Method, device, electronic device and storage medium for predicting related business

Abstract

The present application discloses a method, device, electronic device and storage medium for predicting associated business, and relates to the field of data processing technology, wherein the method comprises: obtaining a weighted coefficient matrix, wherein the weighted coefficient matrix is ​​a matrix generated by weighted operation of the Kendall correlation matrix and the Spearman correlation matrix; the multidimensional dependency model is a model tested by the VaR method; monitoring the sales data of each financial product in the weighted coefficient matrix, and taking the monitored financial products that trigger preset sales conditions as target financial products; obtaining financial products whose weighted coefficients meet preset conditions with the target financial products from the weighted coefficient matrix as associated financial products; and using the multidimensional dependency model to predict sales of the associated financial products and the target financial products. Thus, the correlation between financial products is fully excavated, and marketing workers are promptly reminded to increase the promotion of associated financial products.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method for predicting associated business, a device for predicting associated business, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] As a key performance focus of financial institutions, financial sales and market expansion are one of the most important daily tasks for financial institution employees. The market sales performance of different financial products varies greatly, but there is often a certain correlation between the differences. If this correlation can be found, when a certain financial product has good sales, the promotion of related financial products can be increased to explore business opportunities and achieve good total sales. When a certain financial product fluctuates in the product market, it can more accurately grasp the changing rules of each product related to it and carry out targeted guidance for marketing work. Summary of the invention

[0003] The present application provides a method, device, electronic device and storage medium for predicting associated business, so as to explore the association between various financial products in a financial institution.

[0004] In a first aspect, an embodiment of the present application provides a method for associating business prediction, the method comprising:

[0005] Obtain a weighted coefficient matrix, wherein the weighted coefficient matrix is ​​a matrix generated by weighted operation of the Kendall correlation matrix and the Spearman correlation matrix, wherein the Kendall correlation matrix is ​​a matrix generated by using a pre-trained multidimensional dependency model and used to record the Kendall correlation coefficients of two financial products, and the Spearman correlation matrix is ​​a matrix generated by using the multidimensional dependency model and used to record the Spearman correlation coefficients of two financial products; the multidimensional dependency model is a model tested by the Value at Risk (VaR) method;

[0006] Monitor the sales data of each financial product in the weighted coefficient matrix, and use the monitored financial products that trigger preset sales conditions as target financial products;

[0007] From the weighting coefficient matrix, a financial product whose weighting coefficient meets a preset condition with that of the target financial product is obtained as an associated financial product;

[0008] The multi-dimensional dependency model is used to perform sales forecasts on the associated financial products and the target financial products, and the sales forecast results are used to assist relevant personnel in determining whether to increase the promotion efforts for the associated financial products.

[0009] In a second aspect, an embodiment of the present application further provides a device for associated business prediction, the device comprising:

[0010] A weighted coefficient matrix acquisition module, used to acquire a weighted coefficient matrix, wherein the weighted coefficient matrix is ​​a matrix generated by weighted operation of a Kendall correlation matrix and a Spearman correlation matrix, wherein the Kendall correlation matrix is ​​a matrix generated by using a pre-trained multidimensional dependency model and used to record the Kendall correlation coefficients of two financial products, and the Spearman correlation matrix is ​​a matrix generated by using the multidimensional dependency model and used to record the Spearman correlation coefficients of two financial products; the multidimensional dependency model is a model tested by the Value at Risk (VaR) method;

[0011] A target financial product detection module, used to monitor the sales data of each financial product in the weighted coefficient matrix, and use the monitored financial products that trigger preset sales conditions as target financial products;

[0012] An associated financial product determination module, configured to obtain, from the weighting coefficient matrix, a financial product whose weighting coefficient meets a preset condition with that of the target financial product as an associated financial product;

[0013] The sales forecasting module is used to use the multi-dimensional dependency model to make sales forecasts for the associated financial products and the target financial products. The sales forecast results are used to assist relevant personnel in determining whether to increase the promotion efforts for the associated financial products.

[0014] In a third aspect, an embodiment of the present application further provides an electronic device, the electronic device comprising:

[0015] one or more processors;

[0016] a storage device for storing one or more programs,

[0017] When the one or more programs are executed by the one or more processors, the one or more processors implement the method of the first aspect described above.

[0018] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the method of the first aspect when executed by a processor.

[0019] In a fifth aspect, an embodiment of the present application further provides a computer program product, wherein the computer program product comprises computer executable instructions, and the computer executable instructions are used to implement the method of the first aspect when executed.

[0020] The technical solution provided by this application has the following beneficial effects:

[0021] In this embodiment, after obtaining the weighted coefficient matrix obtained by weighted operation of the Kendall correlation matrix and the Spearman correlation matrix, the sales data of each financial product in the weighted coefficient matrix is ​​monitored, and the monitored financial product that triggers the preset sales volume condition is used as the target financial product. After the target financial product is monitored, the associated financial products whose weighted coefficients meet the preset conditions with the target financial product can be mined from the weighted coefficient matrix, and the sales of the associated financial products are predicted using a multidimensional dependency model. In this way, the correlation between financial products is fully mined. When the sales volume of the target financial product increases significantly, it is determined that the sales volume of its associated financial products will also increase, so that marketing workers can be reminded in time to increase the recommendation of associated financial products, forming a good business opportunity recommendation mechanism. At the same time, the sales volume of each associated financial product in the future is predicted based on the multidimensional dependency model, which is convenient for evaluating the future market prosperity. A more comprehensive and accurate evaluation of the demand status of the financial market is achieved, and based on this, the sales promotion intensity and combination matching method of the overall financial product system of the financial institution are adjusted to increase the sales success rate.

[0022] In addition, compared with the linear correlation coefficient, the two rank correlation coefficients, Kendall correlation coefficient and Spearman correlation coefficient, can more accurately and comprehensively reflect the intrinsic correlation between different financial products from the perspective of probability distribution. This correlation can be used to more accurately determine the related financial products, thereby improving the accuracy of prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flowchart of a method embodiment of an associated business prediction provided in Embodiment 1 of the present application;

[0024] Figure 2 This is a flow chart of an embodiment of a method for predicting associated services provided in Embodiment 2 of the present application;

[0025] Figure 3 This is a flowchart of an embodiment of a method for predicting associated services provided in Embodiment 3 of the present application;

[0026] Figure 4 This is a structural block diagram of an embodiment of a device for predicting associated services provided in Embodiment 4 of the present application;

[0027] Figure 5 It is a structural schematic diagram of an electronic device provided in Example 5 of the present application. DETAILED DESCRIPTION

[0028] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only the parts related to the present application, rather than all structures, are shown in the accompanying drawings.

[0029] It should also be noted that the acquisition, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations.

[0030] Embodiment 1

[0031] Figure 1 This is a flow chart of a method embodiment of an associated business prediction provided in the first embodiment of the present application. This embodiment can be executed by a business processing system of a financial institution, such as Figure 1 As shown, the following steps may be included:

[0032] Step 110, obtaining a weighted coefficient matrix, wherein the weighted coefficient matrix is ​​a matrix generated by performing a weighted operation on the Kendall correlation matrix and the Spearman correlation matrix.

[0033] Among them, the Kendall correlation matrix is ​​a matrix generated by a pre-trained multidimensional dependency model and used to record the Kendall correlation coefficient (Kendall's tau) of two financial products; the Spearman correlation matrix is ​​a matrix generated by the multidimensional dependency model and used to record the Spearman correlation coefficient of two financial products. The Kendall's tau correlation coefficient and the spearman correlation coefficient are both rank correlation coefficients. The spearman correlation coefficient uses the rank size of two variables (equivalent to two financial products in this embodiment) to perform linear correlation analysis, and does not require the distribution of the original variables. It belongs to a non-parametric statistical method. The Kendall's tau correlation coefficient is used to reflect the index of the correlation of categorical variables. It is applicable to the case where both variables are ordered classifications. Specifically, the Kendall's tau correlation coefficient is determined by comparing the size sorting difference between the two sequence data to determine whether the increase or decrease between the two is synchronized, thereby determining the size of the correlation between the two sequences.

[0034] After obtaining the Kendall correlation matrix and the Spearman correlation matrix, this embodiment performs weighted summation of the Kendall's tau correlation coefficient and the Spearman correlation coefficient between two components (i.e., two financial products), and obtains the weighted coefficients of the two components, and finally forms a more comprehensive weighted coefficient matrix. The rows and columns of the weighted coefficient matrix correspond to each component, that is, each financial product of the financial institution, and the elements in the matrix can represent the weighted coefficients between the financial products represented by the row and column where the element is located. Exemplarily, the above-mentioned financial institutions may include banks, and the financial products may include the bank's financial products, deposit services, etc.

[0035] It should be noted that in the implementation of the above weighted summation, the weights of the Kendall's tau correlation coefficient and the spearman correlation coefficient are not limited, and those skilled in the art can configure the weights of the two according to actual needs. For example, since Kendall's tau is more functional, the weight of the Kendall's tau correlation coefficient can be set higher and the spearman correlation coefficient can be set relatively lower in the weighted summation process.

[0036] In one example, the multidimensional dependency model can preferably be a copula model, which is a model that connects the joint distribution function of multiple random variables with their respective marginal distribution functions. In one example, the definition of the copula model can be expressed as follows:

[0037] Assume C(u) = C(u1, ..., u d ) is defined in [0,1] d A copula on , if it satisfies the following three conditions:

[0038] (1) C(u) = C(u1, ..., u d ) For each marginal distribution function u i is monotonically increasing;

[0039] (2)C(1,...,u i ,...,1) for any i=1,...,d,u i ∈[0,1] holds;

[0040] d) Satisfy a i ≤ b i , both have

[0041]

[0042] where for any j∈[1,...,d], there is u j，1 =a j and uj，2 =b j .

[0043] According to Sklar's theorem, such C is uniquely determined in a certain sense.

[0044] It should be noted that there are many types of Copula, and various data types can be used. According to the number of parameters of the copula function, it can be divided into simple, medium, and complex types. The typical representatives are Gaussian copula, D-vine, and R-vine. When choosing, you can adopt the following strategies: when the data dimension is huge, you can use a simple copula to avoid the model being too complex and difficult to calculate; when the data dimension is small, you can use a complex copula to improve the accuracy and applicability of the model; if you don’t know how to choose, you can choose a medium-complexity copula, which can basically handle any situation.

[0045] In this embodiment, the multidimensional dependency model is a model tested by the value at risk (VaR) method. Using the VaR method to test the model can improve the prediction effect of the model.

[0046] According to the correlation parameters in the multidimensional dependency model, a corresponding correlation matrix can be generated. Specifically, the Kendall's tau correlation parameter can be extracted from the multidimensional dependency model, and the Kendall's tau correlation coefficient between the two components can be calculated using the Kendall's tau correlation parameter, thereby constructing the Kendall'stau correlation matrix. The spearman correlation parameter can also be extracted from the multidimensional dependency model, and the spearman correlation coefficient between the two components can be calculated using the spearman correlation parameter, thereby constructing the spearman correlation matrix.

[0047] The spearman correlation coefficient and Kendall's tau correlation coefficient selected in this embodiment are a method for determining the size of the correlation between two sequences by comparing the size order differences between the sequence data to determine whether the increase and decrease amplitudes between the two are synchronized. Unlike the Pearson correlation coefficient, which can only measure linear correlation, the statistical quantity converges faster, and has better adaptability and flexibility.

[0048] Step 120, monitoring the sales data of each financial product in the weighted coefficient matrix, and taking the monitored financial products that trigger the preset sales conditions as target financial products.

[0049] This embodiment can monitor the sales data of each financial product recorded in the weighted coefficient matrix. If the sales data of a certain financial product is monitored to trigger a preset sales condition, the financial product can be determined as a target financial product.

[0050] The preset sales volume condition may be a positive condition, for example, a condition set to identify financial products with rapidly growing sales volume. If the sales volume of a certain financial product in the past period (the sales volume is equal to the sum of the sales data in the past period) meets the positive condition (such as a certain growth threshold), the financial product can be determined as a target financial product. Alternatively, the preset sales volume condition may also be a negative condition, for example, a condition set to identify financial products with low sales volume. If the sales volume of a certain financial product in the past period meets the negative condition (such as a certain negative threshold), the financial product can be determined as a target financial product, and such financial products need to be warned in advance.

[0051] Step 130: From the weighting coefficient matrix, obtain a financial product whose weighting coefficient meets a preset condition with that of the target financial product as an associated financial product.

[0052] After the target financial product is monitored, the financial products whose weighting coefficients meet preset conditions with the target financial product can be obtained as associated financial products according to the weighting coefficients of the pairwise financial products recorded in the weighting coefficient matrix.

[0053] In one embodiment, step 130 may further include the following steps:

[0054] From the weighting coefficient matrix, the target weighting coefficients of the target financial product and other financial products are obtained; financial products whose target weighting coefficients are greater than a preset coefficient threshold are taken as associated financial products; if there is no financial product whose target weighting coefficient is greater than the preset coefficient threshold, the first several financial products with the largest target weighting coefficients are taken as associated financial products.

[0055] In this embodiment, after the target financial product is determined, the identifier of the target financial product can be searched in the weighting coefficient matrix, so as to find the weighting coefficient related to the target financial product as the target weighting coefficient. Then the target weighting coefficient is judged. In one embodiment, a coefficient threshold (for example, 0.5) can be set. If a certain target weighting coefficient is greater than the coefficient threshold (i.e., it meets the preset conditions), the financial products other than the target financial product in the financial products corresponding to the target weighting coefficient (the target weighting coefficient is determined by the components of the row and column, one of which is the target financial product) can be taken as the associated financial products. If all the target weighting coefficients are not greater than the coefficient threshold, the first several financial products with the largest target weighting coefficients can be taken as the associated financial products, for example, the financial products corresponding to the first three largest target weighting coefficients can be taken as the associated financial products.

[0056] Step 140: Use the multi-dimensional dependency model to perform sales forecasting on the associated financial products and the target financial product.

[0057] After determining the associated financial products that have some associated attributes with the current target financial product, these key financial products can be used as key monitoring objects. In this step, a multidimensional dependency model can be used to predict the sales of the above-mentioned associated financial products. The results of the sales prediction are used to assist relevant personnel in determining whether to increase the promotion of the associated financial products, so as to formulate marketing plans in advance.

[0058] Of course, in addition to estimating the sales of each related financial product, you can also estimate the sales of the target financial product.

[0059] After determining the associated financial products that have some associated attributes with the current target financial product, associated prompt information can also be sent to relevant personnel so that the relevant personnel can grasp the association between the financial products in a timely manner.

[0060] In this embodiment, after obtaining the weighted coefficient matrix obtained by weighted operation of the Kendall correlation matrix and the Spearman correlation matrix, the sales data of each financial product in the weighted coefficient matrix is ​​monitored, and the monitored financial product that triggers the preset sales volume condition is used as the target financial product. After the target financial product is monitored, the associated financial products whose weighted coefficients meet the preset conditions with the target financial product can be mined from the weighted coefficient matrix, and the sales of the associated financial products are predicted using a multidimensional dependency model. In this way, the correlation between financial products is fully mined. When the sales volume of the target financial product increases significantly, it is determined that the sales volume of its associated financial products will also increase, so that marketing workers can be reminded in time to increase the recommendation of associated financial products, forming a good business opportunity recommendation mechanism. At the same time, the sales volume of each associated financial product in the future is predicted based on the multidimensional dependency model, which is convenient for evaluating the future market prosperity. A more comprehensive and accurate evaluation of the demand status of the financial market is achieved, and based on this, the sales promotion intensity and combination matching method of the overall financial product system of the financial institution are adjusted to increase the sales success rate.

[0061] In addition, compared with the linear correlation coefficient, the two rank correlation coefficients, Kendall correlation coefficient and Spearman correlation coefficient, can more accurately and comprehensively reflect the intrinsic correlation between different financial products from the perspective of probability distribution. This correlation can be used to more accurately determine the related financial products, thereby improving the accuracy of prediction.

[0062] Embodiment 2

[0063] Figure 2 This is a flow chart of a method embodiment of an associated business prediction provided in Embodiment 2 of the present application. Based on Embodiment 1, this embodiment expands the application of the multi-dimensional dependency model and may include the following steps:

[0064] Step 210, obtaining a weighted coefficient matrix, wherein the weighted coefficient matrix is ​​a matrix generated by performing a weighted operation on the Kendall correlation matrix and the Spearman correlation matrix.

[0065] Among them, the Kendall correlation matrix is ​​a matrix generated by a pre-trained multidimensional dependency model and used to record the Kendall correlation coefficients of two financial products; the Spearman correlation matrix is ​​a matrix generated by the multidimensional dependency model and used to record the Spearman correlation coefficients of two financial products; the multidimensional dependency model is a model tested by the value at risk VaR method.

[0066] Exemplarily, the multidimensional dependency model may preferably be a copula model.

[0067] Step 220, monitoring the sales data of each financial product in the weighted coefficient matrix, and taking the monitored financial products that trigger the preset sales conditions as target financial products.

[0068] Step 230: From the weighting coefficient matrix, obtain a financial product whose weighting coefficient meets a preset condition with that of the target financial product as an associated financial product.

[0069] Step 240, based on a preset number of predictions, when executing each prediction, the multi-dimensional dependency model is used to respectively predict the sales data forecast values ​​of each associated financial product and the target financial product for each day in a preset time period in the future.

[0070] Step 250, based on the number of predictions and the sales data forecast value of each prediction, calculate the average sales forecast value of each associated financial product and the target financial product per day.

[0071] Step 260, obtaining the daily sales forecast average of each financial product of the financial institution, and calculating the sum of the daily sales forecast average of each financial product to obtain the daily sales forecast total of the financial institution.

[0072] In practice, in addition to predicting the sales data of the target financial product and its related financial products, the multidimensional dependency model can actually predict the future sales data of each financial product in the financial institution. In order to improve the accuracy, the number of predictions can be set, and the average of multiple predictions can be taken as the final prediction result of the sales data of the financial product on that day. Furthermore, the average sales forecast of each financial product on each day can be summed up to obtain the total sales forecast of the financial institution on that day.

[0073] For example, it is necessary to predict the sales volume of each financial product in the next 90 days, and the prediction is made 1000 times a day. Assuming there are 100 financial products, the prediction result each time is 100 dimensions (that is, the predicted amount of 100 financial products is obtained each time, and this prediction result is a 100-dimensional vector, and each component of the vector corresponds to the sales data prediction value of each financial product at that time). For 1000 predictions a day, the average sales volume forecast of each financial product can be calculated first, and then the daily sales forecast average of the 100 financial products can be summarized to obtain the total sales volume forecast of the financial institution on that day. The total sales volume forecast of the financial institution for 90 days is recorded, and the sales volume change trend for the next 90 days can be obtained. In other embodiments, the total sales volume forecast of 90 days can also be summarized to obtain the total sales volume forecast of the financial institution for the next 90 days. The above-mentioned predicted sales information can be used as guidance information for the financial institution to adjust financial products from macro to micro operations.

[0074] The total forecast obtained by averaging and summing up the results of multiple forecasts of each financial product will have higher accuracy than directly calculating the historical total value for linear regression. One reason is that linear regression does not take into account the linkage between various financial products, making the model not sophisticated enough, and the premise assumption of traditional regression is too strong, which will greatly interfere with the accuracy. This embodiment introduces the multi-dimensional dependency structure of joint probability, which introduces more data internal structure information than the method of directly predicting the total number, thereby achieving the effect of reducing variance and improving prediction accuracy; secondly, the effect of summing up the forecast components is equivalent to an average, which can effectively reduce volatility. Specifically, the variance will be reduced, thereby compressing the confidence interval and improving accuracy.

[0075] In one embodiment, this embodiment may further include the following steps:

[0076] If the sales forecast average value is lower than the first set threshold, or the sales forecast total value is lower than the second set threshold, a first sales evaluation notice is issued to the relevant management and operation personnel; or, if the sales forecast average value is higher than the third set threshold, or the sales forecast total value is higher than the fourth set threshold, a second sales evaluation notice is issued to the relevant management and operation personnel.

[0077] Among them, one or more thresholds can be set for the predicted sales volume according to marketing needs. For example, if it is used for sales volume warning scenarios, the lowest warning line that the sales volume of a single financial product can bear can be set as the first set threshold. For another example, the lowest warning line that the total predicted sales volume of a financial institution can bear can also be set as the second set threshold.

[0078] After obtaining the sales forecast average value of each financial product, the sales forecast average value of each financial product can be compared with the first set threshold. If it is lower than the first set threshold, it means that there is an abnormality in the sales data of the current financial product, and a sales risk warning for the financial product will be issued to relevant management and operation personnel.

[0079] After obtaining the total daily sales forecast of the financial institution, the total sales forecast can also be compared with the second set threshold. If it is lower than the second set threshold, it means that there is an abnormality in the overall sales data of the current financial institution. In this way, a sales risk warning for the current financial institution will be issued to the relevant management and operation personnel to facilitate the relevant management and operation personnel to make corresponding strategy adjustments.

[0080] In another example, assuming that it is a scenario for sales growth monitoring, a third threshold value can be set for a single financial product. A fourth threshold value can also be set for a financial institution. After obtaining the sales forecast average of each financial product, the sales forecast average of each financial product can be compared with the third threshold value. If it is higher than the third threshold value, it means that the sales data of the current financial product is growing rapidly, and a prompt for rapid sales growth of the financial product is issued to the relevant management and operation personnel to remind the relevant personnel to increase the marketing efforts of the financial product. After obtaining the total sales forecast of the financial institution every day, the total sales forecast can also be compared with the fourth threshold value. If it is higher than the fourth threshold value, it means that the overall sales data of the current financial institution is growing rapidly, and a prompt for sales growth of the current financial institution is issued to the relevant management and operation personnel, so that the relevant management and operation personnel can make corresponding strategic adjustments.

[0081] In this embodiment, the linkage relationship based on the multidimensional dependency model can predict the sales data forecast value of each financial product every day in the future preset time period, and improve the prediction accuracy through multiple predictions, so as to obtain the average sales forecast value of each financial product every day as the final sales forecast result of the financial product on that day. The linkage changes of each financial product can be analyzed based on the average sales forecast value of each financial product every day. It is also possible to calculate the sum of the average sales forecast value of each financial product every day to obtain the total sales forecast value of the current financial institution every day, and provide guidance information from macro-adjustment to micro-operation for the basic work of the financial institution, and prevent business risks. The method of using a multidimensional dependency model for prediction is much more reliable than the method of directly using historical business data to predict future data volume, and achieves the effect of improving prediction accuracy.

[0082] Embodiment 3

[0083] Figure 3 This is a flow chart of a method embodiment of an associated business prediction provided in Embodiment 3 of the present application. This embodiment, based on Embodiment 1 or Embodiment 2, describes the construction process of a multi-dimensional dependency model, which may include the following steps:

[0084] Step 310, collecting historical sales data of various financial products in the financial institution over a period of time.

[0085] In one implementation, the identifier of each financial product can be obtained from a preset list of financial products. Then, the historical sales data for a period of time in the past can be extracted from the business data log record or business data table of each financial product. For example, the historical sales data of each financial product in the past five years can be extracted. The sales data can be reflected in each sales data.

[0086] Step 320: Generate a training data set and a test data set based on the historical sales data of each financial product.

[0087] After collecting the historical sales data of various financial products over a period of time, the historical sales data can be processed and processed to generate training data sets and test data sets for building a multidimensional dependency model.

[0088] In one embodiment, step 320 may further include the following steps:

[0089] Step 320 - 1 , summarize the historical sales data of each financial product according to the set unit time to obtain the summary value of each unit time.

[0090] In one example, the historical sales data of each financial product may be aggregated with "day" as the unit of time to generate a daily sales summary value of each financial product.

[0091] Step 320 - 2 : for each financial product, organize the summary values ​​of each unit time of the financial product into time series data of the financial product.

[0092] For example, for a certain financial product, after obtaining the daily summary value of the financial product in the past period of time, the daily summary value can be arranged in chronological order to form the time series data of the financial product.

[0093] In one embodiment, after obtaining the time series data of each financial product, preliminary data cleaning and processing can be performed on each time series data, for example, missing data can be supplemented or removed. If the data span is too large, the data needs to be standardized or transformed by Box-Cox (a generalized power transformation method).

[0094] Step 320 - 3 , organizing the time series data of all financial products into a first data matrix.

[0095] In one implementation, the time series data after preliminary cleaning and processing can be arranged into a first data matrix according to financial products. In the first data matrix, rows correspond to daily summary values, and columns correspond to various financial products.

[0096] Step 320 - 4 , converting the first data matrix into a second data matrix, wherein the values ​​in the second data matrix are between [0, 1].

[0097] Specifically, the multidimensional dependency model used in this embodiment is a copula model. The structure used by the copula model is a structure used to process the relationship between various probability distribution functions, so its numerical type needs to be a value related to the probability distribution function, between [0,1] (probability value under empirical distribution). Therefore, the values ​​recorded in the first data matrix cannot be directly substituted into the copula structure, and need to be converted into values ​​between [0,1] to finally obtain the second data matrix to meet the data structure requirements of the subsequent copula model.

[0098] In one implementation, the values ​​recorded in each column of the first data matrix can be converted into values ​​between [0, 1] according to the sample quantile value. Specifically, the empirical distribution method can be used to arrange the data in each column in a row as the sample population, and then each data is converted into its corresponding quantile value in the sample population.

[0099] It should be noted that in practice, the 0% and 100% quantiles are not very common. In order to make the data more reasonable, without changing the data sorting, the quantiles corresponding to the minimum and maximum values ​​(the data corresponding to the 0% and 100% quantiles) can be appropriately scaled, for example, to 0.01% and 99.99%, so that the converted data will not have values ​​such as 0 and 1, thereby making the results calculated by the subsequent copula model more accurate.

[0100] Step 320-5: divide the second data matrix into a training data set and a test data set according to a set division ratio.

[0101] In one example, after obtaining the second data matrix, the data in the second data matrix can be divided into a training data set and a test data set in a ratio of 4:1 (ie, a set division ratio) for use by the model.

[0102] Step 330: Use the training data set to fit a preset initial multidimensional dependency model.

[0103] As described above, in one embodiment, the multidimensional dependency model is a copula model, and the maximum likelihood method can be used to fit the initial copula model to estimate the model parameters and obtain the fitting structure of the model.

[0104] In one embodiment, during the model fitting process, the Akaike Information Criterion (AIC) can also be used to select the best model as the fitted multidimensional dependency model. Specifically, the quality of the fitting structure can be judged by the following AIC criterion:

[0105] AIC=2k-2ln(L)

[0106] Where k is the number of parameters of the copula model; L is the likelihood function of the sample, which can be regarded as the copula density. The smaller the AIC value, the better the model.

[0107] The role of AIC is that during the fitting process of the copula model, due to the limitations of computer iterative computing technology, it is often impossible to find the global optimal point of the parameters, but only the local optimal point, and more than one local optimal point may be found (computer iterative technology often needs to give initial values ​​to each parameter, and then the initial values ​​of the iterative parameters are constantly changing and finally converge to the local optimal solution, so different initial values ​​often lead to different local optimal solutions). At this time, the final optimal solution can be determined by comparing the AIC values ​​of each local optimal point. In addition, if there are other dependent structure models as competitors, then in order to compare the copula with these models, it is also possible to judge which model is better based on the AIC criterion.

[0108] Step 340: Use the test data set and the VaR method to perform hypothesis testing on the fitted multidimensional dependency model.

[0109] For a well-fitted multidimensional dependency model, a test data set can also be used for hypothesis testing.

[0110] In one embodiment, step 340 may further include the following steps:

[0111] Step 340-1, using the fitted multidimensional dependency model to generate sales data forecast values ​​for each financial product, and calculating the sum of the sales data forecast values ​​for each financial product as the sales forecast sum of the financial institution; wherein the number of samples of the generated sales data forecast values ​​is the same as the number of samples in the test data set.

[0112] For example, if the current scenario is to predict the sales data of various financial products in a future time period, the length of this future time period is consistent with the length of the corresponding time period of the test set data. The new data generated by the fitted copula model simulation can be used as the sales data forecast value, where the forecast result of each prediction output is a multivariate vector, and each component in the multivariate vector is the sales data forecast value of the corresponding financial product.

[0113] After obtaining the sales data forecast value of each financial product generated by the copula model, the sales data forecast value of each financial product can be added up to obtain the total sales forecast of the current financial institution.

[0114] Step 340 - 2 , calculating the sum of the historical sales data of each sample in the test data set as the test set sum.

[0115] The data samples in the test data set are also multivariate vectors, and each component is the historical sales data of the corresponding financial product. In this step, the historical sales data of each sample in the test data set are added together to obtain the test set sum, which is also the test set sum of the historical sales data of the financial institution.

[0116] Step 340 - 3 , calculating the difference between the sales forecast sum and the test set sum, and determining the risk value based on the difference.

[0117] In this step, by comparing the difference between the sum of sales forecasts of financial institutions and the sum of test sets, a hypothesis test of the prediction effect of the copula model can be performed based on the difference. When performing hypothesis testing, the first step is to determine the value at risk VaR (Value at Risk) based on the difference. VaR refers to the risk quantile value of the value loss that may occur in a financial asset or investment portfolio in a certain period of time in the future under a certain confidence level. In one implementation, the above difference may include one or a combination of conditional variance estimates, mean, standard deviation, etc., which is not limited in this embodiment. After determining the difference between the sum of sales forecasts and the sum of test sets, VaR can be calculated based on the difference and the specified calculation formula.

[0118] If the confidence level is α, VaR can be expressed as VaR α , where VaR α It can be the corresponding risk quantile value. For example, VaR 0.9 It means the 90% quantile. For example, we can use the fitted copula structure to simulate 1000 data samples, add up the predicted sum of these 1000 samples, and take the 90% quantile value of these predicted sums as VaR 0.9 Then observe that the samples in the test data set exceed VaR 0.9 The probability of whether it conforms to the theoretical frequency, that is, 1-0.9=10%.

[0119] Step 340 - 4 , performing hypothesis testing on the fitted multidimensional dependency model based on the risk value and the test data set.

[0120] In one embodiment, VaR can be relied upon α and the data in the test data set, construct three hypothesis tests, namely, unconditional coverage test statistic LRuc, independence test statistic LRind and conditional coverage test statistic LRcc, to test the multidimensional dependency model. Then step 340-4 may further include the following steps:

[0121] Step 340-4-1, compare the observed value in the test data set at time t with the risk value at time t, and determine the indicative value at time t based on the result of the comparison.

[0122] Wherein, the indicative value is the value of the indicative function at time t. In implementation, the indicative function I can be predefined α (t), assuming that the observed value at the current time t in the test data set is x t , the risk value at time t is VaR α (t), then when x t Greater than VaR α (t), I α The indicative value of (t) is 1, otherwise the indicative value is 0.

[0123] Step 340-4-2, constructing an unconditional coverage test statistic LRuc based on the indicative values ​​observed over a period of time and the proportion of observed values ​​greater than the risk value values ​​at the same time.

[0124] Among them, LRuc is used to evaluate the observed value x at the current time t in the test data set t Exceeding VaR α Is the frequency of occurrence of the number of times significantly different from the confidence level α? In the implementation, assume that the sample size in the test data set is n, and I α (t) obeys the Bernoulli distribution with parameter α, so the LRuc statistic obeys the chi-square distribution with parameter 1, and we have:

[0125]

[0126] in, is the actual observed x t Exceeding VaR α (t) ratio, is the value of the corresponding Bernoulli log-likelihood function.

[0127] Step 340-4-3, constructing an independence test statistic LRind according to the changes of the indicative values ​​observed over a period of time.

[0128] Among them, LRind is used to evaluate the observed value x in the test data set t Exceeding VaR α Whether the occurrence times of the number of times are independent of each other, that is, the previous extreme event and the subsequent extreme event do not affect each other. ij =P(I α (t) = j | I α (t-1) = i), that is, the frequency statistics of the next value j when the previous indicative function value is i, nij Indicates I α The number of times i changes to j. So given

[0129]

[0130]

[0131] and, So we have:

[0132]

[0133] Step 340-4-4, constructing a conditional coverage test statistic LRcc according to the LRuc and the LRind.

[0134] In one implementation, LRcc can be a combination of LRuc and LRind, provided that the model setting is established:

[0135]

[0136] Step 340-4-5, respectively use the LRuc, the LRind and the LRcc to test the multidimensional dependency model, and when the LRuc, the LRind and the LRcc all pass the test on the multidimensional dependency model, it is determined that the test on the multidimensional dependency model passes.

[0137] Specifically, the three hypothesis tests consisting of LRuc, LRind and LRcc above constitute a complete theoretical test system. The above LRuc, LRind and LRcc can be used to test the fitted copula model. If the test results of the three tests are all passed, it means that the model has passed the test and can be used. If it fails to pass any test, it means that the model has defects in its basic structure. Even if the prediction accuracy is good, it cannot be used because it may be a coincidence caused by specific data. This provides a more rigorous and accurate basis for judgment.

[0138] Step 350, generating a Kendall correlation matrix and a Spearman correlation matrix based on the multidimensional dependency model, and performing a weighted operation on the Kendall correlation matrix and the Spearman correlation matrix to generate a weighted coefficient matrix.

[0139] Among them, the Kendall correlation matrix is ​​a matrix generated by a pre-trained multidimensional dependency model and is used to record the Kendall correlation coefficients of two financial products; the Spearman correlation matrix is ​​a matrix generated by a multidimensional dependency model and is used to record the Spearman correlation coefficients of two financial products.

[0140] Step 360: monitor the sales data of each financial product in the weighted coefficient matrix, and use the monitored financial products that trigger the preset sales conditions as target financial products.

[0141] Step 370: From the weighting coefficient matrix, obtain a financial product whose weighting coefficient meets a preset condition with that of the target financial product as an associated financial product.

[0142] Step 380: Use the multi-dimensional dependency model to perform sales forecasting on the associated financial products and the target financial product.

[0143] Among them, the results of the above sales forecast are used to assist relevant personnel in determining whether to increase the promotion efforts for the related financial products.

[0144] In this embodiment, the correlation matrix can be extracted from the multidimensional dependency model that has passed the test. When testing the multidimensional dependency model, the three major tests of LRuc, LRin, and LRcc can be used to evaluate the prediction effect of the model. The validity of the model is no longer guaranteed by relying solely on the test of one dimension of prediction accuracy. The verification of the validity of the model is theoretically strengthened to ensure the reliability of the model.

[0145] Embodiment 4

[0146] Figure 4 The structural block diagram of an embodiment of an apparatus for predicting associated services provided in Embodiment 4 of the present application may include the following modules:

[0147] The weighting coefficient matrix acquisition module 410 is used to acquire a weighting coefficient matrix, wherein the weighting coefficient matrix is ​​a matrix generated by weighting the Kendall correlation matrix and the Spearman correlation matrix, wherein the Kendall correlation matrix is ​​a matrix generated by using a pre-trained multidimensional dependency model and used to record the Kendall correlation coefficients of two financial products, and the Spearman correlation matrix is ​​a matrix generated by using the multidimensional dependency model and used to record the Spearman correlation coefficients of two financial products; the multidimensional dependency model is a model tested by the value at risk VaR method;

[0148] The target financial product detection module 420 is used to monitor the sales data of each financial product in the weighted coefficient matrix, and use the monitored financial products that trigger the preset sales conditions as target financial products;

[0149] The associated financial product determination module 430 is used to obtain, from the weight coefficient matrix, a financial product whose weight coefficient meets a preset condition with the target financial product as an associated financial product;

[0150] The sales forecasting module 440 is used to use the multi-dimensional dependency model to make sales forecasts for the associated financial products and the target financial products. The sales forecast results are used to assist relevant personnel in determining whether to increase the promotion efforts for the associated financial products.

[0151] In one embodiment, the apparatus may further include a model building module for building a multidimensional dependency model, including the following submodules:

[0152] The historical data collection submodule is used to collect the historical sales data of various financial products in financial institutions over the past period of time;

[0153] A data set generation submodule, used to generate a training data set and a test data set based on the historical sales data of each financial product;

[0154] A model fitting submodule, used to fit a preset initial multidimensional dependency model using the training data set;

[0155] The model testing submodule is used to perform hypothesis testing on the fitted multidimensional dependency model using the test data set and the VaR method.

[0156] In one embodiment, the model checking submodule may further include the following units:

[0157] A prediction unit, used to generate sales data prediction values ​​for each financial product using the fitted multidimensional dependency model, and calculate the sum of the sales data prediction values ​​for each financial product as the sales prediction sum of the financial institution; wherein the number of samples of the generated sales data prediction values ​​is the same as the number of samples in the test data set;

[0158] A test set sum calculation unit, used to calculate the sum of the historical sales data of each sample in the test data set as the test set sum;

[0159] a value-at-risk value determining unit, configured to calculate a difference between the sales forecast sum and the test set sum, and determine a value-at-risk value based on the difference;

[0160] A hypothesis testing unit is used to perform hypothesis testing on the fitted multidimensional dependency model based on the risk value and the test data set.

[0161] In one embodiment, the hypothesis testing unit is specifically used to:

[0162] Comparing the observed value in the test data set at time t with the risk value at time t, and determining the indicative value at time t according to the comparison result;

[0163] According to the indicative values ​​observed over a period of time, and the proportion of observed values ​​greater than the risk value values ​​at the same time, an unconditional covering test statistic LRuc is constructed;

[0164] According to the changes of the indicative values ​​observed over a period of time, an independence test statistic LRind is constructed;

[0165] Construct a conditional coverage test statistic LRcc according to the LRuc and the LRind;

[0166] The multidimensional dependency model is tested using the LRuc, the LRind, and the LRcc respectively, and when the tests of the multidimensional dependency model by the LRuc, the LRind, and the LRcc all pass, it is determined that the test of the multidimensional dependency model passes.

[0167] In one embodiment, the model building module may further include the following submodules:

[0168] The model selection submodule is used to select the best model as the fitted multidimensional dependency model using the Akaike Information Criterion AIC during the fitting process.

[0169] In one embodiment, the data set generation submodule is specifically used for:

[0170] Summarize the historical sales data of each financial product according to the set unit time to obtain the summary value of each unit time;

[0171] For each financial product, organize the summary values ​​of each unit time of the financial product into time series data of the financial product;

[0172] Organize the time series data of all financial products into the first data matrix;

[0173] Convert the first data matrix into a second data matrix, wherein the values ​​in the second data matrix are between [0, 1];

[0174] The second data matrix is ​​divided into a training data set and a test data set according to a set division ratio.

[0175] In one embodiment, the sales forecasting module 440 is specifically used to:

[0176] Based on a preset number of predictions, when executing each prediction, the multidimensional dependency model is used to respectively predict the sales data forecast values ​​of each associated financial product and the target financial product on a daily basis within a preset time period in the future;

[0177] Based on the number of predictions and the sales data forecast value of each prediction, the daily sales forecast average values ​​of each associated financial product and the target financial product are calculated respectively.

[0178] In one embodiment, the device further comprises:

[0179] The sales forecast total amount acquisition module is used to obtain the daily sales forecast average value of each financial product of the financial institution, and calculate the sum of the daily sales forecast average values ​​of each financial product to obtain the daily sales forecast total amount of the financial institution.

[0180] In one embodiment, the device further includes a sales evaluation notification module, configured to:

[0181] If the sales forecast average value is lower than the first set threshold, or the sales forecast total value is lower than the second set threshold, a first sales assessment notice is issued to relevant management and operation personnel;

[0182] or,

[0183] If the sales forecast average value is higher than the third set threshold, or the sales forecast total value is higher than the fourth set threshold, a second sales assessment notification is issued to relevant management and operation personnel.

[0184] In one embodiment, the associated financial product determination module 430 is specifically used to:

[0185] Obtaining target weighting coefficients of the target financial product and other financial products from the weighting coefficient matrix;

[0186] Financial products whose target weighted coefficient is greater than the preset coefficient threshold are considered as associated financial products;

[0187] If there is no financial product with a target weighting coefficient greater than the preset coefficient threshold, the first several financial products with the largest target weighting coefficients are taken as associated financial products.

[0188] An apparatus for predicting associated services provided in an embodiment of the present application can execute a method for predicting associated services in any one of Embodiments 1 to 3 of the present application, and has functional modules and beneficial effects corresponding to the execution method.

[0189] Embodiment 5

[0190] Figure 5 A schematic diagram of the structure of an electronic device provided in Embodiment 5 of the present application is shown in FIG. Figure 5 As shown, the electronic device includes a processor 510, a memory 520, an input device 530 and an output device 540; the number of the processor 510 in the electronic device can be one or more. Figure 5A processor 510 is taken as an example; the processor 510, the memory 520, the input device 530 and the output device 540 in the electronic device can be connected via a bus or other means. Figure 5 The example of connecting through bus is taken in the following.

[0191] The memory 520, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as the program instructions / modules corresponding to the above-mentioned embodiment 1, embodiment 2 or embodiment 3 in the embodiments of the present application. The processor 510 executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 520, that is, implements the method mentioned in the above-mentioned method embodiment 1, embodiment 2 or embodiment 3.

[0192] The memory 520 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 520 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 520 may further include a memory remotely arranged relative to the processor 510, and these remote memories may be connected to the device / terminal / server via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0193] The input device 530 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the electronic device. The output device 540 may include a display device such as a display screen.

[0194] Embodiment 6

[0195] Embodiment 6 of the present application also provides a storage medium containing computer executable instructions, and when the computer executable instructions are executed by a computer processor, they are used to execute the method in Embodiment 1, Embodiment 2, or Embodiment 3 of the above method.

[0196] Of course, the storage medium containing computer executable instructions provided in the embodiments of the present application is not limited to the operations of the method described above, but can also execute related operations in the method provided in any embodiment of the present application.

[0197] Embodiment 7

[0198] Embodiment 7 of the present application also provides a computer program product, which includes computer executable instructions. When the computer executable instructions are executed by a computer processor, they are used to execute the method in Embodiment 1, Embodiment 2, or Embodiment 3 of the above method.

[0199] Of course, the computer program product provided by the embodiment of the present application, whose computer executable instructions are not limited to the method operations described above, can also execute related operations in the method provided by any embodiment of the present application.

[0200] Through the above description of the implementation method, the technicians in the relevant field can clearly understand that the present application can be implemented with the help of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for an electronic device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0201] It is worth noting that in the embodiment of the above-mentioned device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application.

[0202] Note that the above are only preferred embodiments of the present application and the technical principles used. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present application. Therefore, although the present application is described in more detail through the above embodiments, the present application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A method for predicting associated business, characterized in that: The method comprises: Obtain a weighted coefficient matrix, wherein the weighted coefficient matrix is ​​a matrix generated by weighted operation of the Kendall correlation matrix and the Spearman correlation matrix, the Kendall correlation matrix is ​​a matrix generated by using a pre-trained multidimensional dependency model and used to record the Kendall correlation coefficients of two financial products, and the Spearman correlation matrix is ​​a matrix generated by using the multidimensional dependency model and used to record the Spearman correlation coefficients of two financial products; the multidimensional dependency model is a model tested by the value at risk VaR method; each component corresponding to the rows and columns in the weighted coefficient matrix represents each financial product of a financial institution, and the elements in the weighted coefficient matrix represent the weighted coefficients between the financial products represented by the rows and columns where the elements are located; monitor the sales data of each financial product in the weighted coefficient matrix, and use the monitored financial products that trigger preset sales conditions as target financial products; From the weighting coefficient matrix, a financial product whose weighting coefficient meets a preset condition with that of the target financial product is obtained as an associated financial product; Using the multi-dimensional dependency model to make sales forecasts for the associated financial products and the target financial products, the sales forecast results are used to assist relevant personnel in determining whether to increase the promotion efforts for the associated financial products; The multidimensional dependency model is constructed in the following manner: Collect historical sales data of various financial products in financial institutions over the past period of time; Generate a training data set and a test data set based on the historical sales data of each financial product; Using the training data set to fit a preset initial multidimensional dependency model; The test data set and the VaR method are used to perform hypothesis testing on the fitted multidimensional dependency model.

2. The method according to claim 1, characterized in that The method of using the test data set and the VaR method to perform hypothesis testing on the fitted multidimensional dependency model includes: Using the fitted multidimensional dependency model to generate sales data forecast values ​​for each financial product, and calculating the sum of the sales data forecast values ​​for each financial product as the sales forecast sum of the financial institution; wherein the number of samples of the generated sales data forecast values ​​is the same as the number of samples in the test data set; Calculate the sum of the historical sales data of each sample in the test data set as the test set sum; Calculating the difference between the sales forecast sum and the test set sum, and determining a value at risk based on the difference; A hypothesis test is performed on the fitted multidimensional dependency model based on the risk value and the test data set.

3. The method according to claim 2, characterized in that The performing hypothesis testing on the fitted multidimensional dependency model based on the risk value and the test data set includes: Comparing the observed value in the test data set at time t with the risk value at time t, and determining the indicative value at time t according to the comparison result; According to the indicative values ​​observed over a period of time, and the proportion of observed values ​​greater than the risk value values ​​at the same time, an unconditional covering test statistic LRuc is constructed; According to the changes of the indicative values ​​observed over a period of time, an independence test statistic LRind is constructed; Construct a conditional coverage test statistic LRcc according to the LRuc and the LRind; The multidimensional dependency model is tested using the LRuc, the LRind, and the LRcc respectively, and when the tests of the multidimensional dependency model by the LRuc, the LRind, and the LRcc all pass, it is determined that the test of the multidimensional dependency model passes.

4. The method according to any one of claims 2 to 3, after fitting the preset initial multidimensional dependency model using the training data set, further comprising: In the fitting process, Akaike information criterion AIC was used to select the best model as the fitted multidimensional dependency model.

5. The method according to any one of claims 2-3, characterized in that: The generating of a training data set and a test data set based on the historical sales data of each financial product includes: Summarize the historical sales data of each financial product according to the set unit time to obtain the summary value of each unit time; For each financial product, organize the summary values ​​of each unit time of the financial product into time series data of the financial product; Organize the time series data of all financial products into the first data matrix; Convert the first data matrix into a second data matrix, wherein the values ​​in the second data matrix are between [0, 1]; The second data matrix is ​​divided into a training data set and a test data set according to a set division ratio.

6. The method according to claim 1, characterized in that The adopting the multi-dimensional dependency model to perform sales forecasting on the associated financial products and the target financial product includes: Based on a preset number of predictions, when executing each prediction, the multidimensional dependency model is used to respectively predict the sales data forecast values ​​of each associated financial product and the target financial product on a daily basis within a preset time period in the future; Based on the number of predictions and the sales data forecast value of each prediction, the daily sales forecast average values ​​of each associated financial product and the target financial product are calculated respectively.

7. The method according to claim 6, characterized in that Also includes: The daily sales forecast average value of each financial product of the financial institution is obtained, and the sum of the daily sales forecast average values ​​of each financial product is calculated to obtain the daily sales forecast total value of the financial institution.

8. The method according to claim 7, characterized in that The method further comprises: If the sales forecast average value of a financial product is lower than the first set threshold, or the sales forecast total value is lower than the second set threshold, a first sales assessment notice is issued to the relevant management and operation personnel; or, If the sales forecast average value of a certain financial product is higher than the third set threshold, or the sales forecast total value is higher than the fourth set threshold, a second sales assessment notice is issued to the relevant management and operation personnel.

9. The method according to claim 1, characterized in that: The step of acquiring, from the weighting coefficient matrix, a financial product whose weighting coefficient meets a preset condition with that of the target financial product as an associated financial product comprises: Obtaining target weighting coefficients of the target financial product and other financial products from the weighting coefficient matrix; Financial products whose target weighted coefficient is greater than the preset coefficient threshold are considered as associated financial products; If there is no financial product with a target weighting coefficient greater than the preset coefficient threshold, the first several financial products with the largest target weighting coefficients are taken as associated financial products.

10. A device for predicting associated business, characterized in that: The device comprises: A weighted coefficient matrix acquisition module is used to acquire a weighted coefficient matrix, wherein the weighted coefficient matrix is ​​a matrix generated by weighted operation of a Kendall correlation matrix and a Spearman correlation matrix, wherein the Kendall correlation matrix is ​​a matrix generated by using a pre-trained multidimensional dependency model and used to record the Kendall correlation coefficients of two financial products, and the Spearman correlation matrix is ​​a matrix generated by using the multidimensional dependency model and used to record the Spearman correlation coefficients of two financial products; the multidimensional dependency model is a model tested by the value at risk VaR method; each component corresponding to a row and a column in the weighted coefficient matrix represents each financial product of a financial institution, and an element in the weighted coefficient matrix represents a weighted coefficient between the financial products represented by the row and column where the element is located; A target financial product detection module, used to monitor the sales data of each financial product in the weighted coefficient matrix, and use the monitored financial products that trigger the preset sales conditions as target financial products; An associated financial product determination module, configured to obtain, from the weighting coefficient matrix, a financial product whose weighting coefficient meets a preset condition with that of the target financial product as an associated financial product; A sales forecasting module, used to use the multi-dimensional dependency model to make sales forecasts for the associated financial products and the target financial products, wherein the sales forecast results are used to assist relevant personnel in determining whether to increase the promotion efforts for the associated financial products; The device also includes: a model building module, used to build a multi-dimensional dependency model; The model building module comprises: The historical data collection submodule is used to collect the historical sales data of various financial products in financial institutions over the past period of time; A data set generation submodule, used to generate a training data set and a test data set based on the historical sales data of each financial product; A model fitting submodule, used to fit a preset initial multidimensional dependency model using the training data set; The model testing submodule is used to perform hypothesis testing on the fitted multidimensional dependency model using the test data set and the VaR method.

11. An electronic device, characterized in that: The electronic device comprises: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 9.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

13. A computer program product, comprising computer executable instructions, wherein the computer executable instructions are used to implement the method according to any one of claims 1 to 9 when executed.

Images