Method and system for information prediction and analysis of multi-dimensional content

Abstract

The application provides a multi-dimensional content information prediction and analysis method and system, and belongs to the technical field of information prediction and analysis. The method comprises the following steps: S110, determining an information prediction dimension value PreW; S120, acquiring PreW financial information sequences generated at PreW time nodes; S130, constructing a financial information prediction matrix PreMatrix; S140, judging whether PreW is greater than K, if yes, determining a corresponding information prediction dimension attribute value based on PreW, and selecting a PreW-order sub-matrix from the PreMatrix as a prediction data set; and S150, performing financial information prediction and analysis based on the prediction data set. The system comprises a prediction model database, wherein the prediction model database stores at least one prediction model, and is used for realizing the method. The application can realize financial information simulation and prediction analysis under different dimensions and different prediction targets.

Description

Technical Field

[0001] This invention belongs to the field of information prediction and analysis technology, and in particular relates to a method and system for information prediction and analysis with multi-dimensional content. Background Technology

[0002] Information analysis and prediction is a broad and in-depth information processing activity, focusing primarily on two aspects: first, the in-depth analysis of known information; and second, the scientific prediction of unknown or future information based on this analysis. Specifically, it involves: systematically and extensively collecting known information related to a given topic to meet specific user information needs; processing, organizing, evaluating, and analyzing this information to systematize and organize its content, revealing the laws governing the movement of objective things; and, based on this, using scientific theories, methods, and technologies to make reasonable predictions about unknown or future information about objective things; finally, delivering the prediction results to users through appropriate channels in the form of an information product to meet their needs.

[0003] Any information prediction and analysis requires a foundation of existing, objective data and the establishment of corresponding predictive models to achieve its purpose. However, in many cases, due to inadequate basic datasets or slower-than-expected development of predictive models, coupled with users' objective predictive expectations, the information prediction and analysis process cannot proceed in a timely manner. This situation is particularly evident in the financial information field. The diversity, variability, and differences in the dimensions of financial information, along with the rapid pace of updates to financial products, often lead to delays in the development of predictive models, failing to fully meet current needs for financial information prediction and analysis. Furthermore, the mismatch between different predictive dimensions required by different users and the inability of a single predictive model to match existing predictive data has become a pressing conflict that needs to be resolved. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a multi-dimensional information prediction and analysis method and system, which can realize the simulation and predictive analysis of financial information under different dimensions and prediction objectives.

[0005] In a first aspect of the present invention, a method for predicting and analyzing financial information applied to multi-source heterogeneous terminals is proposed, the method comprising the following steps:

[0006] S110: Determine the information prediction dimension value PreW, wherein the information prediction dimension value PreW is a positive integer;

[0007] S120: Obtain the PreW financial information sequences generated at PreW time points. Every piece of financial information Contains K dimension attribute values , K is a positive integer;

[0008] S130: Constructing the Financial Information Pre-Matrix

[0009] ;

[0010] in,

[0011] S140: Determine whether the information prediction dimension value PreW is not greater than K. If so, determine the corresponding information prediction dimension attribute value based on the information prediction dimension value PreW, and select the PreW order submatrix as the prediction dataset from the financial information prediction matrix PreMatrix.

[0012] S150: Based on the predicted dataset, perform financial information prediction and analysis.

[0013] In practical applications, the K dimension attribute values Includes at least two of the following attribute values:

[0014] The time when financial information is generated, the location where financial data is generated, the type of terminal that generates financial data, the payment amount included in the financial data, the environment in which the financial data is generated, the transaction object corresponding to the financial data, the gender of the user of the terminal that generates the financial data, and the age group of the user of the terminal that generates the financial data.

[0015] As mentioned above, in step S140, in one case, if the information prediction dimension value PreW is not greater than K, then the corresponding information prediction dimension attribute value is determined based on the information prediction dimension value PreW, and the PreW order submatrix is ​​selected from the financial information prediction matrix PreMatrix as the prediction dataset.

[0016] Correspondingly, in step S140, in another case, if the information prediction dimension value PreW is greater than K, then data simulation is performed on the financial information prediction matrix PreMatrix to obtain each piece of financial information. The simulated dimension attribute value;

[0017] Each financial information The simulated dimension attribute values ​​are added to the financial information prediction matrix PreMatrix to form the financial information simulation matrix PreMatrixM;

[0018] The PreW-order submatrix is ​​selected from the financial information simulation matrix PreMatrixM as the prediction dataset.

[0019] More specifically, step S150 employs a prediction model based on a deep neural network to perform the financial information prediction;

[0020] The information prediction dimension value PreW in step S110 is determined by the number of nodes in the output layer of the prediction model based on the deep neural network.

[0021] Step S150 employs multiple different prediction models to perform the financial information prediction;

[0022] Different prediction models correspond to different information prediction dimension values, PreW.

[0023] In a second aspect of the invention, a multi-dimensional content information prediction and analysis system is also proposed, the system including a prediction model database, the prediction model database storing at least one prediction model;

[0024] In terms of specific structure, the system also includes:

[0025] A prediction parameter input unit is used to input prediction parameters, which include information prediction dimension values ​​or information prediction dimension attribute values.

[0026] The model automatic selection unit loads at least one target prediction model from the prediction model database based on the prediction parameters input by the prediction parameter input unit.

[0027] A financial information sequence acquisition unit, wherein the financial information sequence acquisition unit acquires financial information sequences from multi-source heterogeneous terminals;

[0028] A financial information prediction matrix construction unit, wherein the financial information prediction matrix is ​​constructed based on the financial information sequence;

[0029] A prediction dataset extraction unit selects a subset of elements from the financial information prediction matrix as the prediction dataset.

[0030] The predicted dataset is used as input to the target prediction model, which outputs financial information prediction and analysis results.

[0031] The system also includes a financial data simulation unit;

[0032] When the financial information prediction matrix does not meet the preset conditions, the financial data simulation unit is activated to perform data simulation on the financial information prediction matrix to obtain the simulated dimension attribute value of each financial information.

[0033] The simulated dimension attribute value of each piece of financial information is added to the financial information prediction matrix to form the financial information simulation matrix;

[0034] The prediction dataset extraction unit determines the prediction dataset based on the financial information simulation matrix.

[0035] Each piece of financial information in the sequence of financial information obtained from multi-source heterogeneous terminals contains multiple dimensional attribute values, and the multiple dimensional attribute values ​​include at least two of the following attribute values:

[0036] The time when financial information is generated, the location where financial data is generated, the type of terminal that generates financial data, the payment amount included in the financial data, the environment in which the financial data is generated, the transaction object corresponding to the financial data, the gender of the user of the terminal that generates the financial data, and the age group of the user of the terminal that generates the financial data.

[0037] As can be seen, the prediction parameter input unit of the system is used to input prediction parameters, which include information prediction dimension values ​​or information prediction dimension attribute values;

[0038] In one case, when the input prediction parameter is the information prediction dimension value PreW, it can be used to implement the multi-dimensional content information prediction and analysis method described in the first aspect.

[0039] In another scenario, when the input prediction parameter is an information prediction dimension attribute value, it can be implemented as a multi-dimensional content information prediction and analysis method as described in the third aspect below.

[0040] Specifically, the third aspect describes a multi-dimensional content information prediction and analysis method applied to financial information prediction and analysis of multi-source heterogeneous terminals, which specifically includes the following steps:

[0041] S910: Input the information prediction dimension attribute value, and determine the information prediction dimension value PreW based on the information prediction dimension attribute value, wherein the information prediction dimension value PreW is a positive integer;

[0042] S920: Obtain the PreW financial information sequences generated at PreW time points. Every piece of financial information Contains K dimension attribute values , K is a positive integer;

[0043] S930: Constructing the Financial Information Pre-Matrix

[0044] ;

[0045] in,

[0046] S940: Determine whether the information prediction dimension value PreW is not greater than K. If so, select a PreW-order submatrix from the financial information prediction matrix PreMatrix as the prediction dataset based on the information prediction dimension attribute value.

[0047] If not, perform data simulation on the financial information prediction matrix PreMatrix to obtain each piece of financial information. The simulated dimension attribute value;

[0048] Each financial information The simulated dimension attribute values ​​are added to the financial information prediction matrix PreMatrix to form the financial information simulation matrix PreMatrixM;

[0049] Select a PreW-order submatrix from the financial information simulation matrix PreMatrixM as the prediction dataset;

[0050] The PreW-order submatrix contains the information prediction dimension attribute values;

[0051] S950: Load the corresponding prediction model based on the information prediction dimension value PreW, and perform financial information prediction and analysis.

[0052] In the method described in the third aspect above, the prediction model is a prediction model based on a deep neural network;

[0053] The number of nodes in the output layer of the prediction model based on deep neural networks is not less than PreW.

[0054] As can be seen, this invention can realize the simulation and predictive analysis of financial information under different dimensions and prediction objectives.

[0055] Specifically, for financial information with different data attributes generated by multi-source heterogeneous terminals, the technical solution of this invention can set information prediction dimension values ​​and / or information prediction dimension attribute values, and obtain prediction datasets for the corresponding dimensions. Simultaneously, it loads the corresponding prediction model based on the information prediction dimension values ​​and / or information prediction dimension attribute values, thereby making the prediction process more targeted. When the existing financial information sequence data is insufficient to support the prediction output of setting the information prediction dimension values ​​and / or information prediction dimension attribute values, this invention further performs data simulation to obtain the prediction output for each financial information. The simulated dimension attribute values ​​constitute a financial information simulation matrix, thereby maximizing the fulfillment of prediction needs; finally, the technical solution of this invention can adjust the prediction dimension in real time according to the development progress of the prediction model.

[0056] Further advantages of the present invention will be further detailed in the Specific Embodiments section in conjunction with the accompanying drawings. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the 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.

[0058] Figure 1 This is a flowchart illustrating the steps of a multi-dimensional content information prediction and analysis method according to an embodiment of the present invention;

[0059] Figure 2 This is a schematic diagram of the structure of financial information data involved in various embodiments of the present invention;

[0060] Figure 3 This is an architecture diagram of a multi-dimensional content information prediction and analysis system according to an embodiment of the present invention;

[0061] Figure 4 This is an architecture diagram of a multi-dimensional content information prediction and analysis system according to another preferred embodiment of the present invention;

[0062] Figure 5 This is a flowchart illustrating the steps of a multi-dimensional content information prediction and analysis method according to another preferred embodiment of the present invention. Detailed Implementation

[0063] The invention will now be further described in conjunction with the accompanying drawings and specific embodiments.

[0064] See Figure 1 , Figure 1 This is a flowchart illustrating the steps of a multi-dimensional content information prediction and analysis method according to an embodiment of the present invention.

[0065] exist Figure 1 The method includes a judgment process of steps S110-S150, and the specific implementation of each step is as follows:

[0066] S110: Determine the information prediction dimension value PreW, wherein the information prediction dimension value PreW is a positive integer;

[0067] S120: Obtain the PreW financial information sequences generated at PreW time points. Every piece of financial information Contains K dimension attribute values , K is a positive integer;

[0068] S130: Constructing the Financial Information Pre-Matrix

[0069] ;

[0070] in,

[0071] S140: Determine whether the information prediction dimension value PreW is not greater than K. If so, determine the corresponding information prediction dimension attribute value based on the information prediction dimension value PreW, and select the PreW order submatrix as the prediction dataset from the financial information prediction matrix PreMatrix.

[0072] S150: Based on the predicted dataset, perform financial information prediction and analysis.

[0073] The following section details the specific implementation of each step.

[0074] Step S110 requires determining the information prediction dimension value PreW, where PreW is a positive integer.

[0075] To explain the meaning of the information prediction dimension value PreW, see [link to relevant documentation]. Figure 2 First, we will introduce the structural diagrams of the financial information data involved in the various embodiments of the present invention.

[0076] The technical solution of this invention is for financial information prediction and analysis of multi-source heterogeneous terminals.

[0077] Multi-source heterogeneous terminals are mobile terminals and desktop terminals with different software / hardware operating systems. These multi-source heterogeneous terminals have at least one financial application (APP) or client installed.

[0078] In this invention, financial applications (APPs) or clients include all applications (APPs) or clients involved in financial transactions. These applications (APPs) or clients involved in financial transactions will generate financial transaction data, although the types and quantities of financial transaction data generated by different applications (APPs) or clients involved in financial transactions are different.

[0079] As an example, a bank-related financial app installed on a mobile Android device can generate financial transaction data including payments, transfers, facial recognition, and password payments; it can also generate financial transaction data such as facial recognition, password payments, and bank-securities transfer data.

[0080] Different heterogeneous terminals from different sources generate different financial transaction data at different points in time. A single terminal may have multiple financial apps installed, generating financial transaction data with different dimensions at different points in time.

[0081] In this invention, the dimension value of financial transaction data refers to the various attribute values ​​inherent in the financial transaction data itself.

[0082] Specifically, attribute values ​​may include: the time when the financial information was generated, the location where the financial data was generated, the type of terminal that generated the financial data, the payment amount included in the financial data, the environment in which the financial data was generated, the transaction object corresponding to the financial data, the gender of the user of the terminal that generated the financial data, and the age group of the user of the terminal that generated the financial data.

[0083] Figure 2 The following shows the representation of all attribute values ​​of financial data generated by a certain mobile Android terminal in a specific instance ( Figure 2 This illustrates the storage format of computer device memory in a specific implementation of the present invention.

[0084] Time of creation: 2:35 PM; January 1, 2020

[0085] Location (IP): 202.123.456.231

[0086] Device type: Android 4.5.3

[0087] Payment amount: $35.00

[0088] Environment in which it was created: App Store

[0089] Target of the transaction: Skirt

[0090] User gender: male

[0091] User age group: 30-40

[0092] It is understandable that financial transaction data generated by different heterogeneous terminals, at different time points, and by different financial apps will have different attribute values. Depending on the collection permission requirements of the financial terminal and the financial app, the financial transaction data generated by different heterogeneous terminals, at different time points, and by different financial apps may include some or all of the above attributes.

[0093] Figure 2 The dimensional attribute values ​​for all eight dimensions are shown.

[0094] However, in actual information prediction and analysis, the number of dimensions to be predicted is not necessarily 8; it could be 4 or even 9.

[0095] For example, to meet the needs of a marketing campaign, it is necessary to predict the time of generation of relevant financial data, the transaction targets, and the gender of users, so as to accurately push the corresponding transaction targets to users of the corresponding gender at the corresponding time points.

[0096] At this point, the information prediction dimension value PreW determined in step S110 is three;

[0097] Of course, in another case, the output dimension of a certain prediction model is only 4. Therefore, it can only predict the financial attribute values ​​of 4 dimensions each time. In this case, the information prediction dimension value PreW=4 is directly set, and the specific attribute values ​​corresponding to the 4 dimensions are determined later.

[0098] Of course, while financial transaction data generated by different heterogeneous terminals, at different times, and from different financial apps may only contain a portion of the attributes mentioned above, as a practical example, the K-dimensional attribute values... Includes at least two of the following attribute values:

[0099] The time when financial information is generated, the location where financial data is generated, the type of terminal that generates financial data, the payment amount included in the financial data, the environment in which the financial data is generated, the transaction object corresponding to the financial data, the gender of the user of the terminal that generates the financial data, and the age group of the user of the terminal that generates the financial data.

[0100] Next, we construct the financial information prediction matrix PreMatrix:

[0101] ;

[0102] in,

[0103] If the information prediction dimension value PreW is not greater than K, then the corresponding information prediction dimension attribute value is determined based on the information prediction dimension value PreW, and the PreW order submatrix is ​​selected as the prediction dataset from the financial information prediction matrix PreMatrix.

[0104] It is understandable that when the information prediction dimension value PreW is not greater than K, it means that the generated data itself has sufficient dimensional attributes. Therefore, the dataset with the corresponding dimension can be directly selected as the prediction dataset.

[0105] For example, suppose

[0106]

[0107] That is, the data that has been generated has four dimensions of attributes: {time when the financial information was generated, location where the financial data was generated, amount of payment included in the financial data, and gender of the user at the terminal that generated the financial data}.

[0108] That is, K=4; while the predicted information dimension value PreW is determined to be 3 in this case.

[0109] In step S140, based on the information prediction dimension value PreW=3, the corresponding information prediction dimension attribute value is determined to be any three of the following: {financial information generation time, financial data generation location, payment amount included in the financial data, and user gender of the financial data generation terminal}.

[0110] Assume that based on the information prediction dimension value PreW=3, the corresponding information prediction dimension attribute value is determined to be {financial information generation time, financial data generation location, and payment amount included in the financial data};

[0111] The PreW=3 submatrix of the prediction dataset is selected from the financial information prediction matrix PreMatrix. as follows:

[0112]

[0113] or

[0114]

[0115] It can be seen that the PreW order submatrix All that is needed is a PreW-order submatrix that contains the information prediction dimension attribute value in the financial information prediction matrix PreMatrix. That is, there can be multiple PreW-order submatrices. In the subsequent step S150, multiple predictions are performed based on multiple PreW-order submatrices.

[0116] As a further preferred option, it is determined whether the information prediction dimension value PreW is not greater than K. If so, the corresponding information prediction dimension attribute value is determined based on the information prediction dimension value PreW, and multiple PreW-order sub-matrices are selected from the financial information prediction matrix PreMatrix.

[0117] Calculate the eigenvalues ​​of each PreW-order submatrix.

[0118] PreW-order submatrices with eigenvalues ​​less than 1 are used as the prediction dataset.

[0119] In the preferred embodiment described above, using the PreW-order submatrix with eigenvalues ​​less than 1 as the prediction dataset ensures the stability of the prediction dataset. This is because an eigenvalue less than 1 implies compliance with the stability principles of a control system.

[0120] If the information prediction dimension value PreW is greater than K, then data simulation is performed on the financial information prediction matrix PreMatrix to obtain each financial information. The simulated dimension attribute value;

[0121] Each financial information The simulated dimension attribute values ​​are added to the financial information prediction matrix PreMatrix to form the financial information simulation matrix PreMatrixM;

[0122] The PreW-order submatrix is ​​selected from the financial information simulation matrix PreMatrixM as the prediction dataset.

[0123] It is understandable that if the information prediction dimension value PreW is greater than K, it means that the generated data itself does not have enough dimensional attributes and cannot be directly used for prediction. In this case, financial data simulation operation needs to be performed to supplement the dimensions.

[0124] As an example, assume that the generated data only has 5 dimensions: {time of financial information generation, type of financial data generation terminal, payment amount included in financial data, environment in which financial data is generated, and transaction target corresponding to financial data}, i.e., K=5;

[0125] The information prediction dimension value PreW=6 set in this case means that the existing prediction dataset is insufficient and needs to be supplemented through simulation.

[0126] Preferably, data simulation is performed on the financial information prediction matrix PreMatrix to obtain each piece of financial information. The simulated dimension attribute value.

[0127] In the example above, data simulation can be performed based on some or all of the attribute values ​​of {the time when the financial information was generated, the type of the terminal that generated the financial data, the payment amount included in the financial data, the environment in which the financial data was generated, and the transaction target corresponding to the financial data} to obtain the gender of the user of the terminal that generated the financial data.

[0128] The specific data simulation can be obtained using big data analysis, deep learning models, etc., and this embodiment does not impose specific limitations on this.

[0129] As an example, based on the "trading target corresponding to the financial data" (e.g., Skirt) + "the environment in which the financial data was generated" (e.g., a shopping app), we can infer that "the gender of the user who generated this financial data" is female, and use this as a simulated dimension attribute value.

[0130] Similarly, other dimension values ​​such as "the age range of users at the terminal that generated this financial data" can be inferred to simulate dimension attribute values.

[0131] Each financial information The simulated dimension attribute values ​​are added to the financial information prediction matrix PreMatrix to form the financial information simulation matrix PreMatrixM.

[0132] In practical implementation, the set information prediction dimension value PreW also needs to take into account the specific parameters of the available prediction models.

[0133] Specifically, the output dimension of a certain prediction model has only 4 dimensions. Therefore, it can only predict the financial attribute values ​​of 4 dimensions each time. In this case, the information prediction dimension value PreW=4 is directly set, and the specific attribute values ​​corresponding to the 4 dimensions will be determined later.

[0134] That is, the information prediction dimension value PreW in step S110 is determined by the number of nodes in the output layer of the prediction model based on the deep neural network.

[0135] As development progresses, various prediction models with different dimensions can exist. Therefore, step S150 uses various different prediction models to perform the financial information prediction; different prediction models correspond to different information prediction dimension values ​​PreW.

[0136] Therefore, the advantage of this embodiment is that the information prediction dimension attribute value or information prediction dimension value can be set according to the development progress of the existing prediction model and the specific parameters of the actual prediction model, thereby ensuring that the existing multiple prediction models can be applied to the financial information prediction and analysis of multi-source heterogeneous terminals, without having to consider the limitations of the specific data types and quantities generated.

[0137] See next. Figures 3-4 .

[0138] Figure 3 This is an architecture diagram of a multi-dimensional content information prediction and analysis system according to an embodiment of the present invention.

[0139] exist Figure 3 The system includes a prediction model database, which stores at least one prediction model.

[0140] The system also includes:

[0141] A prediction parameter input unit is used to input prediction parameters, which include information prediction dimension values ​​or information prediction dimension attribute values.

[0142] The model automatic selection unit loads at least one target prediction model from the prediction model database based on the prediction parameters input by the prediction parameter input unit.

[0143] A financial information sequence acquisition unit, wherein the financial information sequence acquisition unit acquires financial information sequences from multi-source heterogeneous terminals;

[0144] A financial information prediction matrix construction unit, wherein the financial information prediction matrix is ​​constructed based on the financial information sequence;

[0145] A prediction dataset extraction unit selects a subset of elements from the financial information prediction matrix as the prediction dataset.

[0146] The predicted dataset is used as input to the target prediction model, which outputs financial information prediction and analysis results.

[0147] exist Figure 3 On this basis, Figure 4 This is an architecture diagram of a multi-dimensional content information prediction and analysis system according to another preferred embodiment of the present invention.

[0148] Figure 4 exist Figure 3 Based on this, the system includes a financial data simulation unit.

[0149] When the financial information prediction matrix does not meet the preset conditions, the financial data simulation unit is activated to perform data simulation on the financial information prediction matrix to obtain the simulated dimension attribute value of each financial information; the simulated dimension attribute value of each financial information is added to the financial information prediction matrix to form a financial information simulation matrix.

[0150] The prediction dataset extraction unit determines the prediction dataset based on the financial information simulation matrix.

[0151] Figures 3-4 The system can be used to implement Figure 1 The method described above, namely, the financial information sequence obtained from multi-source heterogeneous terminals, wherein each piece of financial information contains multiple dimensional attribute values, and the multiple dimensional attribute values ​​include at least two of the following attribute values:

[0152] The time when financial information is generated, the location where financial data is generated, the type of terminal that generates financial data, the payment amount included in the financial data, the environment in which the financial data is generated, the transaction object corresponding to the financial data, the gender of the user of the terminal that generates the financial data, and the age group of the user of the terminal that generates the financial data.

[0153] As can be seen, the prediction parameter input unit of the system is used to input prediction parameters, which include information prediction dimension values ​​or information prediction dimension attribute values;

[0154] In one case, when the input prediction parameter is the information prediction dimension value PreW, it can be used to implement the multi-dimensional content information prediction and analysis method described in the first aspect.

[0155] In another scenario, when the input prediction parameters are information prediction dimension attribute values, it can be achieved as follows: Figure 5 A more preferred embodiment of the multi-dimensional content information prediction and analysis method, wherein the embodiment includes steps S910-S950, for the sake of brevity (e.g., most of the description is similar to...), Figure 1 repeat), Figure 5 Only the step numbers are shown; the specific steps are as follows.

[0156] S910: Input the information prediction dimension attribute value, and determine the information prediction dimension value PreW based on the information prediction dimension attribute value, wherein the information prediction dimension value PreW is a positive integer;

[0157] S920: Obtain the PreW financial information sequences generated at PreW time points. Every piece of financial information Contains K dimension attribute values , K is a positive integer;

[0158] S930: Constructing the Financial Information Pre-Matrix

[0159] ;

[0160] in,

[0161] S940: Determine whether the information prediction dimension value PreW is not greater than K. If so, select a PreW-order submatrix from the financial information prediction matrix PreMatrix as the prediction dataset based on the information prediction dimension attribute value.

[0162] If not, perform data simulation on the financial information prediction matrix PreMatrix to obtain each piece of financial information. The simulated dimension attribute value;

[0163] Each financial information The simulated dimension attribute values ​​are added to the financial information prediction matrix PreMatrix to form the financial information simulation matrix PreMatrixM;

[0164] Select a PreW-order submatrix from the financial information simulation matrix PreMatrixM as the prediction dataset;

[0165] The PreW-order submatrix contains the information prediction dimension attribute values;

[0166] S950: Load the corresponding prediction model based on the information prediction dimension value PreW, and perform financial information prediction and analysis.

[0167] Figure 5 The prediction model described in this embodiment is a prediction model based on a deep neural network;

[0168] The number of nodes in the output layer of the prediction model based on deep neural networks is not less than PreW, and each output layer node outputs a predicted value for a dimensional attribute.

[0169] It should be noted that after obtaining the prediction dataset, various existing prediction models can be used, such as multidimensional time series prediction models, multidimensional large-scale prediction models, deep neural network prediction models, etc. This invention will not elaborate on these, as they are not the focus of this invention.

[0170] Specifically, for financial information with different data attributes generated by multi-source heterogeneous terminals, the technical solution of the present invention can set information prediction dimension values ​​and / or information prediction dimension attribute values, obtain prediction datasets of corresponding dimensions, and load corresponding prediction models according to information prediction dimension values ​​and / or information prediction dimension attribute values, thereby making the prediction process more targeted.

[0171] When the existing financial information sequence data is insufficient to support the prediction output of setting information prediction dimension values ​​and / or information prediction dimension attribute values, the present invention further performs data simulation to obtain each financial information. The simulated dimensional attribute values ​​constitute the financial information simulation matrix, thereby maximizing the fulfillment of prediction needs.

[0172] Another advantage of the technical solution of the present invention is that the information prediction dimension attribute value or information prediction dimension value can be set according to the development progress of the existing prediction model and the specific parameters of the actual prediction model. This ensures that the existing multiple prediction models can be applied to the financial information prediction and analysis of multi-source heterogeneous terminals, without having to consider the limitations of the specific data types and quantities generated.

[0173] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

[0174] For any module structures not specifically defined in this invention, the description in the prior art shall prevail. The prior art mentioned in the foregoing background section can be considered part of this invention and used to understand the meaning of certain technical features or parameters. The scope of protection of this invention is determined by the actual content of the claims.

Claims

1. A multi-dimensional information prediction and analysis system; the system includes a prediction model database, wherein the prediction model database stores at least one prediction model; Its features are, The system also includes: A prediction parameter input unit is used to input prediction parameters, which include information prediction dimension values ​​or information prediction dimension attribute values. The model automatic selection unit loads at least one target prediction model from the prediction model database based on the prediction parameters input by the prediction parameter input unit. A financial information sequence acquisition unit, wherein the financial information sequence acquisition unit acquires financial information sequences from multi-source heterogeneous terminals; A financial information prediction matrix construction unit, wherein the financial information prediction matrix is ​​constructed based on the financial information sequence; A prediction dataset extraction unit selects a subset of elements from the financial information prediction matrix as the prediction dataset. The predicted dataset is used as input to the target prediction model, and the target prediction model outputs financial information prediction and analysis results. The prediction parameter input unit determines the information prediction dimension value PreW, where the information prediction dimension value PreW is a positive integer. The financial information sequence acquisition unit acquires PreW financial information sequences generated at PreW time points. Every piece of financial information Contains K dimension attribute values , K is a positive integer; The financial information prediction matrix construction unit constructs the financial information prediction matrix PreMatrix: ； in, The prediction dataset extraction unit determines whether the information prediction dimension value PreW is not greater than K. If so, it determines the corresponding information prediction dimension attribute value based on the information prediction dimension value PreW and selects the PreW order submatrix from the financial information prediction matrix PreMatrix as the prediction dataset.

2. The information prediction and analysis system for multi-dimensional content as described in claim 1, characterized in that, The system also includes a financial data simulation unit; When the financial information prediction matrix does not meet the preset conditions, the financial data simulation unit is activated to perform data simulation on the financial information prediction matrix to obtain the simulated dimension attribute value of each financial information. The simulated dimension attribute value of each piece of financial information is added to the financial information prediction matrix to form the financial information simulation matrix; The prediction dataset extraction unit determines the prediction dataset based on the financial information simulation matrix.

3. The information prediction and analysis system for multi-dimensional content as described in claim 2, characterized in that, Each piece of financial information in the sequence of financial information obtained from multi-source heterogeneous terminals contains multiple dimensional attribute values, and the multiple dimensional attribute values ​​include at least two of the following attribute values: The time when financial information is generated, the location where financial data is generated, the type of terminal that generates financial data, the payment amount included in the financial data, the environment in which the financial data is generated, the transaction object corresponding to the financial data, the gender of the user of the terminal that generates the financial data, and the age group of the user of the terminal that generates the financial data.

Images