WO2022257457A1

WO2022257457A1 - Product data fusion method, apparatus and device, and storage medium

Info

Publication number: WO2022257457A1
Application number: PCT/CN2022/071476
Authority: WO
Inventors: 沈嘉良; 王遥
Original assignee: 平安科技（深圳）有限公司
Priority date: 2021-06-09
Filing date: 2022-01-12
Publication date: 2022-12-15
Also published as: CN113095604B; CN113095604A

Abstract

The present application relates to the field of big data. Provided are a product data fusion method, apparatus and device, and a storage medium, which are used for improving the value prediction accuracy for product data. The product data fusion method comprises: performing asset-category-based classification on product data to be processed, so as to obtain a plurality of classified product data sets; performing normalization and fusion processing on the plurality of classified product data sets, so as to obtain normalized and fused data; performing value autoregression prediction on the plurality of classified product data sets by means of a structural vector autoregressive model, target index data and the normalized and fused data, so as to obtain initial value prediction data; performing inverse normalization processing on the initial value prediction data, so as to obtain candidate value prediction data; and performing correlation fusion processing on the candidate value prediction data, so as to obtain target value prediction data. In addition, the present application further relates to blockchain technology, and product data to be processed can be stored in a blockchain.

Description

Product data fusion method, device, equipment and storage medium

This application claims the priority of the Chinese patent application with the application number 202110639687.3 and the title of the invention "Product data fusion method, device, equipment and storage medium" submitted to the China Patent Office on June 9, 2021, the entire content of which is incorporated by reference incorporated in the application.

technical field

The present application relates to the field of prediction and valuation of big data, and in particular to a fusion method, device, equipment and storage medium of product data.

Background technique

With the development of computer technology, the demand for asset valuation has gradually increased. At present, it is generally through the asset management system to use the cash flow conversion method or the market price comparison method to predict the value of the product data to be processed and obtain the product value data.

However, the inventor realized that because the cash flow conversion method or the market price comparison method predicts the value from a single influencing factor, it cannot accurately predict the value of different types and changing product data, and its prediction flexibility is low , thus resulting in low accuracy of value prediction of product data.

Contents of the invention

The present application provides a product data fusion method, device, equipment and storage medium, which are used to improve the value prediction accuracy of product data.

The first aspect of the present application provides a product data fusion method, including:

Obtaining product data to be processed, and classifying the product data to be processed based on asset categories to obtain multiple classified product data sets;

Perform normalized fusion processing on the plurality of classified product data sets respectively through a preset normalized model, to obtain normalized fusion data corresponding to each classified product data;

Obtain the target index data corresponding to each classified product data, and use the preset structural vector autoregressive model, the target index data corresponding to each classified product data, and the normalized fusion data corresponding to each classified product data, to The plurality of classified product data sets are respectively subjected to auto-regression value prediction to obtain initial value forecast data corresponding to each classified product data set, and the target index data includes economic index data and market index data;

Using the normalization model, denormalize the initial value prediction data corresponding to each classified product data set to obtain candidate value prediction data corresponding to each classified product data set;

Correlation fusion processing is performed on the candidate value prediction data corresponding to each classified product data set to obtain target value prediction data.

The second aspect of the present application provides a product data fusion device, including a memory, a processor, and computer-readable instructions stored on the memory and operable on the processor, and the processor executes the computer The following steps are implemented when the instruction is readable:

The third aspect of the present application provides a computer-readable storage medium, where computer instructions are stored in the computer-readable storage medium, and when the computer instructions are run on the computer, the computer is made to perform the following steps:

The fourth aspect of the present application provides a product data fusion device, including:

A classification module, configured to obtain product data to be processed, and classify the product data to be processed based on asset categories to obtain multiple classified product data sets;

The normalized fusion module is used to perform normalized fusion processing on the multiple classified product data sets respectively through a preset normalized model, so as to obtain normalized fusion data corresponding to each classified product data;

The prediction module is used to obtain the target index data corresponding to each classified product data, and use the preset structural vector autoregressive model, the target index data corresponding to each classified product data and the normalization corresponding to each classified product data Combining the data, performing value autoregressive prediction on the plurality of classified product data sets respectively, and obtaining initial value forecast data corresponding to each classified product data set, the target index data including economic index data and market index data;

The denormalization module is used to perform denormalization processing on the initial value prediction data corresponding to each classified product data set through the normalization model, so as to obtain candidate value prediction data corresponding to each classified product data set;

The correlation fusion module is used to perform correlation fusion processing on the candidate value prediction data corresponding to each classified product data set to obtain target value prediction data.

In the embodiment of the present application, by classifying the product data to be processed based on asset categories, it is convenient for subsequent targeted and effective processing of product data to be processed of different asset types; by separately classifying multiple classified product data sets Normalized fusion processing integrates multiple influencing factors of different asset types, enriches the predictive factors of the value of different types and changing product data, and improves the value prediction accuracy of different types and changing product data; through preset The structural vector autoregressive model, the target index data corresponding to each category of product data, and the normalized fusion data corresponding to each category of product data, the value autoregressive prediction is performed on multiple category product data sets, and the value of each category product data set The corresponding initial value prediction data is denormalized to realize the calculation of the structural relationship between the standard asset price index (that is, the target index data), and to estimate the value of the assets under the user's name in the future based on the predicted initial value prediction data change, and make it meet the risk measurement standards such as sub-additivity, exchangeability, and consistency; perform correlation fusion processing on the candidate value prediction data corresponding to each classified product data set, so that it meets the requirements of additivity, exchangeability, and consistency. Consistency and other risk measurement standards, thereby improving the value prediction accuracy of product data.

Description of drawings

Fig. 1 is the schematic diagram of an embodiment of the fusion method of product data in the embodiment of the present application;

Fig. 2 is the schematic diagram of another embodiment of the fusion method of product data in the embodiment of the present application;

Fig. 3 is a schematic diagram of an embodiment of a fusion device of product data in the embodiment of the present application;

FIG. 4 is a schematic diagram of another embodiment of the product data fusion device in the embodiment of the present application;

Fig. 5 is a schematic diagram of an embodiment of a product data fusion device in the embodiment of the present application.

Detailed ways

Embodiments of the present application provide a product data fusion method, device, equipment, and storage medium, which improve the accuracy of value prediction of product data.

The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above drawings are used to distinguish similar objects, and not necessarily Used to describe a specific sequence or sequence. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the term "comprising" or "having" and any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, system, product or device comprising a sequence of steps or elements is not necessarily limited to those explicitly listed instead, may include other steps or elements not explicitly listed or inherent to the process, method, product or apparatus.

For ease of understanding, the following describes the specific process of the embodiment of the present application. Please refer to FIG. 1. An embodiment of the product data fusion method in the embodiment of the present application includes:

101. Obtain the product data to be processed, and classify the product data to be processed based on asset categories to obtain multiple classified product data sets.

It can be understood that the subject of execution of this application may be a product data fusion device, or a terminal or a server, which is not specifically limited here. The embodiment of the present application is described by taking the server as an execution subject as an example.

Among them, the products in the product data to be processed are specific objects and financial products that can be used for asset valuation, such as: mortgages (houses, cars) and insurance policies, and the product data to be processed include the basic asset information of specific objects and Historical price data, asset basic information and historical price data of financial products, and other data that can be used for asset valuation. The asset category includes a main category and multiple subcategories. For example, the main category is A (insurance policy), and the corresponding subcategories are a1 (social insurance policy) and a2 (commercial insurance policy).

The server can extract the initial product data to be valued from the preset database after obtaining user authorization, or the server can receive the initial product data to be valued sent by the user terminal (target terminal); the server obtains the initial product data to be valued After the product data, the initial product data to be valued is preprocessed for security detection and deduplication to obtain the preprocessed product data; the server can classify the preprocessed product data based on keywords to obtain multiple Classify product data; the server can also classify the pre-processed product data by asset category through a preset classification model, so as to obtain multiple classified product data.

The server can classify the pre-processed product data based on keywords to obtain multiple classified product data. The execution process may specifically include: the server sequentially performs word segmentation and keyword extraction on the pre-processed product data to obtain target keywords; Match the target keyword with the preset asset category vocabulary to get the target asset category vocabulary, obtain the asset category of the target asset category vocabulary, use the resource category as the label of the preprocessed product data corresponding to the target keyword, and preprocess the product data according to the label The product data is classified to obtain multiple classified product data sets, each classified product data set includes product data of multiple subcategories after classification, and one classified product data set corresponds to one main category.

102. Through the preset normalization model, perform normalization fusion processing on multiple classified product data sets respectively, and obtain normalized fusion data corresponding to each classified product data.

Through the preset normalization model, the server performs normalization and fusion processing on multiple classified product data sets respectively: the server obtains the predictive factors corresponding to each classified product data, which includes the The index data, price data and coefficients that affect the overall ups and downs trends and ranges of assets in different periods, the predictors include the normalized benchmark index data and the normalized benchmark index coefficient of the normalized benchmark index data; the server calls pre- The created normalization model is based on the predictive factors, and performs linear regression processing and operation processing on multiple classified product data respectively, so as to obtain the normalized fusion data corresponding to each classified product data. The overall ups and downs trends and ranges of asset data in different periods are normalized and fused.

Among them, the server invokes the pre-created normalization model based on the predictive factors to perform linear regression processing and calculation processing on multiple classified product data, so as to obtain the normalized fusion data corresponding to each classified product data. The specific implementation process can be as follows: The server invokes the pre-created normalization model based on the predictive factors, and performs linear regression processing and calculation processing based on subcategories on multiple classified product data, and obtains the normalized valuation data of the main category corresponding to multiple subcategories. The normalized valuation data of the main category corresponding to multiple subcategories are summed to obtain the normalized fusion data corresponding to each category of product data. For example: take the category product dataset B as the main category B as an example Note that the classified product data set B includes classified product data b11 corresponding to subcategory b1 and classified product data b21 corresponding to subcategory b2, the main category corresponding to subcategory b1 and subcategory b2 is B, and the server calls the pre-created normalization model based on Predictor, perform subcategory-based linear regression processing and calculation processing on the classified product data b11 to obtain the main category normalized valuation data B1 corresponding to the main category B of the subcategory b11, and similarly obtain the normalized valuation data of the main category B2, add the normalized valuation data B1 of the main category and the normalized valuation data B2 of the main category to obtain the normalized fusion data corresponding to the classified product data set B.

By normalizing and merging multiple classified product data sets and integrating various influencing factors of different asset types, the predictive factors for the value of different types and changing product data are enriched, and the value of different types and changing product data is improved. value prediction accuracy.

103. Obtain the target index data corresponding to each category of product data, and use the preset structural vector autoregressive model, the target indicator data corresponding to each category of product data, and the normalized fusion data corresponding to each category of product data to classify multiple categories of products The value autoregressive prediction is performed on the data sets respectively, and the initial value prediction data corresponding to each classified product data set is obtained. The target index data includes economic index data and market index data.

According to the asset category corresponding to each classified product data set, the server matches the index data stored in the preset database, and obtains the economic index data and market index data corresponding to multiple classified product data sets, that is, the target index data, and obtains the value Factor factors for autoregressive prediction, which include disturbance items, independent variable coefficients, and model parameters; the server uses the preset structural vector autoregressive model, based on the normalization of target index data, factor factors, and product data of each category Fusion data, perform fusion processing based on autoregression on multiple classified product data sets, and obtain the initial value prediction data corresponding to each classified product data set, the initial value prediction data is the prediction time (that is, the normalized fusion processing time t+1 The asset value data corresponding to the moment of ), and the initial value forecast data corresponding to each classified product data set include the structural relationship between the standard asset price indices. Among them, economic index data may include but not limited to consumer price index (consumer price index, CPI), gross domestic product (gross domestic product, GDP) and industrial production (industrial production), market index data may include but not limited to flow characteristics, industry conditions, production costs, resource costs and market demand.

Through the preset structural vector autoregressive model, value autoregressive prediction is performed on multiple classified product data sets, and the calculation of the structural relationship between standard asset price indexes is realized, and future users are estimated based on the predicted initial value prediction data Changes in the value of assets under the name, and make them meet risk measurement standards such as sub-additivity, exchangeability, and consistency, which improves the accuracy of value prediction of product data.

104. Using the normalization model, denormalize the initial value prediction data corresponding to each classified product data set to obtain candidate value prediction data corresponding to each classified product data set.

The server obtains the forecast benchmark index data corresponding to the initial value forecast data corresponding to each classified product data set, and the forecast benchmark index coefficient of the forecast benchmark index data, and combines the initial value forecast data, forecast benchmark index data and The prediction benchmark index coefficient is determined as the operation factor of denormalization, and the normalization model is called, based on the preset denormalization formula, the operation factor is calculated to obtain the candidate value prediction data corresponding to each classified product data set, where , the forecast benchmark index data is used to indicate the index factor that affects the denormalization processing of the asset price data based on the preset benchmark price index.

105. Perform correlation fusion processing on the candidate value prediction data corresponding to each classified product data set to obtain target value prediction data.

The specific implementation process for the server to perform correlation fusion processing on the candidate value prediction data corresponding to each classified product data set can be as follows: the server can use the structural relationship (correlation) between the candidate value prediction data corresponding to each classified product data set Perform numerical measurement to obtain the correlation coefficient corresponding to each classified product data set; the server can also extract the data related to the structural relationship from the calculation data generated by the calculation process of the candidate value prediction data corresponding to each classified product data set, thereby Obtain the correlation coefficient corresponding to each classified product data set; the server determines the corresponding correlation coefficient of each classified product data set as the corresponding weighting coefficient of each classified product data set, and calculates the sum value of each classified product data set according to the weighting coefficient, thus obtaining Target value forecast data.

Through the correlation fusion processing of the candidate value prediction data corresponding to each classified product data set, it meets the risk measurement standards such as subadditivity, commutativity and consistency, and improves the value prediction accuracy of product data.

Please refer to Fig. 2, another embodiment of the fusion method of product data in the embodiment of the present application includes:

201. Obtain the product data to be processed, and classify the product data to be processed based on asset categories to obtain multiple classified product data sets.

Specifically, the server receives the asset valuation request sent by the target terminal, reads the product data to be processed from the preset database based on the asset valuation request, and performs data preprocessing on the product data to be processed to obtain preprocessed product data; Through the preset classification model, multi-level convolution-based feature extraction, attention mechanism-based feature fusion, asset category probability calculation, and asset category classification are sequentially performed on the preprocessed product data to obtain multiple classified product datasets.

The server receives the asset valuation request sent by the target terminal, analyzes the asset valuation request, and obtains the key asset valuation information. The asset valuation key information includes product data requirements and asset value forecast demand information, read from the preset database Get the product data to be processed corresponding to the key information of asset valuation; perform data cleaning, data conversion, de-duplication fusion and security detection on the product data to be processed in sequence (that is, data preprocessing includes data cleaning, data conversion, de-duplication and fusion and safety testing) to obtain preprocessed product data.

The server invokes the preset classification model, based on the pre-built multi-level convolutional neural network algorithm, performs hierarchical convolution feature extraction on the preprocessed product data, obtains the features of each level, and based on the preset attention mechanism, extracts the features of each level Carry out attention-based fusion to obtain target features, calculate the asset class probability of target features through the classification network in the preset classification model, and determine the target asset class according to the calculated asset class probability, and preprocess according to the target asset class The product data is classified to obtain multiple classified product data sets, wherein the convolutional neural network algorithms at each level may be the same or different.

Through the multi-level convolutional neural network algorithm, feature extraction is performed on the pre-processed product data, which improves the accuracy of the preferential total energy of the extracted pre-processed product data. By classifying the product data to be processed based on the asset category, it is convenient for subsequent targeted and effective processing of the product data to be processed of different asset types, thereby improving the value prediction accuracy of product data.

202. Through the preset normalization model, perform normalization fusion processing on multiple classified product data sets respectively, to obtain normalized fusion data corresponding to each classified product data.

Specifically, the server obtains asset price data, normalized benchmark index data, and normalized benchmark index coefficients of the multiple classified product data sets respectively corresponding to the data sets. The normalized benchmark index data is used to indicate the On the basis of the preset benchmark price index, the index factor that affects the normalization and fusion processing of asset price data; through the preset normalization model, based on the moment asset price data, normalized benchmark index data and normalized benchmark The index coefficient is to perform normalized value data calculations on multiple classified product data sets, and obtain the normalized fusion data corresponding to each classified product data set.

Among them, the normalized benchmark index data is used to indicate the index factors that affect the normalized fusion processing of asset price data based on the preset benchmark price index. For example, the index factors corresponding to housing prices include location and house age.

For example, taking the classified product data set A as the main category A of the classified product data set as an example, the classified product data set A includes the classified product data a11 corresponding to the subcategory a1 and the classified product data a21 corresponding to the subcategory a2, and the server obtains the classified product Asset price data a1 _t of data a11 at time t, normalized benchmark index data x1 _t,k and normalized benchmark index coefficient c1 _t,k , and asset price data a2 t of classified product data a21 at time _t , normalized benchmark index data x2 _t,k and normalized benchmark index coefficient c2 _t,k , through the preset normalization model, based on the preset calculation formula 1, instant asset price data a1 _t , normalized The benchmark index data x1 _t,k and the normalized benchmark index coefficient c1 _t,k calculate the normalized fusion data A _1t corresponding to the classified product data a11, where the preset calculation formula 1 is:

k and K represent the number of indicators of the normalized benchmark index data, and through the preset normalization model, based on the preset calculation formula 2, moment asset price data a2 _t , normalized benchmark index data x2 _t,k and the normalized benchmark index coefficient c2 _t,k to calculate the normalized fusion data A _2t corresponding to the classified product data a21, where the preset calculation formula 2 is:

k and K represent the number of indicators of the normalized benchmark index data, and the normalized fusion data A _1t and the normalized fusion data A _2t are added to realize the normalized fusion processing on the classified product data set A , to obtain the normalized fusion data A _t corresponding to the classified product data set A.

203. Obtain the target indicator data corresponding to each category of product data, and use the preset structural vector autoregressive model, the target indicator data corresponding to each category of product data and the normalized fusion data corresponding to each category of product data to perform multiple classification products The value autoregressive prediction is performed on the data sets respectively, and the initial value prediction data corresponding to each classified product data set is obtained. The target index data includes economic index data and market index data.

Specifically, the server obtains target index data, disturbance item data, and independent variable coefficients corresponding to multiple classified product data sets, and the target index data includes economic index data and market index data corresponding to multiple classified product data sets; Structural vector autoregressive model, the target index data, disturbance item data and independent variable coefficients corresponding to each category of product data, and the normalized fusion data corresponding to each category of product data, the structure vector autoregressive Regression operation processing to obtain the initial value forecast data corresponding to each classified product data set.

For example, taking the classified product data set A as the main category A and the classified product data set B as the main category B as an example, the normalized fusion data corresponding to the classified product data set A is A _n , the normalized fusion data corresponding to the classified product data set B is B _n , the server obtains the target index data x _Ap corresponding to the classified product data set A, the disturbance item data ε _A , and the normalized fusion data is A _n corresponding to The variable coefficient c _n and the independent variable coefficient e _p corresponding to the target index data x _Ap , and obtain the target index data x _Bp corresponding to the classified product data set B, the disturbance item data ε _B , and the independent variables of the normalized fusion data B _n The coefficient g _n and the independent variable coefficient h _p of the target index data x _Bp , the server uses the calculation formula in the preset structural vector autoregressive model

Among them, n represents the number of periods that change with time, p and P represent the number of indicators of the target indicator data, and the structural vector autoregressive operation is performed to obtain the initial value prediction data A _t+1 corresponding to the classified product data set A, The server uses the calculation formula in the preset structure vector autoregressive model

Among them, n represents the number of periods that change with time, p and P represent the number of indicators of the target index data, and the operation processing of structural vector autoregression is performed to obtain the initial value prediction data B _t+1 corresponding to the classified product data set B.

204. Using the normalization model, denormalize the initial value prediction data corresponding to each classified product data set to obtain candidate value prediction data corresponding to each classified product data set.

Specifically, the server acquires the forecast benchmark index data of the initial value forecast data corresponding to each classified product data set, and the forecast benchmark index coefficients of the forecast benchmark index data, and the forecast benchmark index data is used to indicate the value based on the preset benchmark price index , the index factor that affects the denormalization processing of asset price data; through the normalization model, based on the forecast benchmark index data and forecast benchmark index coefficient, the preset denormalization formula and the initial value prediction corresponding to each classified product data set The data is subjected to operations based on denormalization to obtain candidate value prediction data corresponding to each classified product data set.

For example, taking the initial value prediction data A _t+1 corresponding to the classified product data set A (corresponding to the main category A, the main category A includes subcategory a1 and subcategory a2) as an example, the server obtains the initial value prediction data A _t+1 corresponding to The prediction reference index data y1 _t+1,l and the prediction reference index coefficient m1 _t+1,l of the subcategory a1 at the prediction time t+1, and the subcategory a2 corresponding to the initial value prediction data A _t+1 are predicted The forecast benchmark index data y2 _t+1,l and the forecast benchmark index coefficient m2 _t+1,l at time t+1, through the normalization model, based on the preset denormalization formula 1:

Among them, l and L represent the number of indicators of the forecast benchmark index data, and calculate the classified product for the initial value forecast data A _t+1 , the forecast benchmark index data y1 _t+1,l and the forecast benchmark index coefficient m1 _t+1,l The candidate value prediction data a1 _t+1 corresponding to the subcategory a1 of data set A, and through the normalization model, based on the preset denormalization formula 2:

Among them, l and L represent the number of indicators of the forecast benchmark index data, and calculate the classified product for the initial value forecast data A _t+1 , the forecast benchmark index data y2 _t+1,l and the forecast benchmark index coefficient m2 _t+1,l For the candidate value prediction data a2 _t+1 corresponding to the subcategory a2 of the data set A, the candidate value prediction data a1 _t+1 and the candidate value prediction data a2 _t+1 are determined as the candidate value prediction data corresponding to the classified product data set A.

205. Perform correlation fusion processing on the candidate value prediction data corresponding to each classified product data set to obtain target value prediction data.

Specifically, the server performs correlation analysis on the candidate value prediction data corresponding to each classified product data set to obtain the correlation coefficient corresponding to each classified product data set; through the corresponding correlation coefficient of each classified product data set, the corresponding The candidate value prediction data are weighted and summed to obtain the target value prediction data.

The server calls the preset correlation analysis algorithm to measure the structural relationship (correlation) between the candidate value prediction data corresponding to each classified product data set, and obtains the correlation coefficient corresponding to each classified product data set, and calculates the The correlation coefficient corresponding to the data set is determined as the weighting coefficient corresponding to each classified product data set, and the sum value of each classified product data set is calculated according to the weighted coefficient, so as to realize the correlation fusion of the candidate value prediction data corresponding to each classified product data set processing to obtain target value forecast data.

206. Obtain an error value based on the target value prediction data, match the target optimization strategy corresponding to the error value, and execute an optimization process based on the target optimization strategy.

Among them, the optimization strategy includes an optimization scheme for the normalization model, an optimization scheme for creating a structural vector autoregressive model, and an optimization scheme for the execution process corresponding to the target value prediction data. The server acquires real value data corresponding to the target value forecast data, and calculates an error value between the target value forecast data and the real value data. Through the comparison and analysis of the preset range value and the error value, the target range value corresponding to the error value is obtained, and the key value or structured query statement of the target range value is generated. Through the key value or structured query statement, the preset optimization is performed. The strategy hash table is retrieved to obtain the target optimization strategy. Based on the target optimization strategy, the normalized adjustment model parameters, normalized adjustment factors, structural vector autoregressive adjustment model parameters and structural vector Regression adjustment factor, recreate the normalized model by normalizing the adjustment model parameters and the normalization adjustment factor, recreate the structural vector autoregressive model by adjusting the model parameters and the structural vector autoregressive adjustment factor; receive the target terminal The execution script based on the target optimization strategy is sent, through which the process nodes, data processing methods, and execution programs of the execution process corresponding to the target value prediction data are adjusted and executed.

By optimizing the normalization model, the creation of the structural vector autoregressive model, and the execution process corresponding to the target value prediction data, the accuracy of the normalization model and the structural vector autoregressive model was improved, and the accuracy of the target value prediction data was improved. Accuracy, thereby improving the value prediction accuracy of product data.

In the embodiment of the present application, it is not only convenient to process the product data to be processed of different asset types in a targeted and effective manner, but also integrates various influencing factors of different asset types, and enriches the information of different types and changing product data. The value predictor improves the value prediction accuracy of different types and changing product data, realizes the calculation of the structural relationship between the standard asset price index (that is, the target index data), and estimates it based on the predicted initial value prediction data Changes in the value of assets under the user's name in the future, and make them meet risk measurement standards such as sub-additivity, exchangeability, and consistency; meet risk measurement standards such as additivity, exchangeability, and consistency, thereby improving product data. The accuracy of value prediction can also improve the accuracy of the normalization model and the structure vector autoregressive model by optimizing the creation of the normalization model, the structural vector autoregressive model, and the execution process corresponding to the target value prediction data. The accuracy of target value prediction data is improved, thereby improving the value prediction accuracy of product data.

The product data fusion method in the embodiment of the present application is described above, and the product data fusion device in the embodiment of the present application is described below. Please refer to FIG. 3. An embodiment of the product data fusion device in the embodiment of the present application includes:

A classification module 301, configured to obtain product data to be processed, and classify the product data to be processed based on asset categories to obtain multiple classified product data sets;

The normalized fusion module 302 is used to perform normalized fusion processing on multiple classified product data sets respectively through a preset normalized model to obtain normalized fusion data corresponding to each classified product data;

The prediction module 303 is used to obtain the target index data corresponding to each classified product data, and use the preset structural vector autoregressive model, the target index data corresponding to each classified product data, and the normalized fusion data corresponding to each classified product data. Value autoregressive prediction is performed on multiple classified product data sets, and the initial value prediction data corresponding to each classified product data set is obtained. The target index data includes economic index data and market index data;

The denormalization module 304 is used to denormalize the initial value prediction data corresponding to each classified product data set through a normalization model, and obtain candidate value prediction data corresponding to each classified product data set;

The correlation fusion module 305 is configured to perform correlation fusion processing on the candidate value prediction data corresponding to each classified product data set to obtain target value prediction data.

The function implementation of each module in the above product data fusion device corresponds to the steps in the above product data fusion method embodiment, and its functions and implementation processes will not be repeated here.

In the embodiment of the present application, by classifying the product data to be processed based on asset categories, it is convenient for subsequent targeted and effective processing of product data to be processed of different asset types; by separately classifying multiple classified product data sets Normalized fusion processing integrates multiple influencing factors of different asset types, enriches the predictive factors of the value of different types and changing product data, and improves the value prediction accuracy of different types and changing product data; through preset The structural vector autoregressive model, the target index data corresponding to each category of product data and the normalized fusion data corresponding to each category of product data, the value autoregressive prediction is performed on multiple category product data sets, and the value of each category product data set The corresponding initial value prediction data is denormalized to realize the calculation of the structural relationship between the standard asset price index (that is, the target index data), and to estimate the value of the assets under the user's name in the future based on the predicted initial value prediction data change, and make it meet the risk measurement standards such as additivity, commutativity and consistency; carry out correlation fusion processing on the candidate value prediction data corresponding to each classified product data set, so that it meets sub-additivity, commutativity and consistency Consistency and other risk measurement standards, thereby improving the value prediction accuracy of product data.

Please refer to Figure 4, another embodiment of the product data fusion device in the embodiment of the present application includes:

The normalization fusion module 302 is used to carry out normalization fusion processing to a plurality of classification product data sets respectively by the preset normalization model, and obtain the normalization fusion data corresponding to each classification product data;

A correlation fusion module 305, configured to perform correlation fusion processing on the candidate value prediction data corresponding to each classified product data set, to obtain target value prediction data;

The optimization execution module 306 is configured to obtain an error value based on the target value prediction data, match the target optimization strategy corresponding to the error value, and execute an optimization process based on the target optimization strategy.

Optionally, the normalized fusion module 302 can also be specifically used for:

Obtain asset price data, normalized benchmark index data, and normalized benchmark index coefficients corresponding to multiple classified product data sets respectively, and the normalized benchmark index data is used to indicate the benchmark based on preset Based on the price index, the index factor that affects the normalized fusion processing of asset price data;

Through the preset normalization model, based on the instant asset price data, normalized benchmark index data, and normalized benchmark index coefficients, the value data calculation based on normalization is performed on multiple classified product data sets respectively, and each classification is obtained. The normalized fusion data corresponding to the product dataset.

Optionally, the prediction module 303 can also be specifically used for:

Obtain the target index data, disturbance item data and independent variable coefficients corresponding to multiple classified product data sets respectively, and the target index data includes economic index data and market index data corresponding to multiple classified product data sets respectively;

Through the preset structural vector autoregressive model, the target index data, disturbance item data and independent variable coefficients corresponding to each classified product data, and the normalized fusion data corresponding to each classified product data, the multiple classified product data sets are separately processed The operation processing of structural vector autoregression obtains the initial value prediction data corresponding to each classified product data set.

Optionally, the denormalization module 304 can also be specifically used for:

Obtain the forecast benchmark index data of the initial value forecast data corresponding to each classified product data set, and the forecast benchmark index coefficient of the forecast benchmark index data. The forecast benchmark index data is used to indicate the impact on asset prices based on the preset benchmark price index Index factor for data denormalization processing;

Through the normalization model, based on the forecast benchmark index data and forecast benchmark index coefficients, the preset denormalization formula and the initial value forecast data corresponding to each category product data set, the calculation based on denormalization is performed to obtain the product data of each category The candidate value prediction data corresponding to the set.

Optionally, the correlation fusion module 305 can also be specifically used for:

Correlation analysis is performed on the candidate value prediction data corresponding to each classified product data set, and the correlation coefficient corresponding to each classified product data set is obtained;

Through the correlation coefficients corresponding to each classified product data set, weighted summation is performed on the candidate value prediction data corresponding to each classified product data set to obtain target value prediction data.

Optionally, the classification module 301 can also be specifically used for:

Receive the asset valuation request sent by the target terminal, read the product data to be processed from the preset database based on the asset valuation request, and perform data preprocessing on the product data to be processed to obtain preprocessed product data;

Through the preset classification model, multi-level convolution-based feature extraction, attention mechanism-based feature fusion, asset category probability calculation, and asset category classification are sequentially performed on the preprocessed product data to obtain multiple classified product datasets.

The function implementation of each module and each unit in the above product data fusion device corresponds to the steps in the above product data fusion method embodiment, and its functions and implementation processes will not be repeated here.

The above Figures 3 and 4 describe in detail the product data fusion device in the embodiment of the present application from the perspective of modular functional entities, and the following describes the product data fusion device in the embodiment of the present application in detail from the perspective of hardware processing.

FIG. 5 is a schematic structural diagram of a product data fusion device provided by an embodiment of the present application. The product data fusion device 500 may have relatively large differences due to different configurations or performances, and may include one or more than one processor (central processing units (CPU) 510 (for example, one or more processors) and memory 520, one or more storage media 530 for storing application programs 533 or data 532 (for example, one or more mass storage devices). Wherein, the memory 520 and the storage medium 530 may be temporary storage or persistent storage. The program stored in the storage medium 530 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the product data fusion device 500 . Further, the processor 510 may be configured to communicate with the storage medium 530, and execute a series of instruction operations in the storage medium 530 on the product data fusion device 500.

The product data fusion device 500 may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input and output interfaces 560, and/or, one or more operating systems 531, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art can understand that the product data fusion device structure shown in FIG. 5 does not constitute a limitation to the product data fusion device, and may include more or less components than those shown in the illustration, or combine certain components, or Different component arrangements.

The present application also provides a fusion device for product data, including: a memory and at least one processor, instructions are stored in the memory, and the memory and at least one processor are interconnected through lines; at least one processor invokes the instructions in the memory, so that The product data fusion device executes the steps in the product data fusion method described above.

The present application also provides a computer-readable storage medium, the computer-readable storage medium may be a non-volatile computer-readable storage medium, the computer-readable storage medium may also be a volatile computer-readable storage medium, and the computer-readable storage medium may be Instructions are stored in the read storage medium, and when the instructions are run on the computer, the computer is made to execute the steps of the product data fusion method.

Further, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required by at least one function, etc.; Use the created data etc.

The blockchain referred to in this application is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain (Blockchain), essentially a decentralized database, is a series of data blocks associated with each other using cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify its Validity of information (anti-counterfeiting) and generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

Those skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

If an integrated unit is realized in the form of a software function unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or part of the contribution to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods in various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (read-only memory, ROM), random access memory (random access memory, RAM), magnetic disk or optical disc and other media that can store program codes. .

Above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be applied to the foregoing embodiments The technical solutions described in the examples are modified, or some of the technical features are equivalently replaced; 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 various embodiments of the application.

Claims

A method for merging product data, wherein the method for merging product data includes:

Obtaining product data to be processed, and classifying the product data to be processed based on asset categories to obtain multiple classified product data sets;

Perform normalized fusion processing on the plurality of classified product data sets respectively through a preset normalized model, to obtain normalized fusion data corresponding to each classified product data;

Obtain the target index data corresponding to each classified product data, and use the preset structural vector autoregressive model, the target index data corresponding to each classified product data, and the normalized fusion data corresponding to each classified product data, to The plurality of classified product data sets are respectively subjected to auto-regression value prediction to obtain initial value forecast data corresponding to each classified product data set, and the target index data includes economic index data and market index data;

Using the normalization model, denormalize the initial value prediction data corresponding to each classified product data set to obtain candidate value prediction data corresponding to each classified product data set;

Correlation fusion processing is performed on the candidate value prediction data corresponding to each classified product data set to obtain target value prediction data.
The method for merging product data according to claim 1, wherein said normalized and fused processing is performed on said plurality of classified product data sets through the preset normalization model to obtain the corresponding data of each classified product data. Normalized fusion data, including:

Obtain asset price data, normalized benchmark index data, and normalized benchmark index coefficients of the normalized benchmark index data respectively corresponding to the plurality of classified product data sets, and the normalized benchmark index data is used for Indicates the index factor that affects the normalized fusion processing of asset price data based on the preset benchmark price index;

Through the preset normalization model, based on the asset price data at the time, the normalized benchmark index data and the normalized benchmark index coefficient, the multiple classified product data sets are respectively normalized Value data calculations to obtain the normalized fusion data corresponding to each classified product data set.
The product data fusion method according to claim 1, wherein said acquisition of target index data corresponding to each classified product data, through a preset structural vector autoregressive model, said target index data corresponding to each classified product data Data and the normalized fusion data corresponding to the data of each classified product, respectively perform value autoregressive prediction on the multiple classified product data sets, and obtain the initial value prediction data corresponding to each classified product data set, and the target index data Includes economic indicator data and market indicator data, including:

Obtain target index data, disturbance item data, and independent variable coefficients respectively corresponding to the plurality of classified product data sets, wherein the target index data includes economic index data and market index data respectively corresponding to the plurality of classified product data sets;

Through the preset structural vector autoregressive model, the target index data, disturbance item data and independent variable coefficients corresponding to the various classified product data, and the normalized fusion data corresponding to the various classified product data, the multiple The classification product data sets are respectively processed by structural vector autoregression to obtain the initial value prediction data corresponding to each classification product data set.
The method for merging product data according to claim 1, wherein said normalization model is used to denormalize the initial value prediction data corresponding to each classified product data set to obtain each classified product Candidate value prediction data corresponding to the data set, including:

Acquiring the forecast benchmark index data of the initial value forecast data corresponding to each classified product data set, and the forecast benchmark index coefficient of the forecast benchmark index data, the forecast benchmark index data is used to indicate the value based on the preset benchmark price index Based on the index factors that affect the denormalization processing of asset price data;

Through the normalization model, based on the predictive benchmark index data and the predictive benchmark index coefficient, the preset denormalization formula and the initial value forecast data corresponding to each classified product data set, denormalization based operation to obtain the candidate value prediction data corresponding to each classified product data set.
The product data fusion method according to claim 1, wherein the correlation fusion processing is performed on the candidate value prediction data corresponding to each classified product data set to obtain the target value prediction data, comprising:

Carrying out correlation analysis on the candidate value prediction data corresponding to each classified product data set, to obtain a correlation coefficient corresponding to each classified product data set;

Based on the correlation coefficients corresponding to the classified product data sets, the candidate value prediction data corresponding to the classified product data sets are weighted and summed to obtain the target value prediction data.
The fusion method of product data according to claim 1, wherein said acquiring the product data to be processed, and classifying said product data to be processed based on asset category to obtain multiple classified product data sets, comprising:

Receive the asset valuation request sent by the target terminal, read the product data to be processed from the preset database based on the asset valuation request, and perform data preprocessing on the product data to be processed to obtain preprocessed product data ;

Through the preset classification model, the preprocessed product data is sequentially subjected to multi-level convolution-based feature extraction, attention mechanism-based feature fusion, asset category probability calculation, and asset category classification to obtain multiple classified product data sets.
The product data fusion method according to any one of claims 1-6, wherein the correlation fusion processing is performed on the candidate value prediction data corresponding to each classified product data set, and after the target value prediction data is obtained, Also includes:

Obtaining an error value based on the target value prediction data, matching a target optimization strategy corresponding to the error value, and executing an optimization process based on the target optimization strategy.
A product data fusion device, comprising a memory, a processor, and computer-readable instructions stored on the memory and operable on the processor, and the processor implements the following steps when executing the computer-readable instructions :

Obtaining product data to be processed, and classifying the product data to be processed based on asset categories to obtain multiple classified product data sets;

Perform normalized fusion processing on the plurality of classified product data sets respectively through a preset normalized model, to obtain normalized fusion data corresponding to each classified product data;

Obtain the target index data corresponding to each classified product data, and use the preset structural vector autoregressive model, the target index data corresponding to each classified product data, and the normalized fusion data corresponding to each classified product data, to The plurality of classified product data sets are respectively subjected to auto-regression value prediction to obtain initial value forecast data corresponding to each classified product data set, and the target index data includes economic index data and market index data;

Using the normalization model, denormalize the initial value prediction data corresponding to each classified product data set to obtain candidate value prediction data corresponding to each classified product data set;

Correlation fusion processing is performed on the candidate value prediction data corresponding to each classified product data set to obtain target value prediction data.
The product data fusion device according to claim 8, wherein, the preset normalization model is used to perform normalization and fusion processing on the plurality of classified product data sets respectively, so as to obtain the data corresponding to each classified product data. Normalized fusion data, including:

Obtain asset price data, normalized benchmark index data, and normalized benchmark index coefficients of the normalized benchmark index data respectively corresponding to the plurality of classified product data sets, and the normalized benchmark index data is used for Indicates the index factor that affects the normalized fusion processing of asset price data based on the preset benchmark price index;

Through the preset normalization model, based on the asset price data at the time, the normalized benchmark index data and the normalized benchmark index coefficient, the multiple classified product data sets are respectively normalized Value data calculations to obtain the normalized fusion data corresponding to each classified product data set.
The product data fusion device according to claim 8, wherein said acquisition of the target index data corresponding to each of the classified product data is carried out through a preset structural vector autoregressive model, and the target index corresponding to each of the classified product data Data and the normalized fusion data corresponding to the data of each classified product, respectively perform value autoregressive prediction on the multiple classified product data sets, and obtain the initial value prediction data corresponding to each classified product data set, and the target index data Includes economic indicator data and market indicator data, including:

Obtain target index data, disturbance item data, and independent variable coefficients respectively corresponding to the plurality of classified product data sets, wherein the target index data includes economic index data and market index data respectively corresponding to the plurality of classified product data sets;

Through the preset structural vector autoregressive model, the target index data, disturbance item data and independent variable coefficients corresponding to the various classified product data, and the normalized fusion data corresponding to the various classified product data, the multiple The classification product data sets are respectively processed by structural vector autoregression to obtain the initial value prediction data corresponding to each classification product data set.
The product data fusion device according to claim 8, wherein, through the normalization model, denormalization is performed on the initial value prediction data corresponding to each classified product data set to obtain each classified product Candidate value prediction data corresponding to the data set, including:

Acquiring the forecast benchmark index data of the initial value forecast data corresponding to each classified product data set, and the forecast benchmark index coefficient of the forecast benchmark index data, the forecast benchmark index data is used to indicate the value based on the preset benchmark price index Based on the index factors that affect the denormalization processing of asset price data;

Through the normalization model, based on the predictive benchmark index data and the predictive benchmark index coefficient, the preset denormalization formula and the initial value forecast data corresponding to each classified product data set, denormalization based operation to obtain the candidate value prediction data corresponding to each classified product data set.
The product data fusion device according to claim 8, wherein the correlation fusion processing is performed on the candidate value prediction data corresponding to each classified product data set to obtain the target value prediction data, comprising:

Carrying out correlation analysis on the candidate value prediction data corresponding to each classified product data set, to obtain a correlation coefficient corresponding to each classified product data set;

Based on the correlation coefficients corresponding to the classified product data sets, the candidate value prediction data corresponding to the classified product data sets are weighted and summed to obtain the target value prediction data.
The product data fusion device according to claim 8, wherein said obtaining the product data to be processed, and classifying the product data to be processed based on asset categories, to obtain multiple classified product data sets, including:

Receive the asset valuation request sent by the target terminal, read the product data to be processed from the preset database based on the asset valuation request, and perform data preprocessing on the product data to be processed to obtain preprocessed product data ;

Through the preset classification model, the preprocessed product data is sequentially subjected to multi-level convolution-based feature extraction, attention mechanism-based feature fusion, asset category probability calculation, and asset category classification to obtain multiple classified product data sets.
The product data fusion device according to any one of claims 8-13, wherein, after performing correlation fusion processing on the candidate value prediction data corresponding to each classified product data set, and obtaining the target value prediction data, Also includes:

Obtaining an error value based on the target value prediction data, matching a target optimization strategy corresponding to the error value, and executing an optimization process based on the target optimization strategy.
A computer-readable storage medium, wherein computer instructions are stored in the computer-readable storage medium, and when the computer instructions are run on the computer, the computer is made to perform the following steps:

Obtaining product data to be processed, and classifying the product data to be processed based on asset categories to obtain multiple classified product data sets;

Perform normalized fusion processing on the plurality of classified product data sets respectively through a preset normalized model, to obtain normalized fusion data corresponding to each classified product data;

Obtain the target index data corresponding to each classified product data, and use the preset structural vector autoregressive model, the target index data corresponding to each classified product data, and the normalized fusion data corresponding to each classified product data, to The plurality of classified product data sets are respectively subjected to auto-regression value prediction to obtain initial value forecast data corresponding to each classified product data set, and the target index data includes economic index data and market index data;

Using the normalization model, denormalize the initial value prediction data corresponding to each classified product data set to obtain candidate value prediction data corresponding to each classified product data set;

Correlation fusion processing is performed on the candidate value prediction data corresponding to each classified product data set to obtain target value prediction data.
The computer-readable storage medium according to claim 15, wherein, the preset normalization model is used to perform normalization and fusion processing on the plurality of classified product data sets respectively to obtain the corresponding data of each classified product data. Normalized fusion data, including:

Obtain asset price data, normalized benchmark index data, and normalized benchmark index coefficients of the normalized benchmark index data respectively corresponding to the plurality of classified product data sets, and the normalized benchmark index data is used for Indicates the index factor that affects the normalized fusion processing of asset price data based on the preset benchmark price index;

Through the preset normalization model, based on the asset price data at the time, the normalized benchmark index data and the normalized benchmark index coefficient, the multiple classified product data sets are respectively normalized Value data calculations to obtain the normalized fusion data corresponding to each classified product data set.
The computer-readable storage medium according to claim 15, wherein the acquisition of the target index data corresponding to each classified product data is carried out through a preset structural vector autoregressive model, and the target index data corresponding to each classified product data Data and the normalized fusion data corresponding to the data of each classified product, respectively perform value autoregressive prediction on the multiple classified product data sets, and obtain the initial value prediction data corresponding to each classified product data set, and the target index data Includes economic indicator data and market indicator data, including:

Obtain target index data, disturbance item data, and independent variable coefficients respectively corresponding to the plurality of classified product data sets, wherein the target index data includes economic index data and market index data respectively corresponding to the plurality of classified product data sets;

Through the preset structural vector autoregressive model, the target index data, disturbance item data and independent variable coefficients corresponding to the various classified product data, and the normalized fusion data corresponding to the various classified product data, the multiple The classification product data sets are respectively processed by structural vector autoregression to obtain the initial value prediction data corresponding to each classification product data set.
The computer-readable storage medium according to claim 15, wherein, through the normalization model, the initial value prediction data corresponding to each classified product data set is subjected to denormalization processing to obtain each classified product Candidate value prediction data corresponding to the data set, including:

Acquiring the forecast benchmark index data of the initial value forecast data corresponding to each classified product data set, and the forecast benchmark index coefficient of the forecast benchmark index data, the forecast benchmark index data is used to indicate the value based on the preset benchmark price index Based on the index factors that affect the denormalization processing of asset price data;

Through the normalization model, based on the predictive benchmark index data and the predictive benchmark index coefficient, the preset denormalization formula and the initial value forecast data corresponding to each classified product data set, denormalization based operation to obtain the candidate value prediction data corresponding to each classified product data set.
The computer-readable storage medium according to claim 15, wherein said performing correlation fusion processing on the candidate value prediction data corresponding to each classified product data set to obtain the target value prediction data comprises:

Carrying out correlation analysis on the candidate value prediction data corresponding to each classified product data set, to obtain a correlation coefficient corresponding to each classified product data set;

Based on the correlation coefficients corresponding to the classified product data sets, the candidate value prediction data corresponding to the classified product data sets are weighted and summed to obtain the target value prediction data.
A product data fusion device, wherein the product data fusion device includes:

A classification module, configured to obtain product data to be processed, and classify the product data to be processed based on asset categories to obtain multiple classified product data sets;

The normalized fusion module is used to perform normalized fusion processing on the multiple classified product data sets respectively through a preset normalized model, so as to obtain normalized fusion data corresponding to each classified product data;

The prediction module is used to obtain the target index data corresponding to each classified product data, and use the preset structural vector autoregressive model, the target index data corresponding to each classified product data and the normalization corresponding to each classified product data Combining the data, performing value autoregressive prediction on the plurality of classified product data sets respectively, and obtaining initial value forecast data corresponding to each classified product data set, the target index data including economic index data and market index data;

The denormalization module is used to perform denormalization processing on the initial value prediction data corresponding to each classified product data set through the normalization model, so as to obtain candidate value prediction data corresponding to each classified product data set;

The correlation fusion module is used to perform correlation fusion processing on the candidate value prediction data corresponding to each classified product data set to obtain target value prediction data.