Financial Data Monitoring Method, Device, Equipment and Medium

Abstract

The present invention discloses a financial data monitoring method, device, equipment and medium: receiving a data monitoring instruction including a financial text; performing vector conversion on all financial data and all data types to obtain a fusion vector, and performing vector expansion processing on the fusion vector to obtain a type feature vector and an attribute feature vector; performing feature cross processing on the financial data to obtain at least one set of cross features, and determining a type cross vector and an attribute cross vector corresponding to each set of cross features according to the cross features, the type feature vector and the attribute feature vector; performing attention processing on the cross features based on all type cross vectors and all attribute cross vectors to obtain a feature cross relationship corresponding to the financial text; monitoring the financial text according to the feature cross relationship to obtain a data monitoring result. The present application monitors the financial text from two directions, enhances the connection between financial data and data types, and improves the accuracy and efficiency of financial data monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular, to a financial data monitoring method, device, equipment and medium. Background Art

[0002] With the continuous development of technology, the business of insurance companies has also achieved rapid development accordingly. More and more claims cases need to be reviewed to determine whether the claims cases meet the compensation conditions.

[0003] In the prior art, the information and content of claims cases are often reviewed manually. However, due to the large number and variety of claims cases involved, the efficiency of manual case review is low. And it is easy to make judgment mistakes during the process of manual case review, resulting in low accuracy of manual case review. Summary of the Invention

[0004] Embodiments of the present invention provide a financial data monitoring method, device, equipment and medium to solve the problems of low efficiency and low accuracy in reviewing claims cases by manual review in the prior art.

[0005] A financial data monitoring method includes:

[0006] Receiving a data monitoring instruction including a financial text, where the financial text includes at least one financial data, and one financial data corresponds to one data type;

[0007] Performing vector conversion on all the financial data and all the data types to obtain a fusion vector, and performing vector expansion processing on the fusion vector to obtain a type feature vector and an attribute feature vector;

[0008] Performing feature cross processing on the financial data to obtain at least one set of cross features, and determining a type cross vector and an attribute cross vector corresponding to each set of the cross features according to the cross features, the type feature vector and the attribute feature vector;

[0009] Performing attention processing on the cross features based on all the type cross vectors and all the attribute cross vectors to obtain a feature cross relationship formula corresponding to the financial text;

[0010] Monitoring the financial text according to the feature cross relationship formula to obtain a data monitoring result.

[0011] A financial data monitoring device includes:

[0012] A receiving module, configured to receive a data monitoring instruction including a financial text, where the financial text includes at least one financial data, and one financial data corresponds to one data type;

[0013] An expansion module for performing vector conversion on all the financial data and all the data types to obtain a fusion vector, and performing vector expansion processing on the fusion vector to obtain a type feature vector and an attribute feature vector;

[0014] A cross - feature module for performing feature cross - processing on the financial data to obtain at least one set of cross - features, and determining a type cross - vector and an attribute cross - vector corresponding to each set of the cross - features according to the cross - features, the type feature vector, and the attribute feature vector;

[0015] An attention processing module for performing attention processing on the cross - features based on all the type cross - vectors and all the attribute cross - vectors to obtain a feature cross - relation formula corresponding to the financial text;

[0016] A result module for monitoring the financial text according to the feature cross - relation formula to obtain a data monitoring result.

[0017] A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the above - mentioned financial data monitoring method is implemented.

[0018] A computer - readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the above - mentioned financial data monitoring method is implemented.

[0019] The present invention provides a financial data monitoring method, device, equipment, and medium. The method realizes monitoring of financial texts from two directions of type feature vectors and attribute feature vectors by expanding the fusion vector. By performing feature cross - processing on financial data, the cross - relation between financial data of different data types can be focused on. And by performing attention processing on the cross - features through type cross - vectors and attribute cross - vectors, the feature cross - relation in different cross - features can be learned. Thus, the connection between different financial data is enhanced. In this way, monitoring the financial text based on the feature cross - relation improves the accuracy and efficiency of financial data monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0021] Figure 1Schematic diagram of the application environment of the financial data monitoring method in an embodiment of the present invention;

[0022] Figure 2 Flowchart of the financial data monitoring method in an embodiment of the present invention;

[0023] Figure 3 Flowchart of step S40 in the financial data monitoring method in an embodiment of the present invention;

[0024] Figure 4 Flowchart of step S30 in the financial data monitoring method in an embodiment of the present invention;

[0025] Figure 5 Principle block diagram of the financial data monitoring device in an embodiment of the present invention;

[0026] Figure 6 Schematic diagram of a computer device in an embodiment of the present invention. Detailed implementation manners

[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0028] The financial data monitoring method provided by the embodiments of the present invention can be applied to the application environment as Figure 1 shown. Specifically, the financial data monitoring method is applied in a financial data monitoring device, and the financial data monitoring device includes a client and a server as Figure 1 shown. The client communicates with the server through a network, and is used to solve the problems of low efficiency and low accuracy of manual review in the prior art. Among them, the server can be an independent server or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms. The client, also known as the user side, refers to a program that provides local services for the client corresponding to the server. The client can be installed on but not limited to various personal computers, laptops, smartphones, tablets, and portable wearable devices.

[0029] In one embodiment, as Figure 2 shown, a financial data monitoring method is provided, and the method is applied in Figure 1Taking the server in [ID] as an example, the steps are as follows:

[0030] S10: Receive a data monitoring instruction including a financial text, where the financial text includes at least one financial data, and one financial data corresponds to one data type.

[0031] Understandably, the data monitoring instruction can be sent by the user through a mobile terminal, or automatically generated by the server according to the financial text after the user sends the financial text to the server through the mobile terminal. All financial data of the claim case are recorded in the financial text. The financial data may include the basic information of the claimant of the claim case, such as 20 years old, male, and teacher, etc. And each financial data corresponds to one data type. For example, 20 years old corresponds to the age type, male corresponds to the gender type, and teacher corresponds to the occupation type.

[0032] S20: Perform vector conversion on all the financial data and all the data types to obtain a fusion vector, and perform vector expansion processing on the fusion vector to obtain a type feature vector and an attribute feature vector.

[0033] Understandably, the fusion vector is a vector used to represent the financial text. The type feature vector is a vector of the data type obtained after expanding the fusion vector. The attribute feature vector is a vector of the financial data obtained after expanding the fusion vector.

[0034] Specifically, after obtaining all the financial data and all the data types, perform vector conversion on all the financial data and all the data types to obtain a fusion vector, that is, all the information of the financial text is included in this fusion vector. That is, use a preset language model (such as the Bert model) to extract text features from all the financial data and all the data types, perform vector conversion on the text features, and the obtained text feature vector is the fusion vector. Perform vector expansion processing on the fusion vector, that is, perform vector expansion on the fusion vector according to all the data types and all the financial data in the financial text, and perform vector conversion on all the data types and all the financial data to obtain the type feature vector of all the data types and the attribute feature vector of all the financial data.

[0035] S30: Perform feature cross processing on the financial data to obtain at least one set of cross features, and determine the type cross vector and the attribute cross vector corresponding to each set of the cross features according to the cross features, the type feature vector, and the attribute feature vector.

[0036] Understandably, the cross features are obtained by randomly combining financial data with financial data and data types with data types. For example, a set of cross features of financial data can be male - 25, 19 - student, female - customer service, etc., and a set of cross features of data types can be gender - age, age - occupation, gender - occupation, etc. The type cross vector is the type feature vector corresponding to the data type in the cross features. The attribute cross vector is the attribute feature vector corresponding to the financial data in the cross features.

[0037] Specifically, after obtaining the type feature vector and the attribute feature vector, feature crossing is performed on the financial data, that is, random combination is performed among the financial data, so as to obtain all the cross features corresponding to the financial data. While performing feature crossing among the financial data, the data types corresponding to the financial data also perform feature crossing, so as to obtain all the cross features corresponding to the data types. And the financial data and data types in the cross features are matched with the financial data and data types in the financial text. When the match is successful, that is, the financial data and data types in the cross features are the same as the financial data and data types in the financial text. The attribute feature vector corresponding to the financial data in the financial text and the type feature vector corresponding to the data type are determined as the attribute cross vector and the type cross vector corresponding to the cross features. In this way, the type cross vector and the attribute cross vector corresponding to each group of the cross features are determined through the above method.

[0038] S40: Perform attention processing on the cross features based on all the type cross vectors and all the attribute cross vectors to obtain the feature cross relationship formula corresponding to the financial text.

[0039] Understandably, the feature cross relationship formula is used to represent the sum of the products between all the cross features and all the attribute cross vectors. This feature cross relationship formula is obtained by performing attention processing on all the type cross vectors and all the attribute cross vectors through a preset attention model. The preset attention model is an attention network model based on a multi - layer perceptron, that is, a model with an additional layer of attention network in the multi - layer perceptron. The multi - layer perceptron (MLP, Multilayer Perceptron) is a feed - forward artificial neural network model, which maps multiple input data to a single output data.

[0040] Specifically, after obtaining the type cross-vector and the attribute cross-vector corresponding to each group of the cross-features, input the type cross-vector and the attribute cross-vector corresponding to each group of the cross-features into a preset attention model. Perform type attention processing on all the type cross-vectors through the preset attention model to obtain a type attention processing result. Perform attribute attention processing on all the attribute cross-vectors through the preset attention model and the type attention processing result to obtain an attribute attention processing result. Determine a feature cross-relation expression corresponding to the financial text according to all the attribute cross-vectors, all the attribute attention processing results, and all the cross-features.

[0041] S50: Monitor the financial text according to the feature cross-relation expression to obtain a data monitoring result.

[0042] Understandably, the data monitoring result is a result used to characterize whether the case can be normally compensated. Further, the relation expression is a factorization machine algorithm, and the expression of the factorization machine algorithm is where x represents the fusion vector of the entire financial text, x i , x j represents the i-th financial data and the j-th financial data in the financial text, w i represents the weight of the cross-feature x i x j , w0 is a bias, v i represents the attribute cross-vector corresponding to the financial data x j , v j represents the attribute cross-vector corresponding to the financial data x j . The vector relation expression is used to determine the sum of the products of all the cross-features and the attribute cross-vectors corresponding to all the cross-features, and the expression of the vector relation expression is The expression of the feature cross-relation expression is The expression of the target relation expression is where x i , x j represents the i-th financial data and the j-th financial data in the financial text, (vf i ⊙ vf j ) is the product of the attribute cross-vectors corresponding to the cross-feature x i x j , vf i is the attribute cross-vector corresponding to x i , vf j is the attribute cross-vector corresponding to x j , and is the attribute attention processing result.

[0043] Specifically, after obtaining the feature cross relationship expression corresponding to the financial text, obtain the relationship expression corresponding to the financial text from the server, and replace the vector relationship expression in the relationship expression with the feature cross relationship expression to obtain the target relationship expression. Input the financial text into the target relationship expression, and monitor the financial text through the target relationship expression, that is, calculate the financial data in the financial text through the target relationship expression, so as to obtain the data monitoring result.

[0044] In the data monitoring method in the embodiment of the present invention, the method realizes the monitoring of the financial text from two directions of the type feature vector and the attribute feature vector by expanding the fusion vector. By performing feature cross on the financial data, the cross relationship between financial data of different data types can be focused on. And through the type cross vector and the attribute cross vector to perform attention processing on the cross feature, the feature cross relationship in different cross features can be learned. Thus, the connection between different financial data is enhanced. In this way, monitoring the financial text based on the feature cross relationship improves the accuracy and efficiency of financial data monitoring.

[0045] In one embodiment, as Figure 3 shown, in step S40, that is, based on all the type cross vectors and all the attribute cross vectors to perform attention processing on the cross feature to obtain the feature cross relationship expression corresponding to the financial text, including:

[0046] S401, perform type attention processing on the cross feature through the type cross vector to determine the type attention value.

[0047] It can be understood that the type attention value is obtained by performing attention processing on the type cross vector through a preset attention model.

[0048] Specifically, after determining the type cross vector and the attribute cross vector corresponding to each group of the cross features, perform matrix conversion on all the type cross vectors corresponding to this group of cross features, that is, convert the high-dimensional sparse type cross vector into a low-dimensional dense type matrix vector. And input the type matrix vector into the preset attention model, and perform type attention processing on the type matrix vector through the preset attention model, so as to determine the type attention value corresponding to this group of cross features. In this way, the type attention values corresponding to each group of cross features are obtained in turn through the above method.

[0049] S402, perform attribute attention processing on the cross feature through the attribute cross vector and the type attention value to determine the attribute attention value.

[0050] S403, based on all the attribute attention values, all the cross features and all the attribute cross vectors, determine the feature cross relationship expression corresponding to the financial text.

[0051] Understandably, the attribute attention value is obtained by performing attention processing on the attribute cross vector through a preset attention model.

[0052] Specifically, after obtaining the type attention value corresponding to each group of cross features, matrix transformation is performed on all the attribute cross vectors corresponding to this group of cross features, that is, converting the high-dimensional sparse attribute cross vector into a low-dimensional dense attribute matrix vector. And the attribute matrix vector and the type attention value corresponding to the same cross feature are input into the preset attention model, and the preset attention model performs attribute attention processing on the attribute matrix vector and the type attention value, so as to determine the attribute attention value corresponding to this group of cross features. In this way, the attribute attention value corresponding to each group of cross features is obtained in turn through the above method.

[0053] Furthermore, all the attribute cross vectors, attribute attention values, and this cross feature corresponding to the same cross feature are calculated, that is, all the attribute cross vectors corresponding to the same cross feature are multiplied pairwise, and the pairwise multiplication result, the attribute attention value, and this cross feature are multiplied, so as to obtain the relational expression corresponding to this cross feature. The expression of the relational expression is In this way, the relational expressions corresponding to each group of cross features are obtained in turn through the above method, and the relational expressions corresponding to each group of cross features are added together, and the feature cross relational expression corresponding to the financial text can be obtained, that is

[0054] In the embodiment of the present invention, type attention processing is performed on the cross feature through the type cross vector, so as to realize the determination of the type attention value, which facilitates the subsequent attention processing of the attribute cross vector. Based on the type attention value, attribute attention processing is performed on the cross feature through the attribute cross vector, so as to determine the attribute attention value, which enhances the connection between financial data. According to all the attribute attention values, all the cross features, and all the attribute cross vectors, the feature cross relational expression is determined, thereby improving the accuracy of monitoring financial texts.

[0055] In one embodiment, in step S401, that is, type attention processing is performed on the cross feature through the type cross vector to determine the type attention value, including:

[0056] S4011: Determine the first type vector corresponding to the first financial data and the second type vector corresponding to the second financial data from the type cross vectors corresponding to the cross feature.

[0057] Understandably, the cross feature includes the first financial data and the second financial data. The first type vector is the type cross vector corresponding to the first financial data. The second type vector is the type cross vector corresponding to the second financial data.

[0058] Specifically, after determining the type cross-vector and the attribute cross-vector corresponding to each group of the cross-features, determine the first financial data and the second financial data included in each group of cross-features, and determine the first type vector corresponding to the first financial data and the second type vector corresponding to the second financial data from the type cross-vector corresponding to the cross-features. In this way, the first type vector corresponding to the first financial data and the second type vector corresponding to the second financial data in each group of cross-features are determined in sequence through the above method.

[0059] Exemplarily, when the group of cross-features is age-gender, determine the age cross-vector corresponding to age from the age cross-vector and the gender cross-vector corresponding to age-gender, and determine it as the first type vector, and determine the gender cross-vector corresponding to gender as the second type vector. When the group of cross-features is gender-occupation, determine the gender cross-vector corresponding to gender from the gender cross-vector and the occupation cross-vector corresponding to gender-occupation, and determine it as the first type vector, and determine the occupation cross-vector corresponding to occupation as the second type vector.

[0060] S4012: Determine the first hidden layer vector according to the first type vector, the second type vector, and the cross-feature.

[0061] It can be understood that the first hidden layer vector is a low-dimensional dense vector matrix. The type attention value is a weight value used to characterize the data type.

[0062] Specifically, after obtaining the first type vector corresponding to the first financial data and the second type vector corresponding to the second financial data, first perform a pairwise multiplication process on the first type vector and the second type vector to obtain a pairwise multiplication result of types. Then multiply the pairwise multiplication result of types by the cross-feature to obtain the first hidden layer vector corresponding to the cross-feature. In this way, the first hidden layer vectors corresponding to each group of cross-features are obtained in sequence through the above method.

[0063] S4013: Perform type attention processing on all the first hidden layer vectors to obtain the type attention value.

[0064] It can be understood that the type attention value is a weight value used to characterize the data type.

[0065] Specifically, input the first hidden layer vectors corresponding to all cross features into a preset attention model. Through the preset attention model, perform type attention processing on all the first hidden layer vectors, that is, only perform attention processing on the first hidden layer vectors corresponding to the data type. Process the first hidden layer vectors through the first hidden layer in the preset attention model. The hidden units in the first hidden layer calculate all the first hidden layer vectors according to the weights and input the calculation results into the second hidden layer. The hidden units in the second hidden layer calculate the calculation results of all the first hidden layer vectors again according to weights different from those of the first hidden layer and input the calculation results of the second hidden layer into the attention network layer. Through the attention network layer, perform attention processing on the calculation results of all the second hidden layers, that is, predict important calculation results and strengthen important calculation results, and then the type attention processing results can be obtained. Then, through the pooling layer, convert the type attention processing results with variable length into type attention processing results with fixed length. And perform activation processing on the preset attention model through the ReLU function, and then perform normalization processing on the pooled type attention processing results through the normalization layer to obtain the type attention value. Among them, the number of hidden layers and the hidden units (neurons) in the hidden layer can be set according to the actual situation. For example, in this example, it is set to 2 layers and 16 hidden units (neurons).

[0066] In the embodiment of the present invention, the cross features are calculated through the first type vector corresponding to the first financial data and the second type vector corresponding to the second financial data, thereby realizing the acquisition of the first hidden layer vectors, and further realizing the conversion of the high-dimensional sparse type cross vectors into low-dimensional dense type matrix vectors. Through the preset attention model, perform attention processing on the first hidden layer vectors, realize the determination of the type attention value of the data type, and further facilitate the subsequent processing of the financial data.

[0067] In one embodiment, in step S402, that is, perform attribute attention processing on the cross features through the attribute cross vector and the type attention value to determine the attribute attention value, including:

[0068] S4021, determine the first attribute vector corresponding to the first financial data and the second attribute vector corresponding to the second financial data from the attribute cross vectors corresponding to the cross features.

[0069] It can be understood that the first attribute vector is the attribute cross vector corresponding to the first financial data. The second attribute vector is the attribute cross vector corresponding to the second financial data.

[0070] Specifically, after determining the type cross-vector and the attribute cross-vector corresponding to each group of the cross-features, the first financial data and the second financial data included in each group of cross-features are determined, and the first attribute vector corresponding to the first financial data and the second attribute vector corresponding to the second financial data are determined from the attribute cross-vector corresponding to the cross-features. In this way, the first attribute vector corresponding to the first financial data and the second attribute vector corresponding to the second financial data in each group of cross-features are determined in sequence through the above method.

[0071] Exemplarily, when the group of cross-features is 20 - male, the attribute cross-vector corresponding to 20 is determined from the attribute cross-vector of 20 and the attribute cross-vector of male corresponding to 20 - male, and is determined as the first attribute vector, and the attribute cross-vector of male is determined as the second attribute vector. When the group of cross-features is female - customer service, the attribute cross-vector corresponding to female is determined from the female cross-vector and the customer service cross-vector corresponding to female - customer service, and is determined as the first attribute vector, and the attribute cross-vector of customer service is determined as the second attribute vector.

[0072] S4022. Determine the second hidden layer vector according to the type attention value, the first attribute vector, and the second attribute vector.

[0073] It can be understood that the second hidden layer vector is a low-dimensional dense vector matrix. The attribute attention value is a weight value used to characterize the financial data.

[0074] Specifically, after obtaining the first attribute vector corresponding to the first financial data and the second attribute vector corresponding to the second financial data, the first attribute vector and the second attribute vector are first subjected to bitwise multiplication processing to obtain the result of bitwise multiplication of attributes. Then, the result of bitwise multiplication of attributes, the cross-feature, and the type attention value are multiplied to obtain the second hidden layer vector corresponding to the cross-feature. In this way, the second hidden layer vectors corresponding to each group of cross-features are obtained in sequence through the above method.

[0075] S4023. Perform attribute attention processing on all the second hidden layer vectors to obtain the attribute attention value.

[0076] It can be understood that the attribute attention value is a weight value used to characterize the financial data.

[0077] Specifically, input the second hidden layer vectors corresponding to all cross features into a preset attention model. Through the preset attention model, perform type attention processing on all second hidden layer vectors, that is, only perform attention processing on the second hidden layer vectors corresponding to financial data. Process the second hidden layer vectors through the first hidden layer in the preset attention model. The hidden units in the first hidden layer calculate all second hidden layer vectors according to weights and input the calculation results into the second hidden layer. The hidden units in the second hidden layer calculate the calculation results of all second hidden layer vectors again according to weights different from those in the first hidden layer and input the calculation results of the second hidden layer into the attention network layer. Through the attention network layer, perform attention processing on the calculation results of all second hidden layers, that is, predict important calculation results and strengthen important calculation results, and then the attribute attention processing results can be obtained. Then, through the pooling layer, convert the attribute attention processing results with variable length into attribute attention processing results with fixed length. And perform activation processing on the preset attention model through the ReLU function, and then perform normalization processing on the pooled attribute attention processing results through the normalization layer, and the attribute attention values can be obtained. Among them, the number of hidden layers and the hidden units (neurons) in the hidden layer can be set according to the actual situation, such as setting to 2 layers and 16 neurons.

[0078] In the embodiment of the present invention, through the first attribute vector corresponding to the first financial data, the second attribute vector corresponding to the second financial data, and the cross features, the acquisition of the second hidden layer vectors is realized, and further the conversion of the high-dimensional sparse attribute cross vectors into low-dimensional dense attribute matrix vectors is realized. Through the preset attention model, perform attention processing on the second hidden layer vectors, realize the determination of the attribute attention values of the financial data, and further improve the accuracy of case prediction and the efficiency of case prediction.

[0079] In one embodiment, as Figure 4 shown, in step S30, that is, according to the cross features, the type feature vectors, and the attribute feature vectors, determining the type cross vectors and attribute cross vectors corresponding to each group of the cross features includes:

[0080] S301, obtain the data types corresponding to the first financial data and the data types corresponding to the second financial data.

[0081] S302, according to the data types corresponding to the first financial data and the data types corresponding to the second financial data, determine the type feature vectors corresponding to the data types of the first financial data and the type feature vectors corresponding to the data types of the second financial data.

[0082] It can be understood that the cross features include the first financial data and the second financial data.

[0083] Specifically, after obtaining the type feature vector and the attribute feature vector, the cross features are parsed to obtain the first financial data and the second financial data. The first financial data and the second financial data included in the cross features are matched with the financial data in the financial text. When the match is successful, that is, when the first financial data and the second financial data in the cross features are the same as the financial data in the financial text, the type feature vector corresponding to the data type of the financial data in the financial text is determined as the type feature vector corresponding to the first financial data and the type feature vector corresponding to the second financial data.

[0084] S303. Determine the attribute feature vector corresponding to the first financial data and the attribute feature vector corresponding to the second financial data according to the first financial data and the second financial data.

[0085] S304. Based on the type feature vector corresponding to the data type of the first financial data, the type feature vector corresponding to the data type of the second financial data, the attribute feature vector corresponding to the first financial data, and the attribute feature vector corresponding to the second financial data, determine the type cross vector and the attribute cross vector corresponding to each group of cross features.

[0086] Specifically, the first financial data and the second financial data included in the cross features are matched with the financial data in the financial text. When the match is successful, that is, when the first financial data and the second financial data in the cross features are the same as the financial data in the financial text, the attribute feature vector corresponding to the financial data in the financial text is determined as the attribute feature vector corresponding to the first financial data and the attribute feature vector corresponding to the second financial data.

[0087] Further, according to the first financial data and the second financial data corresponding to the cross feature, the type feature vector corresponding to the data type of the first financial data and the type feature vector corresponding to the data type of the second financial data are determined as the type cross vector corresponding to each group of cross features. The attribute feature vector corresponding to the first financial data and the attribute feature vector corresponding to the second financial data are determined as the attribute cross vector corresponding to each group of cross features, and thus the type cross vector and the attribute cross vector corresponding to each group of cross features can be determined. In this way, the type cross vector and the attribute cross vector corresponding to all cross features are determined by the above method.

[0088] In an embodiment of the present invention, based on the data types corresponding to the first financial data and the second financial data, the type feature vectors corresponding to the first financial data and the second financial data are quickly determined. Based on the first financial data and the second financial data, the attribute feature vectors corresponding to the first financial data and the second financial data are quickly determined, and then the type cross vectors and attribute cross vectors corresponding to each group of cross features are determined, and the relationship between different financial data is enhanced.

[0089] In one embodiment, in step S50, that is, monitoring the financial text according to the feature cross relationship expression to obtain a data monitoring result, including:

[0090] S501, obtaining the relationship expression corresponding to the financial text, where the relationship expression includes a fixed relationship expression and a vector relationship expression.

[0091] S502, replacing the vector relationship expression with the feature cross relationship expression to obtain a target relationship expression, and monitoring the financial data according to the target relationship expression to obtain the data monitoring result.

[0092] It can be understood that the relationship expression is the original algorithm for monitoring financial texts, that is, the factorization machine algorithm. The fixed relationship expression is a linear regression equation. The linear regression equation is a statistical analysis method that uses regression analysis in mathematical statistics to determine the quantitative relationship of interdependence between two or more variables. The expression is w0 is the bias number, w i is the weight value of each feature x i and x i is the i-th feature. The target relationship expression is an algorithm improved based on the relationship expression. Among them, the vector cross relationship expression cannot handle the mutual relationship between the data type and the financial data, so the vector cross relationship expression is improved. That is, the feature cross relationship expression, which can better express the relationship between the data type and the financial data.

[0093] Specifically, after obtaining the feature cross relationship formula corresponding to the financial text, retrieve the stored relationship formula from the server. This relationship formula is the factorization machine algorithm, which includes a fixed relationship formula and a vector relationship formula. Replace the vector cross relationship formula in the relationship formula with the feature cross relationship formula, and thus the target relationship formula can be obtained. Then, calculate the financial data in the financial text through the target relationship formula. First, perform vector conversion on all financial data and all data types to obtain attribute feature vectors and type feature vectors. Then, perform feature crossing on the financial data to obtain cross features, and determine the attribute cross vectors and type cross vectors corresponding to the cross features for the attribute feature vectors and type feature vectors corresponding to the financial data and data types. Then, perform attention calculation on all attribute cross vectors and all type cross vectors to determine the data monitoring result corresponding to the financial text.

[0094] In the embodiment of the present invention, by replacing the vector relationship formula with the feature cross relationship formula, the determination of the target relationship formula is realized. Based on the target relationship formula, all financial data is calculated, thereby realizing the monitoring of the financial text, improving the accuracy of the monitoring of the financial text, and enhancing the efficiency of the monitoring of the financial text.

[0095] In one embodiment, in step S502, that is, monitoring the financial data according to the target relationship formula to obtain a data monitoring result, including:

[0096] S5021, generate a text monitoring script based on the target relationship formula and execute the text monitoring script to obtain the monitoring value corresponding to the financial text.

[0097] It can be understood that the monitoring value is obtained by monitoring the financial text through the target relationship formula and is used to represent the numerical value of the monitoring of the financial text. The text monitoring script is generated by calculating the financial data through the target relationship formula.

[0098] Specifically, after obtaining the target relationship formula, generate a text monitoring script for monitoring the financial text according to the target table relationship formula. Executing the text monitoring script, that is, calculating the financial data in the financial text through the target relationship formula, thereby obtaining the monitoring value corresponding to the financial text. That is, the monitoring value is the numerical value obtained by calculating the cross features of all financial data in the financial text through the target relationship formula.

[0099] S5022, obtain a preset threshold, and determine the data monitoring result according to the monitoring value and the preset threshold.

[0100] It can be understood that the preset threshold is set in advance to determine whether the case corresponding to the financial text can be normally compensated.

[0101] Specifically, after obtaining the monitoring value corresponding to the financial text, retrieve the preset threshold from the server or obtain the preset threshold from a third-party platform, and compare the monitoring value with the preset threshold. When the monitoring value corresponding to the data monitoring result is greater than or equal to the preset threshold, determine the monitoring value corresponding to the greater than or equal to the preset threshold as the claim data monitoring result, that is, the claim data monitoring result is used to indicate that the case can be normally compensated. When the monitoring value corresponding to the data monitoring result is less than the preset threshold, record the monitoring value corresponding to the less than the preset threshold as the non-claim data monitoring result, that is, the non-claim data monitoring result indicates that the case cannot be normally compensated.

[0102] In the embodiment of the present invention, a text monitoring script is generated through a target relational expression, thereby realizing the acquisition of the text monitoring script, and further realizing the determination of the monitoring value. The monitoring value is compared with the preset threshold, so as to obtain the data monitoring result of the financial text monitoring, and further improve the accuracy of the financial text monitoring.

[0103] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0104] In an embodiment, a financial data monitoring device is provided, and the financial data monitoring device corresponds one-to-one to the financial data monitoring method in the above embodiment. As Figure 5 shown, the financial data monitoring device includes a receiving module 11, an expansion module 12, a cross-feature module 13, an attention processing module 14, and a result module 15. The detailed descriptions of each functional module are as follows:

[0105] The receiving module 11 is configured to receive a data monitoring instruction including a financial text, where the financial text includes at least one financial data, and one financial data corresponds to one data type;

[0106] The expansion module 12 is configured to perform vector conversion on all the financial data and all the data types to obtain a fusion vector, and perform vector expansion processing on the fusion vector to obtain a type feature vector and an attribute feature vector;

[0107] The cross-feature module 13 is configured to perform feature cross-processing on the financial data to obtain at least one set of cross-features, and determine a type cross-vector and an attribute cross-vector corresponding to each set of cross-features according to the cross-features, the type feature vector, and the attribute feature vector;

[0108] The attention processing module 14 is configured to perform attention processing on the cross-features based on all the type cross-vectors and all the attribute cross-vectors to obtain a feature cross-relational expression corresponding to the financial text;

[0109] A result module 15, configured to monitor the financial text according to the feature cross relationship formula to obtain a data monitoring result.

[0110] In one embodiment, the attention processing module 14 includes:

[0111] A type attention unit, configured to perform type attention processing on the cross features through the type cross vector to determine a type attention value;

[0112] An attribute attention unit, configured to perform attribute attention processing on the cross features through the attribute cross vector and the type attention value to determine an attribute attention value;

[0113] A relationship determination unit, configured to determine a feature cross relationship formula corresponding to the financial text based on all the attribute attention values, all the cross features, and all the attribute cross vectors.

[0114] In one embodiment, the type attention unit includes:

[0115] A type vector determination subunit, configured to determine a first type vector corresponding to the first financial data and a second type vector corresponding to the second financial data from the type cross vectors corresponding to the cross features;

[0116] A first hidden layer vector subunit, configured to determine a first hidden layer vector according to the first type vector, the second type vector, and the cross features;

[0117] A type attention value subunit, configured to perform type attention processing on all the first hidden layer vectors to obtain the type attention value.

[0118] In one embodiment, the attribute attention unit includes:

[0119] An attribute vector determination subunit, configured to determine a first attribute vector corresponding to the first financial data and a second attribute vector corresponding to the second financial data from the attribute cross vectors corresponding to the cross features;

[0120] A second hidden layer vector subunit, configured to determine a second hidden layer vector according to the type attention value, the first attribute vector, and the second attribute vector;

[0121] An attribute attention value subunit, configured to perform attribute attention processing on all the second hidden layer vectors to obtain the attribute attention value.

[0122] In one embodiment, the cross feature module 13 includes:

[0123] A data type acquisition unit for acquiring the data type corresponding to the first financial data and the data type corresponding to the second financial data;

[0124] A type feature vector unit for determining the type feature vector corresponding to the data type of the first financial data and the type feature vector corresponding to the data type of the second financial data according to the data type corresponding to the first financial data and the data type corresponding to the second financial data;

[0125] An attribute feature vector unit for determining the attribute feature vector corresponding to the first financial data and the attribute feature vector corresponding to the second financial data according to the first financial data and the second financial data;

[0126] A cross vector determination unit for determining the type cross vector and the attribute cross vector corresponding to each group of cross features based on the type feature vector corresponding to the data type of the first financial data, the type feature vector corresponding to the data type of the second financial data, the attribute feature vector corresponding to the first financial data, and the attribute feature vector corresponding to the second financial data.

[0127] In one embodiment, the result module 15 includes:

[0128] An acquisition unit for acquiring the relational expression corresponding to the financial text, where the relational expression includes a fixed relational expression and a vector relational expression;

[0129] A monitoring result unit for replacing the vector relational expression with the feature cross relational expression to obtain a target relational expression, and monitoring the financial data according to the target relational expression to obtain the data monitoring result.

[0130] In one embodiment, the monitoring result unit includes:

[0131] An execution subunit for generating a text monitoring script based on the target relational expression and executing the text monitoring script to obtain the monitoring value corresponding to the financial text;

[0132] A comparison subunit for obtaining a preset threshold and determining the data monitoring result according to the monitoring value and the preset threshold.

[0133] For the specific limitations of the financial data monitoring device, reference may be made to the limitations of the financial data monitoring method in the foregoing text, which will not be elaborated herein. Each module in the above financial data monitoring device can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in the processor of the computer device in hardware form or independent thereof, or stored in the memory of the computer device in software form, so as to facilitate the processor to call and execute the operations corresponding to each of the above modules.

[0134] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as Figure 6 shown. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the data used in the financial data monitoring method in the above embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a financial data monitoring method.

[0135] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the financial data monitoring method in the above embodiment.

[0136] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, it implements the financial data monitoring method in the above embodiment.

[0137] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0138] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0139] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A financial data monitoring method, characterized in that, Including: Receiving a data monitoring instruction including a financial text, where the financial text includes at least one financial data, and one financial data corresponds to one data type; Performing vector conversion on all the financial data and all the data types to obtain a fusion vector, and performing vector extension processing on the fusion vector to obtain a type feature vector and an attribute feature vector; Performing feature cross-processing on the financial data to obtain at least one set of cross features, and determining a type cross vector and an attribute cross vector corresponding to each set of the cross features according to the cross features, the type feature vector, and the attribute feature vector; Performing attention processing on the cross features based on all the type cross vectors and all the attribute cross vectors to obtain a feature cross relationship expression corresponding to the financial text; Monitoring the financial text according to the feature cross relationship expression to obtain a data monitoring result; The cross features include a first financial data and a second financial data; The determining a type cross vector and an attribute cross vector corresponding to each set of the cross features according to the cross features, the type feature vector, and the attribute feature vector includes: Obtaining the data type corresponding to the first financial data and the data type corresponding to the second financial data; Determining the type feature vector corresponding to the data type of the first financial data and the type feature vector corresponding to the data type of the second financial data according to the data type corresponding to the first financial data and the data type corresponding to the second financial data; Determining the attribute feature vector corresponding to the first financial data and the attribute feature vector corresponding to the second financial data according to the first financial data and the second financial data; Determining the type cross vector and the attribute cross vector corresponding to each set of cross features based on the type feature vector corresponding to the data type of the first financial data, the type feature vector corresponding to the data type of the second financial data, the attribute feature vector corresponding to the first financial data, and the attribute feature vector corresponding to the second financial data; The monitoring the financial text according to the feature cross relationship expression to obtain a data monitoring result includes: Obtaining a relationship expression corresponding to the financial text, where the relationship expression includes a fixed relationship expression and a vector relationship expression; Replacing the vector relationship expression with the feature cross relationship expression to obtain a target relationship expression, and monitoring the financial data according to the target relationship expression to obtain the data monitoring result; The monitoring the financial data according to the target relationship expression to obtain a data monitoring result includes: Generating a text monitoring script based on the target relationship expression and executing the text monitoring script to obtain a monitoring value corresponding to the financial text; Obtaining a preset threshold, and determining the data monitoring result according to the monitoring value and the preset threshold.

2. The financial data monitoring method according to claim 1, wherein The performing attention processing on the cross features based on all the type cross vectors and all the attribute cross vectors to obtain a feature cross relationship expression corresponding to the financial text includes: Perform type attention processing on the cross features through the type cross vectors to determine type attention values; Perform attribute attention processing on the cross features through the attribute cross vectors and the type attention values to determine attribute attention values; Based on all the attribute attention values, all the cross features, and all the attribute cross vectors, determine the feature cross relationship corresponding to the financial text.

3. The financial data monitoring method according to claim 2, characterized in that, The cross features include first financial data and second financial data; The performing attention processing on the cross features through the type cross vectors to determine type attention values includes: Determine a first type vector corresponding to the first financial data and a second type vector corresponding to the second financial data from the type cross vectors corresponding to the cross features; Determine a first hidden layer vector according to the first type vector, the second type vector, and the cross features; Perform type attention processing on all the first hidden layer vectors to obtain the type attention values.

4. The financial data monitoring method according to claim 3, characterized in that, The performing attention processing on the cross features through the attribute cross vectors and all the type attention values to determine attribute attention values includes: Determine a first attribute vector corresponding to the first financial data and a second attribute vector corresponding to the second financial data from the attribute cross vectors corresponding to the cross features; Determine a second hidden layer vector according to the type attention values, the first attribute vector, and the second attribute vector; Perform attribute attention processing on all the second hidden layer vectors to obtain the attribute attention values.

5. A financial data monitoring device, characterized in that, Includes: A receiving module, configured to receive a data monitoring instruction including a financial text, where the financial text includes at least one financial data, and one financial data corresponds to one data type; An expansion module, configured to perform vector conversion on all the financial data and all the data types to obtain fusion vectors, and perform vector expansion processing on the fusion vectors to obtain type feature vectors and attribute feature vectors; A cross feature module, configured to perform feature cross processing on the financial data to obtain at least one group of cross features, and determine type cross vectors and attribute cross vectors corresponding to each group of the cross features according to the cross features, the type feature vectors, and the attribute feature vectors; An attention processing module, configured to perform attention processing on the cross features based on all the type cross vectors and all the attribute cross vectors to obtain the feature cross relationship corresponding to the financial text; A result module, configured to monitor the financial text according to the feature cross relationship to obtain a data monitoring result; The cross features include first financial data and second financial data; the cross feature module includes: A data type acquisition unit, configured to acquire the data type corresponding to the first financial data and the data type corresponding to the second financial data; A type feature vector unit, configured to determine the type feature vector corresponding to the data type of the first financial data and the type feature vector corresponding to the data type of the second financial data according to the data type corresponding to the first financial data and the data type corresponding to the second financial data; An attribute feature vector unit, configured to determine the attribute feature vector corresponding to the first financial data and the attribute feature vector corresponding to the second financial data according to the first financial data and the second financial data; A cross vector determination unit, configured to determine the type cross vector and the attribute cross vector corresponding to each group of cross features based on the type feature vector corresponding to the data type of the first financial data, the type feature vector corresponding to the data type of the second financial data, the attribute feature vector corresponding to the first financial data, and the attribute feature vector corresponding to the second financial data; The result module includes: An acquisition unit, configured to acquire a relationship expression corresponding to a financial text, where the relationship expression includes a fixed relationship expression and a vector relationship expression; A monitoring result unit, configured to replace the vector relationship expression with the feature cross relationship expression to obtain a target relationship expression, and monitor the financial data according to the target relationship expression to obtain the data monitoring result; The monitoring result unit includes: An execution subunit, configured to generate a text monitoring script based on the target relationship expression and execute the text monitoring script to obtain a monitoring value corresponding to the financial text; A comparison subunit, configured to acquire a preset threshold, and determine the data monitoring result according to the monitoring value and the preset threshold.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, the financial data monitoring method according to any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, the financial data monitoring method according to any one of claims 1 to 4 is implemented.

Images