A risk prediction method, device, storage medium and program product

Abstract

This invention discloses a risk prediction method, device, storage medium, and program product, relating to the fields of fintech and artificial intelligence. The method includes: collecting multiple current state descriptions of a target object on multiple dedicated public information platforms; calculating multiple internal description change feature values ​​of the target object based on multiple current internal description feature values ​​corresponding to each current state description and multiple historical internal description feature values ​​of the target object; calculating multiple external description feature values ​​of the target object based on title data associated with the target object published on multiple general information publishing websites; and inputting each internal description change feature value and each external description feature value into a pre-built risk prediction model to obtain a risk prediction result for the target object. The different model layers in the risk prediction model are configured to execute on different servers in a server cluster, improving the efficiency and accuracy of risk prediction.

Description

Technical Field

[0001] This invention relates to the fields of financial technology and artificial intelligence, and in particular to a risk prediction method, device, storage medium, and program product. Background Technology

[0002] Currently, in order to proactively manage the potential uncertainties of a target object, risk prediction is usually performed on the target object.

[0003] In existing technologies, risk prediction is typically performed on a risk prediction model deployed on a single server, based on the target object's current state description information.

[0004] However, risk prediction models deployed on a single server are easily limited by the server's hardware computing power, memory, and concurrency processing limits, resulting in slow model inference response and low risk prediction efficiency. Secondly, since the current state description of the target object cannot accurately reflect its actual state, using existing technologies to predict the risk of the target object often fails to yield accurate results. Summary of the Invention

[0005] This invention provides a risk prediction method, apparatus, device, storage medium, and program product, which improves the efficiency and accuracy of risk prediction for target objects.

[0006] In a first aspect, embodiments of the present invention provide a risk prediction method, including:

[0007] When the risk warning conditions for the target object are met, according to the pre-established data collection strategy table, multiple current status descriptions of the target object within a preset time interval are collected on multiple dedicated public platforms for object information.

[0008] Based on the current state description information, calculate multiple current internal description feature values ​​of the target object, and based on the multiple current internal description feature values ​​and the multiple historical internal description feature values ​​of the target object obtained from historical calculation, calculate multiple internal description change feature values ​​of the target object.

[0009] Collect title data associated with the target object published within a preset time interval from multiple general information publishing websites, and calculate multiple external descriptive feature values ​​of the target object within the preset time interval based on each title data.

[0010] Each internal description feature value and each external description feature value are input into a pre-built risk prediction model to obtain the risk prediction result of the target object. The different model layers in the risk prediction model are configured to be executed on different servers in the server cluster.

[0011] Secondly, embodiments of the present invention also provide a risk prediction device, comprising:

[0012] The information acquisition module is used to collect multiple current status descriptions of the target object within a preset time interval on multiple dedicated public information release platforms for the target object, based on a pre-established data collection strategy table, when the risk warning conditions for the target object are met.

[0013] The change value determination module is used to calculate multiple current internal description feature values ​​of the target object based on the current state description information, and to calculate multiple internal description change feature values ​​of the target object based on the multiple current internal description feature values ​​and the multiple historical internal description feature values ​​of the target object obtained from historical calculation.

[0014] The feature value determination module is used to collect title data associated with the target object published within a preset time interval from multiple general information publishing websites, and to calculate multiple external descriptive feature values ​​of the target object within the preset time interval based on each title data.

[0015] The risk prediction module is used to input the internal description change feature values ​​and the external description feature values ​​into the pre-built risk prediction model to obtain the risk prediction results of the target object. The different model layers in the risk prediction model are configured to be executed on different servers in the server cluster.

[0016] Thirdly, embodiments of the present invention also provide an electronic device, the electronic device comprising:

[0017] At least one processor; and

[0018] A memory that is communicatively connected to at least one processor; wherein,

[0019] The memory stores a computer program that can be executed by at least one processor, such that the at least one processor is able to perform the risk prediction method provided in any embodiment of the present invention.

[0020] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing computer instructions for causing a processor to execute and implement the risk prediction method provided in any embodiment of the present invention.

[0021] Fifthly, embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the risk prediction method provided in any embodiment of the present invention.

[0022] The technical solution of this invention calculates multiple internal description change feature values ​​of a target object using multiple current internal description feature values ​​and multiple historical internal description feature values; calculates multiple external description feature values ​​of the target object within a preset time interval based on title data associated with the target object; and inputs each internal description change feature value and each external description feature value into a pre-built risk prediction model to obtain the risk prediction result of the target object. This avoids the situation in the prior art where risk prediction of the target object is performed based on the current state description information of the target object using a risk prediction model deployed on a single server, resulting in low efficiency and accuracy in risk prediction. This invention improves the efficiency and accuracy of risk prediction of the target object.

[0023] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart of a risk prediction method provided in Embodiment 1 of the present invention;

[0026] Figure 2 This is a flowchart of another risk prediction method provided in Embodiment 2 of the present invention;

[0027] Figure 3 This is a schematic diagram of the structure of a risk prediction device provided in Embodiment 3 of the present invention;

[0028] Figure 4 This is a schematic diagram of the structure of an electronic device provided in Embodiment 4 of the present invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0031] Example 1

[0032] Figure 1 This is a flowchart of a risk prediction method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where risk prediction is performed on the operational health status or business condition of a target object. Figure 1 As shown, the risk prediction method disclosed in this embodiment is executed by a server cluster and includes:

[0033] S110. When the risk warning conditions for the target object are met, according to the pre-established data collection strategy table, collect multiple current status descriptions of the target object within a preset time interval on multiple dedicated public platforms for object information.

[0034] In this embodiment, the target object can be set according to user needs, such as an already launched software product, or an enterprise that entrusts a bank to handle financial business or receives financial services from a bank. Risk warning conditions can be understood as conditions requiring risk prediction for the target object. For example, when the target object is an already launched software product, if the download volume of the target object within a set time period is lower than a preset minimum download volume, or the negative review rate is higher than a preset maximum negative review rate, then the risk warning conditions for the target object are determined to be met. When the target object is an enterprise, if the target object's credit score is lower than a preset score threshold, or if it applies for large loans multiple times consecutively, then the risk warning conditions for the target object are determined to be met.

[0035] The data collection strategy table can be used to record the mapping relationship between the status description information of the target object and the dedicated public platform for publishing object information. The dedicated public platform for publishing object information can be understood as a platform that can provide the status description information of the target object. The status description information can reflect the operational status of the target object. For example, when the target object is a software product that has been launched, the status description information may include product performance information such as normal service time, total service time, resource consumption, and total resources; when the target object is a company, the status description information may include operational description information such as net profit, total operating revenue, current liabilities, and total liabilities.

[0036] In this step, specifically, the data collection strategy table can be queried to obtain the dedicated public publishing platform for object information corresponding to each current state description. Then, on the dedicated public publishing platform for object information corresponding to each current state description, multiple current state descriptions of the target object within a preset time interval can be collected.

[0037] S120. Based on the current state description information, calculate multiple current internal description feature values ​​of the target object, and based on the multiple current internal description feature values ​​and the multiple historical internal description feature values ​​of the target object obtained from historical calculation, calculate multiple internal description change feature values ​​of the target object.

[0038] In this embodiment, the current internal description feature value can be used to assess the current operating status of the target object. Historical internal description feature values ​​can be used to assess the historical operating status of the target object. There can be various types of internal description feature values. For example, when the target object is an online software product, the internal description feature values ​​may include service availability and resource consumption ratio; when the target object is an enterprise, the internal description feature values ​​may include return on net assets and debt-to-equity ratio.

[0039] Specifically, this step involves calculating the ratios of multiple sets of current state description information to obtain multiple current internal descriptive feature values ​​of the target object, and calculating the ratios of multiple sets of historical state description information to obtain multiple historical internal descriptive feature values ​​of the target object. Taking a company as an example, if the current state description information of the target object includes net profit and net assets, the return on net assets of the target object can be determined based on the ratio of net profit to net assets.

[0040] Then, the change value of each current internal description feature value relative to the corresponding historical internal description feature value can be calculated to obtain the internal description change feature value corresponding to each current internal description feature value of the target object.

[0041] S130. Collect title data associated with the target object published within a preset time interval from multiple general information publishing websites, and calculate multiple external descriptive feature values ​​of the target object within the preset time interval based on each title data.

[0042] In this step, specifically, target title data containing preset words and the preset topics to which each preset word belongs can be identified from the title data associated with the target object. Then, based on the preset words contained in each target title data and the preset topics to which each preset word belongs, the number of preset words contained in all target title data, the number of preset topics to which all target title data belongs, the weighted number of words corresponding to all target title data, and the weighted number of topics corresponding to all target title data can be determined. Finally, at least one of the following can be used as multiple external descriptive feature values ​​of the target object within a preset time interval: the number of target title data, the number of preset words contained in all target title data, the number of preset topics to which all target title data belongs, the weighted number of words corresponding to all target title data, and the weighted number of topics corresponding to all target title data. These external descriptive feature values ​​can be used to reflect the evaluation of the target object's operational status by external personnel. External personnel can include news media and relevant personnel with actual connections to the target object. For example, when the target object is an online software product, relevant personnel with actual connections to the target object can include software users; when the target object is a company, relevant personnel with actual connections to the target object can include those involved in transactions.

[0043] S140. Input the internal description change feature values ​​and the external description feature values ​​into the pre-built risk prediction model to obtain the risk prediction results of the target object.

[0044] The risk prediction model can be trained using multiple historical internal descriptive change feature values ​​and multiple historical external descriptive feature values ​​corresponding to multiple sample objects as training samples. The risk prediction model can be a decision tree model.

[0045] In this step, specifically, since deploying the risk prediction model on a single server is easily limited by the hardware computing power, memory, and concurrency processing limit of the single server, the model inference response is slow. Therefore, the technical solution of this embodiment configures different model layers in the risk prediction model to be executed on different servers in the server cluster, so as to improve the efficiency of using the risk prediction model to predict the risk of the target object.

[0046] The technical solution of this embodiment solves the problem that the existing technology, which relies solely on a risk prediction model deployed on a single server to predict the risk of a target object within a preset time interval, suffers from low efficiency and accuracy. This is achieved by: collecting multiple current state descriptions of the target object within a preset time interval based on a pre-established data collection strategy table; calculating multiple current internal description feature values ​​of the target object based on each current state description; calculating multiple internal description change feature values ​​of the target object based on these current internal description feature values ​​and historically calculated multiple internal description feature values ​​of the target object; collecting title data associated with the target object published within a preset time interval from multiple general information publishing websites; calculating multiple external description feature values ​​of the target object within a preset time interval based on each title data; and inputting these internal description change feature values ​​and external description feature values ​​into a pre-built risk prediction model to obtain the risk prediction result for the target object. This approach improves the efficiency and accuracy of risk prediction for target objects.

[0047] Example 2

[0048] Figure 2 This is a flowchart of another risk prediction method provided by Embodiment 2 of the present invention. This embodiment is a further optimization and extension based on the above embodiments and can be combined with various optional technical solutions in the above embodiments.

[0049] like Figure 2 As shown, this embodiment discloses a risk prediction method, including:

[0050] S210. When the risk warning conditions for the target object are met, according to the pre-established data collection strategy table, collect multiple current status descriptions of the target object within a preset time interval on multiple dedicated public platforms for object information.

[0051] Optionally, before collecting multiple current status descriptions of the target object within a preset time interval on multiple dedicated public platforms for object information according to a pre-established data collection strategy table, the process further includes: sequentially acquiring current status descriptions from all status descriptions and determining at least one alternative status information table that records the current status descriptions; acquiring alternative dedicated public platforms for object information to publish each alternative status information table on each dedicated public platform for object information, and acquiring the data accuracy index of each alternative status information table published by each alternative dedicated public platform for object information; acquiring the target dedicated public platform for object information and the target status information table corresponding to the maximum data accuracy index; adding the mapping relationship between the current status descriptions, the target dedicated public platforms for object information, and the target status information table to the data collection strategy table; and returning to execute the operation of sequentially acquiring current status descriptions from all operational descriptions until the processing of all status descriptions is completed.

[0052] Specifically, each alternative status information table can be evaluated from the perspectives of accuracy, completeness, consistency, timeliness, and reliability to obtain the data accuracy index of each alternative status information table.

[0053] By using the above settings, we can obtain more accurate information about the current state of the target object, thereby improving the accuracy of risk prediction for the target object.

[0054] S220. Based on the current state description information, calculate multiple current internal description feature values ​​of the target object, and based on the multiple current internal description feature values ​​and the multiple historical internal description feature values ​​of the target object obtained from historical calculation, calculate multiple internal description change feature values ​​of the target object.

[0055] Optionally, before calculating the multiple internal description change feature values ​​of the target object based on multiple current internal description feature values ​​and multiple historical internal description feature values ​​of the target object obtained from historical calculations, the method further includes: obtaining multiple historical time intervals before a preset time interval, and collecting multiple historical state description information of the target object in each historical time interval on multiple dedicated public platforms for object information according to a pre-established data collection strategy table; and calculating multiple historical internal description feature values ​​of the target object in each historical time interval based on each historical state description information.

[0056] Further, based on multiple current internal description feature values ​​and multiple historical internal description feature values ​​of the target object calculated historically, multiple internal description change feature values ​​of the target object are calculated. This can include: classifying each current internal description feature value into a first type of current internal description feature value and a second type of current internal description feature value, and classifying each historical internal description feature value into a first type of historical internal description feature value and a second type of historical internal description feature value. Then, based on multiple first type current internal description feature values ​​of the target object within a preset time interval, a first type current multivariate vector can be generated, and based on multiple first type historical internal description feature values ​​of the target object within each historical time interval, a first type historical multivariate vector corresponding to each historical time interval can be generated. Afterwards, the average value of each first type historical multivariate vector can be calculated to obtain a first mean, and a subtraction operation can be performed between the first type current multivariate vector and the first mean to obtain a first relative mean change value corresponding to the target object.

[0057] Finally, the first historical time interval closest to the end time of the preset time interval and the second historical time interval closest to the end time of the first historical time interval can be obtained from multiple historical time intervals. The first change feature value is obtained by subtracting the first-type historical multivariate vector corresponding to the first historical time interval from the first-type current multivariate vector, and the second change feature value is obtained by subtracting the first-type historical multivariate vector corresponding to the second historical time interval from the first-type current multivariate vector. The first relative mean change value, the first change feature value, and the second change feature value are collectively used as multiple first-type internal descriptive change feature values ​​corresponding to the target object.

[0058] Similarly, a second-type current multivariate vector can be generated based on multiple second-type current internal descriptive feature values ​​of the target object within a preset time interval. Then, a second-type historical multivariate vector corresponding to each historical time interval can be generated based on multiple second-type historical internal descriptive feature values ​​of the target object within each historical time interval. These second-type internal descriptive feature values ​​can be of various types, such as the producer price index, fixed asset investment completion amount, and financing scale. Next, the average of each second-type historical multivariate vector can be calculated to obtain a second mean. Subtraction between the second-type current multivariate vector and the second mean yields a first relative mean change value corresponding to the target object. Finally, subtracting the second-type historical multivariate vector corresponding to the first historical time interval from the second-type current multivariate vector yields a third change feature value, and subtracting the second-type historical multivariate vector corresponding to the second historical time interval from the second-type current multivariate vector yields a fourth change feature value. The second relative mean change value, the third change feature value, and the fourth change feature value are collectively used as multiple second-type internal descriptive change feature values ​​corresponding to the target object.

[0059] Taking a company as an example, Table 1 is a table for recording first-type internal description feature values ​​according to an embodiment of the present invention. As shown in Table 1, multiple first-type current internal description feature values ​​of the target company in the current quarter, as well as multiple first-type historical internal description feature values ​​in the first, second, and third historical quarters, can be calculated from the dimensions of profitability, capital structure, operating capacity, and cash reserves. The first historical quarter is the quarter preceding the current quarter, the second historical quarter is the quarter preceding the first historical quarter, and the third historical quarter is the quarter preceding the second historical quarter. In practical applications, the first-type internal description feature values ​​of the target company in the third quarter of the previous year can be calculated from January to April each year; the first-type internal description feature values ​​of the target company in the entire previous year and the first quarter of the current year can be calculated from May to August each year; the first-type internal description feature values ​​of the target company in the first half of the current year can be calculated from September to October each year; and the first-type internal description feature values ​​of the target company in the third quarter of the current year can be calculated from November to December each year.

[0060] Then, after determining the first type of current multivariate vector corresponding to the current quarter, and the first type of historical multivariate vector corresponding to each historical quarter, the formula can be used to... Determine the first relative mean change value corresponding to the target object. Wherein, △X' t X is the first relative mean. t X is the first type of current multivariate vector corresponding to the current quarter. t-1 X is the first type of historical multivariate vector corresponding to the first historical quarter. t-2 X is the first type of historical multivariate vector corresponding to the second historical quarter. t-3 This is the first type of historical multivariate vector corresponding to the third historical quarter. Finally, it can be expressed by the formula △X. t =X t -X t-1 Determine the first change characteristic value corresponding to the target object using the formula △X. t-1 =X t-1 -X t-2 Determine the second change characteristic value corresponding to the target object. Wherein, △X t The first characteristic value of change, △X t-1 This is the second characteristic value of change.

[0061]

[0062]

[0063] Similarly, multiple current internal description features (IDLs) of the target object in the current quarter, as well as multiple historical IDLs in the first, second, and third historical quarters, can be calculated from the perspective of quarterly year-on-year growth rate. These IDLs can include the Producer Price Index, fixed asset investment, and total social financing. In practical applications, the IDLs of the target object in the third quarter of the previous year can be calculated in January of each year; in the fourth quarter of the previous year from February to April; in the first quarter of the current year from May to July; in the second quarter of the current year from August to October; and in the third quarter of the current year from November to December. Then, based on these IDLs, multiple IDL change features corresponding to the target object can be determined.

[0064] The advantage of this setup is that, since the healthy operating status of the target object has a certain continuity, by calculating multiple historical internal descriptive feature values ​​of the target object in each historical time interval, the impact of the historical internal descriptive feature values ​​of the target object on the risk prediction results can be fully considered, thus providing a basis for accurately predicting the risk of the target object.

[0065] S230. Collect title data associated with the target object published within a preset time interval from multiple general information publishing websites, and determine the target title data containing preset words in the title data associated with the target object, and the preset topic to which each preset word in the target title data belongs, based on a preset dictionary.

[0066] The preset dictionary includes a correspondence between preset topics and preset vocabulary. Risk vocabulary can be understood as words used to reflect the risks faced by the target object.

[0067] S240. Based on the preset words contained in each target title data, determine the number of preset words contained in all target title data, and based on the preset topic to which each preset word in each target title data belongs, determine the number of preset topics to which all target title data belongs.

[0068] In this step, specifically, the number of preset words contained in each target title data can be determined based on the preset words contained in each target title data, and the number of preset words contained in each target title data can be added together to determine the total number of preset words contained in all target title data.

[0069] Based on the preset topics to which each preset word in each target title data belongs, determine the number of preset topics to which each target title data belongs, and add the number of preset topics to which each target title data belongs to to obtain the total number of preset topics to which all target title data belongs.

[0070] For example, the number of preset words contained in all target title data can be determined using the following formula:

[0071] NW it =∑ j∈D(t) NW ij

[0072] Among them, NW it D(t) represents the preset number of words contained in all target title data, and NW represents the preset time interval. ij This represents the number of preset keywords contained in the title data published by the issuer within a preset time interval, where i represents the issuer and j represents the title data.

[0073] The number of preset topics to which all target title data belongs is determined by the following formula:

[0074] NT it =∑ j∈D(t) NT ij

[0075] Among them, NT it The preset number of topics to which all target title data belongs, NT ij The number of preset topics included in the title data published by the issuer.

[0076] S250. Based on the preset topic to which each preset word in each target title data belongs, determine the number of preset words belonging to different preset topics in each target title data, and perform a weighted calculation on the number of preset words belonging to different preset topics in each target title data to obtain the weighted word count corresponding to each target title data.

[0077] Optionally, a weighted calculation is performed on the number of preset words belonging to different preset topics in each target title data to obtain the weighted word count corresponding to each target title data. This includes: obtaining the number of preset words corresponding to each preset topic in all target title data, and determining the weight coefficient corresponding to each preset topic based on the number of preset words corresponding to each preset topic and the total number of title data associated with the target object; and performing a weighted calculation on the number of preset words belonging to different preset topics in each target title data based on the weight coefficient corresponding to each preset topic to obtain the weighted word count corresponding to each target title data.

[0078] Specifically, the total number of title data can be divided by the number of preset words corresponding to each preset theme to obtain the result of the division with respect to each preset theme. Then, the logarithm corresponding to each division result can be calculated to obtain the weight coefficient corresponding to each preset theme.

[0079] By setting the above parameters, a larger weight coefficient can be assigned to low-frequency preset topics, and a smaller weight coefficient can be assigned to high-frequency preset topics. This fully considers the impact of the frequency of occurrence of preset topics on risk prediction and provides a basis for accurate risk prediction of target objects.

[0080] Optionally, the number of preset words belonging to different preset topics in each target title data is weighted and calculated to obtain the weighted number of words corresponding to each target title data. This includes: determining the weight coefficient corresponding to each preset topic based on the arrangement position of each preset topic in the preset dictionary; and weighting the number of preset words belonging to different preset topics in each target title data based on the weight coefficient corresponding to each preset topic to obtain the weighted number of words corresponding to each target title data.

[0081] Specifically, since the earlier a preset topic appears in the preset dictionary, the more important it is, the weight coefficient corresponding to each preset topic can be determined based on its position in the preset dictionary and the total number of preset topics included in the preset dictionary.

[0082] For example, when a preset topic is ranked in the first third of a preset dictionary, the weight coefficient of the preset topic can be determined to be 3; when a preset topic is ranked in the middle third of a preset dictionary, the weight coefficient of the target preset topic can be determined to be 2; and when a preset topic is ranked in the last third of a preset dictionary, the weight coefficient of the target preset topic can be determined to be 1.

[0083] The above settings allow for full consideration of the impact of the importance of the preset topic on the risk prediction results, providing a foundation for accurate risk prediction of the target object.

[0084] S260. Determine the number of weighted words corresponding to all target title data based on the number of weighted words corresponding to each target title data.

[0085] In this step, specifically, the weighted word counts corresponding to each target title data can be added together to obtain the weighted word counts corresponding to all target title data.

[0086] For example, when determining the weighted number of words corresponding to each target title data based on the preset number of words corresponding to each preset topic, the weighted number of words corresponding to all target title data can be determined by the following formula:

[0087]

[0088] Among them, NWITF it NW represents the weighted number of words corresponding to all target title data. ijk N represents the number of preset words belonging to preset topic k in the title data j published by issuer i within a preset time interval, where k is any preset topic in the preset dictionary. k The number of preset words corresponding to the preset theme, where N is the total number of title data associated with the target object. The weighting coefficients are those corresponding to the preset topic k.

[0089] By determining the weight coefficient corresponding to each preset topic based on its position in the preset dictionary, the weighted number of words corresponding to all target title data can be determined using the following formula:

[0090] NWTDW it =∑ j∈D(t) ∑ k NW ijk Score k

[0091] Among them, NWITF it Score is the weighted number of words corresponding to all target title data. k The weighting coefficients are those corresponding to the preset topic k.

[0092] S270. Perform a weighted calculation on the preset topics to which each target title data belongs, to obtain the number of weighted topics corresponding to each target title data, and determine the number of weighted topics corresponding to all target title data based on the number of weighted topics corresponding to each target title data.

[0093] In this step, specifically, the preset topics to which each target title data belongs can be weighted according to the weight coefficients corresponding to each preset topic, to obtain the weighted topic count corresponding to each target title data. Then, the weighted topic counts corresponding to each target title data can be added together to obtain the weighted topic count corresponding to all target title data.

[0094] For example, when determining the weighted number of words corresponding to each target title data based on the number of preset words corresponding to each preset topic, the weighted number of topics corresponding to all target title data can be determined by the following formula:

[0095]

[0096] Among them, NTITF it TMinioned is the weighted number of topics corresponding to all target title data. ijk As a dummy variable, if publisher i publishes title data j within a preset time interval containing preset keywords belonging to preset theme k, then TMenioned ijk The value is 1. If the title data j published by issuer i within the preset time interval does not contain preset words belonging to the preset theme k, then TMenioned ijk The value is 0.

[0097] By determining the weight coefficient corresponding to each preset topic based on its position in the preset dictionary, the weighted number of topics corresponding to all target title data can be determined using the following formula:

[0098] NTTDW it =∑ j∈D(t) ∑ k TMenioned ijk Score k

[0099] Among them, NTTDW it This represents the weighted number of topics corresponding to all target title data.

[0100] S280. The number of target title data, the number of preset words contained in all target title data, the number of preset topics to which all target title data belongs, the number of weighted words corresponding to all target title data, and the number of weighted topics corresponding to all target title data are used together as the external descriptive feature value of the target object within the preset time interval.

[0101] In this step, specifically, the sum of each target title data can be calculated to obtain the number of target title data related to the target object published within the preset time interval.

[0102] For example, the number of target title data associated with the target object published within a preset time interval can be determined using the following formula:

[0103] NN it =∑ j∈D(t) IsNeg ij

[0104] Among them, NN it IsNeg represents the number of target title data associated with the target object published within a preset time interval. ij As a dummy variable, IsNeg is defined if the title data j published by publisher i within a preset time interval contains preset keywords. ij The value is 1. If the title data j published by issuer i within the preset time interval does not contain the preset words, then IsNeg is set to 1. ij The value is 0.

[0105] The above settings fully consider the impact of the quantity of target title data, the number of preset words contained in all target title data, the number of preset topics to which all target title data belongs, and the preset topic type to which the target title data belongs on the risk prediction results, thus providing a foundation for accurately predicting the risks of target objects.

[0106] S290. Input the internal description change characteristic values ​​and the external description characteristic values ​​into the pre-built risk prediction model to obtain the risk prediction results of the target object.

[0107] In this model, different model layers are configured to run on different servers in the server cluster.

[0108] Optionally, before inputting the internal description change feature values ​​and external description feature values ​​into a pre-built risk prediction model to obtain the risk prediction result for the target object, the method further includes: obtaining multiple historical internal description change feature values ​​and multiple historical external description feature values ​​corresponding to multiple sample objects; using a five-fold cross-validation method, training five risk test models based on the historical internal description change feature values ​​and the historical external description feature values, and obtaining the recall rates corresponding to the five risk test models respectively; calculating the average of the recall rates to obtain the average recall rate, and taking the recall rate with the smallest deviation from the average recall rate and the largest actual value as the target recall rate; and constructing a risk prediction model based on the hyperparameter combination corresponding to the target recall rate, according to the historical internal description change feature values ​​and the historical external description feature values.

[0109] Specifically, the historical internal descriptive feature values ​​and the historical external descriptive feature values ​​can be randomly divided into five subsets of roughly equal size. Then, by using the five subsets in turn as the validation set and the subsets not selected as the validation set as the training set, five risk testing models and their corresponding recall rates can be trained.

[0110] The advantage of this setup is that by using the recall rate that minimizes the deviation from the average recall rate and has the largest actual value as the target recall rate, and constructing a risk prediction model based on the hyperparameter combination corresponding to the target recall rate, the generalization ability and prediction accuracy of the risk prediction model can be improved, providing a foundation for accurately predicting the risk of target objects.

[0111] The technical solution of this embodiment takes into full account the impact of the number of target title data, the number of preset words contained in all target title data, the number of preset topics to which all target title data belongs, the number of weighted words corresponding to all target title data, and the number of weighted topics corresponding to all target title data on the risk prediction results, and provides a basis for accurately predicting the risk of the target object within a preset time interval.

[0112] The requirements state that the target object information collected by this invention is information and data authorized by the target object owner or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with the relevant laws, regulations, and standards of the relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for the target object to choose to authorize or refuse.

[0113] Example 3

[0114] Figure 3 This is a schematic diagram of a risk prediction device according to Embodiment 3 of the present invention. This embodiment is applicable to situations where risk prediction is performed on the operational health status or business conditions of a target object. The risk prediction device can be implemented in hardware and / or software and can be configured in electronic devices such as computers.

[0115] like Figure 3 As shown, the risk prediction device disclosed in this embodiment includes: an information acquisition module 31, a change value determination module 32, a feature value determination module 33, and a risk prediction module 34, wherein:

[0116] Information acquisition module 31 is used to collect multiple current status descriptions of the target object within a preset time interval on multiple dedicated public information release platforms for multiple objects, based on a pre-established data collection strategy table, when the risk warning conditions for the target object are met.

[0117] The change value determination module 32 is used to calculate multiple current internal description feature values ​​of the target object based on each current state description information, and to calculate multiple internal description change feature values ​​of the target object based on the multiple current internal description feature values ​​and multiple historical internal description feature values ​​of the target object obtained from historical calculation.

[0118] The feature value determination module 33 is used to collect title data associated with the target object published within a preset time interval from multiple general information publishing websites, and to calculate multiple external descriptive feature values ​​of the target object within the preset time interval based on each title data.

[0119] The risk prediction module 34 is used to input the internal description change feature values ​​and the external description feature values ​​into the pre-built risk prediction model to obtain the risk prediction results of the target object.

[0120] The technical solution in this embodiment, through the cooperation of the information acquisition module 31, the change value determination module 32, the feature value determination module 33, and the risk prediction module 34, solves the problem that the existing technology only uses a risk prediction model deployed on a single server to predict the risk of the target object based on the state description information of the target object at the current moment, resulting in low efficiency and accuracy in risk prediction of the target object. It improves the efficiency and accuracy of risk prediction of the target object.

[0121] Optionally, the eigenvalue determination module 33 includes:

[0122] The preset vocabulary recognition unit is used to determine, based on a preset dictionary, target title data containing preset vocabulary in title data associated with the target object, and the preset topic to which each preset vocabulary in the target title data belongs;

[0123] The preset dictionary includes the correspondence between preset topics and preset vocabulary;

[0124] The topic quantity determination unit is used to determine the number of preset words contained in all target title data based on the preset words contained in each target title data, and to determine the number of preset topics to which all target title data belongs based on the preset topics to which each preset word contained in each target title data belongs.

[0125] The vocabulary weighting unit is used to determine the number of preset words belonging to different preset topics in each target title data according to the preset topics to which each preset word belongs, and to perform weighted calculation on the number of preset words belonging to different preset topics in each target title data to obtain the weighted vocabulary number corresponding to each target title data.

[0126] The vocabulary count summation unit is used to determine the weighted vocabulary count corresponding to all target title data based on the weighted vocabulary count corresponding to each target title data.

[0127] The topic quantity weighting unit is used to perform weighted calculation on the preset topics to which each target title data belongs, to obtain the weighted topic quantity corresponding to each target title data, and to determine the weighted topic quantity corresponding to all target title data based on the weighted topic quantity corresponding to each target title data.

[0128] The external feature determination unit is used to take the number of target title data, the number of preset words contained in all target title data, the number of preset topics to which all target title data belongs, the number of weighted words corresponding to all target title data, and the number of weighted topics corresponding to all target title data as the external descriptive feature values ​​of the target object within a preset time interval.

[0129] Optional, word-weighted units include:

[0130] The first weight determination subunit is used to obtain the number of preset words corresponding to each preset topic in all target title data, and determine the weight coefficient corresponding to each preset topic based on the number of preset words corresponding to each preset topic and the total number of title data associated with the target object.

[0131] The first vocabulary weighting subunit is used to calculate the number of preset words belonging to different preset topics in each target title data according to the weight coefficients corresponding to each preset topic, so as to obtain the weighted vocabulary number corresponding to each target title data.

[0132] The second weight determination subunit is used to determine the weight coefficient corresponding to each preset topic based on the arrangement position of each preset topic in the preset dictionary.

[0133] The second vocabulary weighting subunit is used to calculate the number of preset words belonging to different preset topics in each target title data according to the weight coefficients corresponding to each preset topic, so as to obtain the weighted vocabulary number corresponding to each target title data.

[0134] Optionally, the device also includes a model training module, which is used to: acquire multiple historical internal descriptive change feature values ​​and multiple historical external descriptive feature values ​​corresponding to multiple sample objects; use a five-fold cross-validation method to train five risk testing models and their corresponding recall rates based on the historical internal descriptive change feature values ​​and the historical external descriptive feature values; calculate the average recall rate by averaging the recall rates, and take the recall rate with the smallest deviation from the average recall rate and the largest actual value as the target recall rate; and construct a risk prediction model based on the hyperparameter combination corresponding to the target recall rate and the historical internal descriptive change feature values ​​and the historical external descriptive feature values.

[0135] Optionally, the device further includes a strategy table construction module, which is used to: sequentially obtain the current status description information from all status description information, and determine at least one alternative status information table that records the current status description information; obtain the alternative object information dedicated publishing and public platform used to publish each alternative status information table from each object information dedicated publishing and public platform, and obtain the data accuracy index of each alternative status information table published by each alternative object information dedicated publishing and public platform; obtain the target object information dedicated publishing and public platform and the target status information table corresponding to the maximum data accuracy index; add the mapping relationship between the current status description information, the target object information dedicated publishing and public platform and the target status information table to the data collection strategy table; and return to execute the operation of sequentially obtaining the current status description information from all business description information until the processing of all status description information is completed.

[0136] Optionally, the device also includes a historical feature value acquisition module, which is used to: acquire multiple historical time intervals before a preset time interval, and collect multiple historical state description information of the target object in each historical time interval on multiple dedicated public platforms for object information according to a pre-established data collection strategy table; and calculate multiple historical internal description feature values ​​of the target object in each historical time interval based on each historical state description information.

[0137] The risk prediction device provided in this embodiment of the invention can execute the risk prediction method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method execution. Content not described in detail in this embodiment can be referred to the description in any method embodiment of this application.

[0138] Example 4

[0139] Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement embodiments of the present invention is shown. For example... Figure 4As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0140] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0141] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as risk prediction methods.

[0142] In some embodiments, the risk prediction method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the risk prediction method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the risk prediction method by any other suitable means (e.g., by means of firmware).

[0143] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0144] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0145] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0146] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0147] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0148] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0149] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0150] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A risk prediction method, characterized in that, Executed by a server cluster, the method includes: When the risk warning conditions for the target object are met, according to the pre-established data collection strategy table, multiple current status descriptions of the target object within a preset time interval are collected on multiple dedicated public platforms for object information. Based on the current state description information, calculate multiple current internal description feature values ​​of the target object, and based on the multiple current internal description feature values ​​and the multiple historical internal description feature values ​​of the target object obtained from historical calculation, calculate multiple internal description change feature values ​​of the target object. Collect title data associated with the target object published within a preset time interval from multiple general information publishing websites, and calculate multiple external descriptive feature values ​​of the target object within the preset time interval based on each title data. Each internal description feature value and each external description feature value are input into a pre-built risk prediction model to obtain the risk prediction result of the target object. The different model layers in the risk prediction model are configured to be executed on different servers in the server cluster.

2. The method according to claim 1, characterized in that, Based on the data from each title, calculate multiple external descriptive feature values ​​of the target object within a preset time interval, including: Based on the preset dictionary, determine the target title data containing preset words in the title data associated with the target object, and the preset topic to which each preset word in the target title data belongs; The preset dictionary includes the correspondence between preset topics and preset vocabulary; Based on the preset words contained in each target title data, determine the number of preset words contained in all target title data, and based on the preset topic to which each preset word in each target title data belongs, determine the number of preset topics to which all target title data belongs. Based on the preset topic to which each preset word in each target title data belongs, determine the number of preset words belonging to different preset topics in each target title data, and perform a weighted calculation on the number of preset words belonging to different preset topics in each target title data to obtain the weighted word count corresponding to each target title data. Determine the weighted number of words corresponding to all target title data based on the weighted number of words corresponding to each target title data. The preset topics to which each target title data belongs are weighted and calculated to obtain the number of weighted topics corresponding to each target title data. Based on the number of weighted topics corresponding to each target title data, the number of weighted topics corresponding to all target title data is determined. The number of target title data, the number of preset words contained in all target title data, the number of preset topics to which all target title data belongs, the number of weighted words corresponding to all target title data, and the number of weighted topics corresponding to all target title data are used together as the external descriptive feature value of the target object within the preset time interval.

3. The method according to claim 2, characterized in that, The number of preset words belonging to different preset topics in each target title data is weighted and calculated to obtain the weighted word count corresponding to each target title data, including: Obtain the number of preset words corresponding to each preset theme from all target title data, and determine the weight coefficient corresponding to each preset theme based on the number of preset words corresponding to each preset theme and the total number of title data associated with the target object. Based on the weight coefficients corresponding to each preset theme, the number of preset words belonging to different preset themes in each target title data is weighted and calculated to obtain the weighted number of words corresponding to each target title data.

4. The method according to claim 3, characterized in that, The number of preset words belonging to different preset topics in each target title data is weighted and calculated to obtain the weighted word count corresponding to each target title data, including: Based on the arrangement position of each preset topic in the preset dictionary, determine the weight coefficient corresponding to each preset topic; Based on the weight coefficients corresponding to each preset theme, the number of preset words belonging to different preset themes in each target title data is weighted and calculated to obtain the weighted number of words corresponding to each target title data.

5. The method according to claim 1, characterized in that, Before inputting the internal descriptive change feature values ​​and the external descriptive feature values ​​into the pre-built risk prediction model to obtain the risk prediction results for the target object, the following steps are also included: Obtain multiple historical internal description feature values ​​and multiple historical external description feature values ​​corresponding to multiple sample objects; Five-fold cross-validation was used to train five risk testing models based on the historical internal description change feature values ​​and the historical external description feature values, and the recall rates corresponding to the five risk testing models were obtained respectively. The average recall rate is obtained by calculating the average of the recall rates mentioned above. The recall rate with the smallest deviation from the average recall rate and the largest actual value is taken as the target recall rate. Based on the hyperparameter combination corresponding to the target recall rate, a risk prediction model is constructed according to the feature values ​​of changes in each historical internal description and the feature values ​​of each historical external description.

6. The method according to any one of claims 1-5, characterized in that, Before collecting multiple current state descriptions of the target object within a preset time interval on multiple dedicated public platforms for publishing object information, based on a pre-established data collection strategy table, the process also includes: In all state description information, the current state description information is obtained sequentially, and at least one alternative state information table that records the current state description information is determined. On each dedicated information release platform for each object, obtain the dedicated information release platform for each candidate object used to release each candidate status information table, and obtain the data accuracy index of each candidate status information table released by each dedicated information release platform for each candidate object. A dedicated public platform and target status information table are used to obtain target object information corresponding to the maximum data accuracy index; Add the mapping relationship between the current status description information, the dedicated public platform for object information, and the target status information table to the data collection strategy table; Return to the previous step and sequentially retrieve the current status description information from all the business description information until the processing of all status description information is completed.

7. The method according to any one of claims 1-5, characterized in that, Before calculating the multiple internal description change feature values ​​of the target object based on multiple current internal description feature values ​​and multiple historical internal description feature values ​​of the target object obtained from historical calculations, the following steps are also included: Obtain multiple historical time intervals prior to the preset time interval, and collect multiple historical operational descriptions of the target object within each historical time interval on multiple dedicated public platforms for publishing object information, based on a pre-established data collection strategy table; Based on the historical operational description information, calculate multiple historical internal description feature values ​​of the target object within each historical time interval.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the risk prediction method according to any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the risk prediction method according to any one of claims 1-7.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the risk prediction method according to any one of claims 1-7.

Images