund risk assessment page within a certain period of time. Sharing refers to a customer sharing the filled study abroad fund risk assessment page information to others after completing the filling of the study abroad fund risk assessment page, and the number of shares refers to the number of times a study abroad destination (a study abroad country or a study abroad school) is shared. Saving refers to a customer saving the filled study abroad fund risk assessment page information on the platform or locally after completing the filling of the study abroad fund risk assessment page, and the number of saves refers to the number of times a study abroad destination (a study abroad country or a study abroad school) is saved. The timestamp is the specific time when the user performs the click, share, and save behaviors, which is used for subsequent time decay calculation.

[0136] Step S602, a training sample set is established according to the historical data.

[0137] In an embodiment of the present application, the present application extracts features from the historical data of each study abroad destination, generates a feature vector, and then manually or automatically labels the feature vector according to the actual popularity in the historical data to generate a training sample. The popularity label can be set according to the historical ranking data or the heat index. Assuming that the historical popularity of a target school is high, it can be marked as “popular”, and a cold school is marked as “cold”.

[0138] Step S603, the training sample set and a preset heat calculation formula are used to train a preset machine learning model to obtain the popular study abroad destination prediction model.

[0139] In an embodiment of the present application, the hotness calculation formula includes a time decay factor and respective weight values of the number of clicks, the number of shares, and the number of saves. When training the popular overseas study destination prediction model, the hotness value calculated by the hotness calculation formula is used as the target variable of the model, the prediction result of the model approaches the target variable, and the time decay factor and the weight values are continuously optimized according to the feedback of the loss function until the prediction accuracy of the trained model meets the requirements.

[0140] In an embodiment of the present application, the popular overseas study destination prediction model can be trained using a decision tree model (such as random forest), a time series model (such as LSTM), or a gradient boosting tree (such as XGBoost or LightGBM).

[0141] In an embodiment of the present application, the hotness calculation formula is specifically:

[0142] Hotness = (w1 x number of clicks + w2 x number of shares + w3 x number of saves) / time decay factor.

[0143] where w1 is the click weight, w2 is the share weight, and w3 is the save weight. w1, w2, w3, and the time decay factor are coefficients set according to specific conditions. Generally, the higher the weight, the greater the contribution, and in this algorithm, w3 > w1 > w2. The time decay factor is used to reduce the influence of past time periods on the current hotness, ensuring that the hotness ranking more accurately reflects recent trends. A relatively high time decay factor means that the rankings of popular countries and regions in the past have less influence on the hotness.

[0144] The hotness calculation formula mainly serves as a target function (Target Function) in the training process, guiding the model to learn how to predict the hotness value based on input features. The hotness formula converts behavior data such as the number of clicks, shares, and saves into a specific hotness value through weights (w1, w2, w3). This hotness value becomes the target variable of the model, i.e., the training goal of the model is to predict a value close to the calculation result of the hotness formula based on input data (number of clicks, number of shares, number of saves, time stamp, etc.). During training, the model will continuously adjust the model parameters based on the difference between the predicted hotness value and the target hotness value (such as using the mean square error loss function) to gradually reduce the prediction error. This process is achieved through optimization algorithms (such as gradient descent), ultimately enabling the model to accurately predict hotness based on input data.

[0145] In an embodiment of the present application, the original data needs to be normalized before model training. Numerical features such as click count, share count, and save count are normalized to avoid large feature values that can cause model bias. The present application also converts timestamps into useful numerical features. For example, the date difference (number of days from the current date to the date of the behavior) is introduced into the feature vector for the calculation of the time decay factor.

[0146] In an embodiment of the present application, the preprocessed data is input into the selected model for training. Through repeated iterations, the model learns the relationship between the features in the sample and the popularity label, and optimizes the weight coefficients (w1, w2, w3) and the time decay factor. During training, the loss function can use the cross-entropy loss function, and the optimization algorithm can use gradient descent or other optimization methods to update the model parameters to minimize the loss function.

[0147] In an embodiment of the present application, during model training, the model automatically optimizes the weight coefficients (w1, w2, w3) of clicks, shares, and saves based on the data. Through model learning, the weight coefficients are gradually adjusted to the optimal value, ensuring that the model has a reasonable evaluation of the contribution of different features. For example, if the number of saves has a greater impact on popularity, the model may increase the weight of w3 and reduce the weights of w1 or w2.

[0148] In an embodiment of the present application, the time decay factor is an important parameter in the model that affects the impact of past data on current popularity prediction. By introducing time features (such as timestamp difference), the model can automatically learn the optimal time decay factor. Recent behavior data will have a greater impact on popularity, while distant data will gradually be decayed. The present application can also prevent model overfitting by introducing a regularization term (such as L1 or L2 regularization), while optimizing the learning of the time decay factor, so that the decay factor is not too large or too small.

[0149] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0150] Based on the same inventive concept, the embodiments of the present application also provide a fund risk prediction device, which can be used to implement the fund risk prediction method described in the above embodiments, as described in the following embodiments. Since the principle of solving problems of the fund risk prediction device is similar to that of the fund risk prediction method, the embodiments of the fund risk prediction device can refer to the embodiments of the fund risk prediction method, and the repeated parts will not be described herein. The term "unit" or "module" used below can be a combination of software and / or hardware that can implement a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware or a combination of software and hardware is also possible and is conceived.

[0151] Figure 7 FIG. 1 is a first structural block diagram of a fund risk prediction device according to an embodiment of the present application, as shown in the figure, in an embodiment of the present application, the fund risk prediction device 7 of the present application comprises a basic information acquisition unit 701, a customer financial data acquisition unit 702, a multi-dimensional feature vector generation unit 703 and a risk prediction unit 704. Figure 7

[0152] The basic information acquisition unit 701 is used to acquire the basic information of the customer filled in the fund risk assessment page of the customer. The basic information of the customer includes one or more of the target country information, the target school information, the start and end time information, the education stage information, the reserve fund information and the personal consumption level information.

[0153] The customer financial data acquisition unit 702 is used to acquire the customer financial data of the customer from the bank system. The customer financial data includes one or more of the maximum historical drawdown data, the annual income data, the loan information, the historical credit record, the historical customer risk assessment record and the historical in-house financial product purchase record.

[0154] The multi-dimensional feature vector generation unit 703 is connected with the basic information acquisition unit 701 and acquires the basic information from the basic information acquisition unit 701. The multi-dimensional feature vector generation unit 703 is also connected with the customer financial data acquisition unit 702 and acquires the customer financial data from the customer financial data acquisition unit 702. The multi-dimensional feature vector generation unit 703 is used to generate the fund feature data according to the basic information, generate the customer financial feature data according to the customer financial data, and generate the multi-dimensional feature vector according to the fund feature data and the customer financial feature data.

[0155] ​The risk prediction unit 704 is connected with the multi-dimensional feature vector generation unit 703, and obtains the multi-dimensional feature vector from the multi-dimensional feature vector generation unit 703. The risk prediction unit 704 is configured to input the multi-dimensional feature vector into a preset overseas study fund risk prediction model to obtain a risk prediction result output by the overseas study fund risk prediction model, wherein the overseas study fund risk prediction model is obtained by training a preset classification model by using training samples, the training samples are multi-dimensional feature vectors for model training generated based on historical overseas study fund risk assessment data, and the multi-dimensional feature vectors for model training are labeled with risk prediction results based on expert experience to obtain.

[0156] In an embodiment of the present application, the risk prediction result includes a first prediction result for indicating that there is an overseas study fund risk and a second prediction result for indicating that there is no overseas study fund risk.

[0157] Figure 8 FIG. 7 is a second structural block diagram of the fund risk prediction device according to an embodiment of the present application, as shown in the figure, in an embodiment of the present application, the fund risk prediction device 7 of the present application further includes an overseas study total cost prediction unit 705, an overseas study fund gap determination unit 706, and an overseas study financial product recommendation unit 707. Figure 8

[0158] The overseas study total cost prediction unit 705 is connected with the overseas study basic information acquisition unit 701, and obtains overseas study target country information, overseas study target school information, overseas study start and end time information, overseas study education stage information, and student personal consumption level information from the overseas study basic information acquisition unit 701. The overseas study total cost prediction unit 705 is configured to, if the risk prediction result is the first prediction result, predict an overseas study total cost according to the overseas study target country information, the overseas study target school information, the overseas study start and end time information, the overseas study education stage information, the student personal consumption level information, and a preset overseas study cost prediction model.

[0159] In an embodiment of the present application, the overseas study total cost prediction unit 705 includes a cost feature vector generation module and a cost prediction module.

[0160] ​The cost feature vector generation module is configured to generate a cost feature vector according to the information about the target country for studying abroad, the information about the target school for studying abroad, the information about the start and end time for studying abroad, the information about the education stage for studying abroad, and the information about the personal consumption level of the student.

[0161] The total cost for studying abroad prediction unit 705 is connected with the target country for studying abroad information acquisition unit 701, the target school for studying abroad information acquisition unit 702, the start and end time for studying abroad information acquisition unit 703, the education stage for studying abroad information acquisition unit 704, and the personal consumption level of the student information acquisition unit 706, and obtains the information about the target country for studying abroad, the information about the target school for studying abroad, the information about the start and end time for studying abroad, the information about the education stage for studying abroad, and the information about the personal consumption level of the student from the target country for studying abroad information acquisition unit 701, the target school for studying abroad information acquisition unit 702, the start and end time for studying abroad information acquisition unit 703, the education stage for studying abroad information acquisition unit 704, and the personal consumption level of the student information acquisition unit 706.

[0162] The financial product recommendation unit 707 is connected with the total cost for studying abroad prediction unit 705, and obtains the total cost for studying abroad from the total cost for studying abroad prediction unit 705. The financial product recommendation unit 707 is configured to determine a financial product recommendation scheme according to the total cost for studying abroad and the maximum historical drawdown data accepted by the customer, and send the financial product recommendation scheme to the customer.

[0163] In an embodiment of the present application, the financial product recommendation unit 707 comprises a financial product combination determination module, a historical average annual yield determination module, and a financial product screening module.

[0164] The financial product combination determination module is configured to determine a financial product combination for studying abroad that can meet the total cost for studying abroad according to the information about the reserve fund for studying abroad and the information about the start and end time for studying abroad. The historical average annual yield determination module is configured to determine a historical average annual yield according to the maximum historical drawdown data accepted by the customer. The financial product screening module is configured to screen each financial product for studying abroad in the financial product combination for studying abroad according to the historical average annual yield and the average annual yield of each financial product for studying abroad in the financial product combination for studying abroad, and obtain the financial product recommendation scheme.

[0165] In an embodiment of the present application, the financial product combination for studying abroad comprises fixed-income financial products and equity financial products, and the ratio between the fixed-income financial products and the equity financial products is determined according to historical customer risk assessment records.

[0166] In an embodiment of the present application, the fund risk prediction device 7 of the present application further comprises a popular data acquisition unit and a page processing unit.

[0167] The popular data acquisition unit is configured to acquire popular study abroad country data and popular study abroad school data predicted by a preset popular study abroad destination prediction model. The page processing unit is configured to embed the popular study abroad country data and the popular study abroad school data in drop-down options of input boxes of the study abroad fund risk assessment page, for reference by a customer and input of a target study abroad country and a target study abroad school in the input boxes.

[0168] In an embodiment of the present application, the fund risk prediction device 7 further comprises a historical data acquisition unit, a training sample set establishment unit and a popular study abroad destination prediction model training unit.

[0169] The historical data acquisition unit is configured to acquire historical study abroad fund risk assessment data and parse historical data of each study abroad destination from the historical study abroad fund risk assessment data, the historical data including a click number, a share number, a save number and a timestamp, and the study abroad destination including a study abroad country and a study abroad school. The training sample set establishment unit is configured to establish a training sample set according to the historical data. The popular study abroad destination prediction model training unit is configured to train a preset machine learning model according to the training sample set and a preset heat calculation formula to obtain the popular study abroad destination prediction model.

[0170] The heat calculation formula contains a time decay factor and weight values of the click number, the share number and the save number. During model training, the heat value calculated by the heat calculation formula is set as a target variable of the model, the prediction result of the model is close to the target variable, and the time decay factor and the weight values are continuously optimized according to feedback of a loss function until the prediction accuracy of the trained model reaches a requirement.

[0171] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a computer device is also provided. As shown in the Figure 9 The computer device 8 comprises a memory 801, a processor 802, a communication interface 803 and a communication bus 804, and a computer program stored in the memory 801 and executable on the processor 802, wherein the processor 802 implements the steps in the fund risk prediction method when executing the computer program.

[0172] The memory 801, as a non-transitory computer readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs and units, such as the corresponding program units in the above-mentioned fund risk prediction method embodiments. The memory 801 can include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required by a function; and the data storage area can store data created by the processor and the like. In an embodiment of the present application, the memory 801 can include a high-speed random access memory. In another embodiment of the present application, the memory 801 can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory 801 can optionally include a memory remotely arranged with respect to the processor, and these remote memories can be connected to the processor 802 through a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0173] The processor 802 executes various functional applications and work data processing of the processor 802 by running the non-transitory software programs, instructions and modules stored in the memory, that is, implements the above-mentioned fund risk prediction method. In some embodiments, the processor 802 can be a central processing unit (CPU). In other embodiments, the processor 802 can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components or the like chips, or a combination of the above-mentioned chips.

[0174] The one or more units are stored in the memory 801, and when executed by the processor 802, the fund risk prediction method in the above-mentioned embodiments is executed.

[0175] The above-mentioned computer device 8 can correspond to the above-mentioned corresponding related description and effect for understanding, and will not be repeated here.

[0176] To achieve the above object, according to another aspect of the present application, there is further provided a computer readable storage medium storing a computer program, which, when executed on a computer processor, implements the steps of the above fund risk prediction method. It is understood by those skilled in the art that all or part of the processes of the above fund risk prediction method can be completed by a computer program instructing relevant hardware. The program can be stored in a computer readable storage medium, and when executed, can include the processes of the above fund risk prediction method. The storage medium can be a magnetic disc, an optical disc, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), etc. The storage medium can also include a combination of the above types of memories.

[0177] To achieve the above object, according to another aspect of the present application, there is further provided a computer program product including computer program / instructions, which, when executed by a processor, implements the steps of the above fund risk prediction method.

[0178] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by general computing devices, which can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Alternatively, they can be implemented by program codes executable by computing devices, so that they can be stored in storage devices and executed by computing devices, or they can be respectively manufactured into individual integrated circuit modules, or multiple modules or steps among them can be manufactured into a single integrated circuit module. Thus, the present application is not limited to any specific combination of hardware and software.

[0179] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method of predicting a fund risk, characterized by, The method comprises the following steps: obtaining the basic information of a customer filled in a study abroad fund risk assessment page; obtaining the customer financial data of the customer from a bank system; generating study abroad feature data according to the basic information, generating customer financial feature data according to the customer financial data, and generating a multi-dimensional feature vector according to the study abroad feature data and the customer financial feature data; inputting the multi-dimensional feature vector into a preset study abroad fund risk prediction model to obtain a risk prediction result output by the study abroad fund risk prediction model, wherein the study abroad fund risk prediction model is obtained by training a preset classification model using training samples, the training samples are multi-dimensional feature vectors for model training generated based on historical study abroad fund risk assessment data, and the risk prediction result labels are obtained based on expert experience on the multi-dimensional feature vectors for model training.

2. The method of claim 1, wherein, The risk prediction result includes a first prediction result for indicating the existence of a study abroad fund risk and a second prediction result for indicating the non-existence of a study abroad fund risk. The fund risk prediction method further comprises: if the risk prediction result is the first prediction result, predicting a total study abroad cost according to the study abroad target country information, the study abroad target school information, the study abroad start and end time information, the study abroad education stage information, the student personal consumption level information, and a preset study abroad cost prediction model; determining a study abroad fund gap according to the total study abroad cost and study abroad reserve fund information; determining a study abroad financial product recommendation scheme according to the study abroad fund gap and the maximum historical drawdown data accepted by the customer, and sending the study abroad financial product recommendation scheme to the customer.

3. The method of claim 2, wherein, The method of predicting a total study abroad cost according to the study abroad target country information, the study abroad target school information, the study abroad start and end time information, the study abroad education stage information, the student personal consumption level information, and a preset study abroad cost prediction model comprises: generating a cost feature vector according to the study abroad target country information, the study abroad target school information, the study abroad start and end time information, the study abroad education stage information, and the student personal consumption level information; inputting the cost feature vector into the study abroad cost prediction model to obtain a total study abroad cost prediction value output by the study abroad cost prediction model, wherein the study abroad cost prediction model is obtained by training a preset classification model using training samples, the training samples are cost feature vectors for model training generated based on historical data, and the total study abroad cost labels are obtained based on real study abroad costs on the cost feature vectors for model training.

4. The method of claim 2, wherein, The method of determining a study abroad financial product recommendation scheme according to the study abroad fund gap and the maximum historical drawdown data accepted by the customer comprises: determining a study abroad financial product combination capable of meeting the study abroad fund gap according to the study abroad reserve fund information and the study abroad start and end time information; determining a historical average annual yield according to the maximum historical drawdown data accepted by the customer; According to the historical average annual yield and the average annual yield of each study abroad financial product in the study abroad financial product portfolio, each study abroad financial product in the study abroad financial product portfolio is screened to obtain the study abroad financial product recommendation scheme.

5. The method of claim 4, wherein, The study abroad financial product portfolio includes fixed-income financial products and equity financial products. The ratio between the fixed-income financial products and the equity financial products is determined according to historical customer risk assessment records.

6. The method of claim 1, wherein, Before obtaining the study abroad basic information filled in by the customer in the study abroad fund risk assessment page, the fund risk prediction method further comprises: Obtain the popular study abroad country data and the popular study abroad school data predicted by the preset popular study abroad destination prediction model; Embed the popular study abroad country data and the popular study abroad school data in the drop-down options of the input box of the study abroad fund risk assessment page for the customer to refer to and complete the input of the target country and the target school in the input box.

7. The method of claim 6, wherein, The fund risk prediction method further comprises: Obtain historical study abroad fund risk assessment data and parse historical data of each study abroad destination from the historical data, the historical data including click count, share count, save count and timestamp, and the study abroad destination including study abroad country and study abroad school; Establish a training sample set according to the historical data; According to the training sample set and a preset heat calculation formula, train on the basis of a preset machine learning model to obtain the popular study abroad destination prediction model, the heat calculation formula containing a time decay factor and weight values of click count, share count and save count; during model training, the heat value calculated by the heat calculation formula is the target variable of the model, the prediction result of the model approaches the target variable, and the time decay factor and the weight values are continuously optimized according to the feedback of the loss function until the prediction accuracy of the trained model meets the requirements.

8. The method of claim 1, wherein, The study abroad basic information includes one or more of study abroad target country information, study abroad target school information, study abroad start and end time information, study abroad education stage information, study abroad reserve fund information and student personal consumption level information.

9. The method of claim 1, wherein, The customer financial data includes one or more of maximum historical drawdown data accepted by the customer, customer family annual income data, loan information, historical credit records, historical customer risk assessment records and historical industry financial product purchase records.

10. A fund risk prediction apparatus characterized by comprising: It comprises: A study abroad basic information acquisition unit for acquiring study abroad basic information filled in by the customer in the study abroad fund risk assessment page; A customer financial data acquisition unit for acquiring customer financial data of the customer from a bank system; A multi-dimensional feature vector generation unit for generating study abroad feature data according to the study abroad basic information, generating customer financial feature data according to the customer financial data, and generating a multi-dimensional feature vector according to the study abroad feature data and the customer financial feature data; A risk prediction unit is configured to input the multi-dimensional feature vector into a preset overseas study fund risk prediction model to obtain a risk prediction result output by the overseas study fund risk prediction model, wherein the overseas study fund risk prediction model is obtained by training a preset classification model using training samples, the training samples are multi-dimensional feature vectors generated based on historical overseas study fund risk assessment data for model training, and the multi-dimensional feature vectors are labeled with risk prediction results based on expert experience.

11. The fund risk prediction apparatus according to claim 10, wherein The risk prediction result includes a first prediction result indicating that there is an overseas study fund risk and a second prediction result indicating that there is no overseas study fund risk. The fund risk prediction device further includes: An overseas study total cost prediction unit is configured to, if the risk prediction result is the first prediction result, predict an overseas study total cost according to overseas study target country information, overseas study target school information, overseas study start and end time information, overseas study education stage information, student personal consumption level information, and a preset overseas study cost prediction model. An overseas study fund gap determination unit is configured to determine an overseas study fund gap according to the overseas study total cost and overseas study reserve fund information. An overseas study financial product recommendation unit is configured to determine an overseas study financial product recommendation scheme according to the overseas study fund gap and historical maximum drawdown data accepted by a client, and send the overseas study financial product recommendation scheme to the client.

12. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the steps of the fund risk prediction method in any one of claims 1 to 9.

13. A computer readable storage medium having stored thereon computer programs / instructions, characterized in that, The computer program / instruction is executed by the processor to implement the steps of the fund risk prediction method in any one of claims 1 to 9.

14. A computer program product comprising computer programs / instructions, characterized in that, The computer program / instruction is executed by the processor to implement the steps of the fund risk prediction method in any one of claims 1 to 9.

Images