A data processing method, device, storage medium and computer program product

Abstract

The application discloses a data processing method and device, a storage medium and a computer program product, relates to the technical field of machine learning, and comprises the following steps: acquiring associated data sets related to investment concentration degrees in each investment and financial management product; performing normalization processing on the associated data sets, and using a prediction model last used for each business of each investment and financial management product to predict a first early warning threshold of each business investment concentration degree in a corresponding investment and financial management product according to the normalized associated data sets of the corresponding investment and financial management product; the prediction model is obtained based on an integrated learning voting algorithm; a multi-layer investment early warning system of an asset management system is constructed according to the first early warning thresholds of all businesses in each investment and financial management product and the weights corresponding to different investment levels in an asset management institution; and the investment concentration degrees corresponding to different levels are early warned by using the multi-layer investment early warning system.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, and in particular to a data processing method, apparatus, storage medium, and computer program product. Background Technology

[0002] As the requirements for investment risk control become increasingly stringent, asset management institutions typically need to issue early warnings about the investment concentration of their wealth management products and adjust their investment structure accordingly to achieve risk control management.

[0003] Managing the risks of wealth management products requires both ensuring compliance with financial regulations and maximizing asset liquidity, transaction success rate, and investment return. Existing regulatory rules prioritize investment concentration as the most crucial control factor. However, current technologies typically only provide early warnings for the investment concentration of individual wealth management products, and methods for handling investment concentration largely rely on expert experience and post-event compensation mechanisms. This results in low accuracy for investment concentration warnings.

[0004] Therefore, how to efficiently and accurately provide early warnings about investment concentration has become a pressing technical problem that needs to be solved. Summary of the Invention

[0005] This invention provides a data processing method, apparatus, storage medium, and computer program product to solve the technical problems of insufficient efficiency and accuracy in investment concentration early warning in the prior art.

[0006] Firstly, to solve the above-mentioned technical problems, the present invention provides a data processing method as follows:

[0007] Obtain the associated datasets related to investment concentration from each investment and wealth management product managed by the asset management institution;

[0008] The associated dataset is normalized, and the prediction model previously used for each business type in each investment and wealth management product is used to predict the first warning threshold of investment concentration for each business in the current corresponding investment and wealth management product based on the normalized associated dataset; wherein, the prediction model is trained based on an ensemble learning voting algorithm;

[0009] Based on the first warning threshold of all businesses in each investment and wealth management product and the weights corresponding to different investment levels in the asset management institution, a multi-level investment warning system is constructed for the asset management system; wherein, the first warning threshold corresponds to the bottom warning of the multi-level investment warning system;

[0010] The aforementioned multi-layered investment early warning system is used to provide early warnings about the investment concentration at different levels within the asset management institution.

[0011] One possible implementation involves obtaining a dataset related to investment concentration from each investment product managed by the asset management institution, including:

[0012] Collect investment transaction flow data for each investment and wealth management product in the asset management institution; and extract investment transaction feature data for each business type from the investment transaction flow data according to the business type.

[0013] Collect basic product information for each investment and wealth management product;

[0014] Collect the concentration ratio data of each single business type in each investment and wealth management product;

[0015] Collect regulatory data for each business product;

[0016] Collect asset ratio data released by peers;

[0017] The investment transaction characteristic data, the regulatory clause data, the product basic information, the concentration ratio data, and the asset ratio data released by peers are used to construct the associated dataset.

[0018] One possible implementation involves normalizing the associated dataset, including:

[0019] Based on the investment and transaction characteristics included in the business type, a one-dimensional array consisting of data of all investment and transaction characteristics included in the business type is constructed and used as the first dataset;

[0020] The asset valuation percentage of each business segment of each investment and wealth management product is used to construct a second dataset;

[0021] The product type of each investment and wealth management product and its corresponding regulatory data are constructed into a third dataset consisting of a two-dimensional array;

[0022] The asset ratio data released by peers for each investment and wealth management product is used as a reference influencing factor for the prediction model of the corresponding investment and wealth management product, and a fourth dataset is constructed.

[0023] Normalization is performed on the first dataset, the second dataset, the third dataset, and the fourth dataset respectively;

[0024] The normalized first dataset, normalized second dataset, normalized third dataset, and normalized fourth dataset are multiplied by their respective preset reference weights to obtain the normalized first dataset, normalized second dataset, normalized third dataset, and normalized fourth dataset.

[0025] One possible implementation method for obtaining the prediction model includes:

[0026] Obtain the historical associated dataset of the aforementioned business type;

[0027] The historical associated dataset is normalized.

[0028] The normalized historical association dataset is divided into a training set and a validation set according to a preset ratio;

[0029] The training set is used to train the classification business early warning model constructed by the ensemble learning voting algorithm, and the training set is used to verify the trained classification business early warning model until the accuracy of the trained classification business early warning model in predicting the concentration early warning threshold of the corresponding business type reaches the set value, thus obtaining the trained prediction model of the business type.

[0030] One possible implementation involves constructing a multi-tiered investment early warning system for the asset management system based on the first early warning threshold for all business types and the weights corresponding to different investment levels within the asset management institution. This system includes:

[0031] Based on the first warning threshold of each business in each investment and wealth management product and the frequency weight of each business type in each investment and wealth management product, the second warning threshold of investment concentration for each investment and wealth management product is determined.

[0032] Based on the second warning threshold for each investment and wealth management product and the size weight of each investment and wealth management product in the asset management institution, the third warning threshold for the investment concentration of the asset management institution is determined.

[0033] The multi-layered investment early warning system is constructed based on the first early warning threshold, the second early warning threshold, and the third early warning threshold.

[0034] One possible implementation, after using the multi-layered investment early warning system to provide early warnings about investment concentration at different levels within the asset management institution, further includes:

[0035] Obtain the associated dataset generated during the use of the multi-layered early warning system;

[0036] A correlation analysis is performed on the associated datasets generated after using the multi-layer early warning system to obtain non-alarm associated datasets that are positively correlated with the alarm associated datasets that have triggered alarms by the multi-layer early warning system but have not triggered alarms.

[0037] Using the alarm-associated dataset and the alarm-untriggered dataset, the previously used prediction model is iteratively trained to obtain the iterative prediction model, and the first warning threshold, the second warning threshold, and the third warning threshold are updated.

[0038] One possible implementation involves performing correlation analysis on the associated datasets generated during the use of the multi-layered early warning system to obtain a non-alarm associated dataset that is positively correlated with the alarm-related datasets triggered by the multi-layered early warning system but has not triggered an alarm. This includes:

[0039] The associated datasets generated during the use of the multi-layered early warning system will be divided into alarmed associated datasets and non-alarmed datasets.

[0040] Calculate the cosine similarity between the alarmed associated data in the alarmed dataset and the non-alarmed associated data in the non-alarmed dataset;

[0041] Non-alarm related data with a cosine similarity greater than a set threshold are used as positively correlated data of the corresponding alarm related data.

[0042] The set of all positively correlated data is defined as the non-alarm correlated dataset.

[0043] Secondly, embodiments of the present invention provide a data processing apparatus, comprising:

[0044] The acquisition unit is used to acquire the associated dataset related to investment concentration in each investment and wealth management product managed by the asset management institution.

[0045] The processing unit is used to normalize the associated dataset and predict the first warning threshold of investment concentration of each business in the corresponding investment and wealth management product based on the prediction model used last time for each business in each investment and wealth management product and the associated dataset after normalization of the corresponding investment and wealth management product; wherein, the prediction model is trained based on the ensemble learning voting algorithm.

[0046] A construction unit is used to construct a multi-layered investment early warning system for the asset management system based on the first early warning threshold of all businesses in each investment and wealth management product and the weights corresponding to different investment levels in the asset management institution; wherein, the first early warning threshold corresponds to the bottom-level early warning of the multi-layered investment early warning system;

[0047] The early warning unit is used to provide early warnings about the investment concentration at different levels within the asset management institution using the multi-layered investment early warning system.

[0048] In one possible implementation, the acquisition unit is further configured to:

[0049] Collect investment transaction flow data for each investment and wealth management product in the asset management institution; and extract investment transaction feature data for each business type from the investment transaction flow data according to the business type.

[0050] Collect basic product information for each investment and wealth management product;

[0051] Collect the concentration ratio data of each single business type in each investment and wealth management product;

[0052] Collect regulatory data for each business product;

[0053] Collect asset ratio data released by peers;

[0054] The investment transaction characteristic data, the regulatory clause data, the product basic information, the concentration ratio data, and the asset ratio data released by peers are used to construct the associated dataset.

[0055] In one possible implementation, the processing unit is further configured to:

[0056] Based on the investment and transaction characteristics included in the business type, a one-dimensional array consisting of data of all investment and transaction characteristics included in the business type is constructed and used as the first dataset;

[0057] The asset valuation percentage of each business segment of each investment and wealth management product is used to construct a second dataset;

[0058] The product type of each investment and wealth management product and its corresponding regulatory data are constructed into a third dataset consisting of a two-dimensional array;

[0059] The asset ratio data released by peers for each investment and wealth management product is used as a reference influencing factor for the prediction model of the corresponding investment and wealth management product, and a fourth dataset is constructed.

[0060] Normalization is performed on the first dataset, the second dataset, the third dataset, and the fourth dataset respectively;

[0061] The normalized first dataset, normalized second dataset, normalized third dataset, and normalized fourth dataset are multiplied by their respective preset reference weights to obtain the normalized first dataset, normalized second dataset, normalized third dataset, and normalized fourth dataset.

[0062] In one possible implementation, the building unit is further configured to:

[0063] Obtain the historical associated dataset of the aforementioned business type;

[0064] The historical associated dataset is normalized.

[0065] The normalized historical association dataset is divided into a training set and a validation set according to a preset ratio;

[0066] The training set is used to train the classification business early warning model constructed by the ensemble learning voting algorithm, and the training set is used to verify the trained classification business early warning model until the accuracy of the trained classification business early warning model in predicting the concentration early warning threshold of the corresponding business type reaches the set value, thus obtaining the trained prediction model of the business type.

[0067] In one possible implementation, the building unit is further configured to:

[0068] Based on the first warning threshold of each business in each investment and wealth management product and the frequency weight of each business type in each investment and wealth management product, the second warning threshold of investment concentration for each investment and wealth management product is determined.

[0069] Based on the second warning threshold for each investment and wealth management product and the size weight of each investment and wealth management product in the asset management institution, the third warning threshold for the investment concentration of the asset management institution is determined.

[0070] The multi-layered investment early warning system is constructed based on the first early warning threshold, the second early warning threshold, and the third early warning threshold.

[0071] In one possible implementation, the building unit is further configured to:

[0072] After using the multi-layered investment early warning system to issue early warnings on the investment concentration corresponding to different levels in the asset management institution, the associated dataset generated during the use of the multi-layered early warning system is obtained.

[0073] A correlation analysis is performed on the associated datasets generated after using the multi-layer early warning system to obtain non-alarm associated datasets that are positively correlated with the alarm associated datasets that have triggered alarms by the multi-layer early warning system but have not triggered alarms.

[0074] Using the alarm-associated dataset and the alarm-untriggered dataset, the previously used prediction model is iteratively trained to obtain the iterative prediction model, and the first warning threshold, the second warning threshold, and the third warning threshold are updated.

[0075] In one possible implementation, the building unit is further configured to:

[0076] The associated datasets generated during the use of the multi-layered early warning system will be divided into alarmed associated datasets and non-alarmed datasets.

[0077] Calculate the cosine similarity between the alarmed associated data in the alarmed dataset and the non-alarmed associated data in the non-alarmed dataset;

[0078] Non-alarm related data with a cosine similarity greater than a set threshold are used as positively correlated data of the corresponding alarm related data.

[0079] The set of all positively correlated data is defined as the non-alarm correlated dataset.

[0080] Thirdly, embodiments of the present invention also provide a data processing apparatus, comprising:

[0081] At least one processor, and

[0082] Memory connected to the at least one processor;

[0083] The memory stores instructions that can be executed by the at least one processor, which executes the instructions stored in the memory to perform the method described in the first aspect above.

[0084] Fourthly, embodiments of the present invention also provide a readable storage medium, comprising:

[0085] memory,

[0086] The memory is used to store instructions that, when executed by a processor, cause the apparatus including the readable storage medium to perform the method described in the first aspect above.

[0087] Fifthly, embodiments of the present invention also provide a computer program product comprising computer-executable instructions for causing a computer to perform the method described in the first aspect.

[0088] Through the technical solutions in one or more of the above embodiments of the present invention, the embodiments of the present invention have at least the following technical effects:

[0089] In the embodiments provided by this invention, a correlation dataset related to investment concentration is obtained for each business product managed by the asset management institution; the correlation dataset is normalized, and the prediction model previously used for each business product in each investment and wealth management product is used to predict the first warning threshold of investment concentration for each business in the current corresponding investment and wealth management product based on the normalized correlation dataset; wherein, the prediction model is trained based on an ensemble learning voting algorithm, so that a corresponding prediction model can be constructed according to different business products using the ensemble learning voting algorithm, and the first warning threshold of investment concentration for the corresponding business is predicted using the latest iteration of the prediction model; then, based on the first warning threshold of all businesses in the investment and wealth management product and the weights corresponding to different investment levels in the asset management institution, a multi-layer investment warning system for the asset management system is constructed; wherein, the first warning threshold corresponds to the bottom warning of the multi-layer investment warning system; the multi-layer investment warning system is used to issue warnings for investment concentration corresponding to different levels in the asset management institution. This adaptively adjusts the warning thresholds of investment concentration at different levels in the multi-layer investment warning system, improves the accuracy of concentration risk warnings, reduces the probability of triggering concentration risk at each level in the asset management institution, and improves investment success rate, making the investment and wealth management products managed by the asset management institution have better liquidity. Meanwhile, since the aforementioned multi-layered investment early warning system does not rely on human experience and post-event compensation for concentration control, nor is it limited to a single business type or product, it can comprehensively control concentration at different levels of asset management institutions, thereby improving the efficiency of concentration risk management. Attached Figure Description

[0090] Figure 1 A flowchart of a data processing method provided in an embodiment of the present invention;

[0091] Figure 2 This is a schematic diagram of the structure of a multi-layer early warning system provided in an embodiment of the present invention;

[0092] Figure 3 A schematic diagram of the iterative prediction model and the updated multi-layer early warning system provided in the embodiments of the present invention;

[0093] Figure 4 This is a schematic diagram of the structure of a data processing device provided in an embodiment of the present invention. Detailed Implementation

[0094] This invention provides a data processing method, apparatus, storage medium, and computer program product to address the technical problems of inefficient and inaccurate investment concentration early warning in existing technologies. The acquisition, storage, use, and processing of data in the proposed solution all comply with relevant national laws and regulations.

[0095] To better understand the above technical solutions, the technical solutions of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solutions of the present invention, rather than limitations on the technical solutions of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0096] Please refer to Figure 1 This invention provides a data processing method, the processing procedure of which is as follows.

[0097] Step 101: Obtain the associated dataset related to investment concentration for each investment and wealth management product managed by the asset management institution.

[0098] Asset management institutions can be securities companies, insurance asset management companies, public fund companies, private fund companies, etc., with no specific restrictions. Investment and wealth management products may include one or more businesses, each belonging to a business type. Business types may include stocks, funds, bonds, futures, options, cash deposits, repurchase agreements, non-standard assets, etc.

[0099] In some embodiments, obtaining the associated dataset related to investment concentration in each investment and wealth management product managed by the asset management institution can be achieved in the following ways:

[0100] Collect investment transaction flow data for each investment and wealth management product in the asset management institution; and extract investment transaction feature data for each business type from the investment transaction flow data.

[0101] The investment transaction flow data corresponding to each business in the investment and wealth management product is divided into investment transaction flow data for each business type, and investment transaction feature data for each business type is extracted from it. Investment transaction feature data may include, for example, transaction price and transaction quantity.

[0102] Collect basic product information for each investment and wealth management product.

[0103] The basic information of investment and wealth management products includes data on product type, collateral, and investment direction. Product type data includes cash management products, equity products, fixed-income products, money market funds, and so on.

[0104] Collect the concentration ratio data of each investment and wealth management product for each single business type.

[0105] The concentration ratio data for single business types in investment and wealth management products includes: the concentration ratio of single product investment in bonds, the concentration ratio of single product investment in equity securities, the concentration ratio of single product investment in cash assets, the concentration ratio of institutional investment in large-denomination bank certificates of deposit, etc.

[0106] The aforementioned internal data, such as investment transaction flow data, basic product information, and concentration ratio data for a single business product, can be obtained by logging into the server database of the corresponding asset management institution. For example, by logging into the asset management institution's server database, one can obtain investment flow data, transaction flow data, and detailed asset disclosure flow data for business products such as stocks, bonds, funds, repurchase agreements, bidding, futures, options, cash deposits, and non-standard assets.

[0107] Collect regulatory data for each business product;

[0108] Collect asset ratio data released by peers, such as the details of various investment and wealth management products disclosed within the same industry during the term, the proportion of investment asset types in investment and wealth management products, and details of penalties for violations, etc.

[0109] The aforementioned regulatory data and external data such as the asset ratios published by peers can be obtained through API calls, email network optical character recognition (OCR) recognition, and other methods.

[0110] It is important to understand that different asset management institutions offer different investment and wealth management products, and the composition of the aforementioned related dataset also varies accordingly. Therefore, it should not be interpreted that the related dataset includes all types mentioned in the examples above.

[0111] The investment transaction characteristic data, the regulatory clause data, the product basic information, the concentration ratio data, and the asset ratio data released by peers are used to construct the associated dataset.

[0112] After obtaining the aforementioned associated dataset, step 102 can be executed.

[0113] Step 102: Normalize the associated dataset and use the prediction model previously used for each business in each investment and wealth management product to predict the first warning threshold of investment concentration for each business in the current corresponding investment and wealth management product based on the normalized associated dataset; wherein, the prediction model is trained based on the ensemble learning voting algorithm.

[0114] Normalization of associated datasets can be achieved in the following ways:

[0115] Based on the investment and trading characteristics included in the business type, a one-dimensional array is constructed, consisting of the data of all investment and trading characteristics included in the business type, and this array serves as the first dataset. For example, if business type 1 is a stock, all investment and trading characteristics of a stock include transaction price, transaction volume, etc. The data of these investment and trading characteristics constitute a one-dimensional array [A1, A2, ..., An]. This one-dimensional array can be called the first dataset A of business type 1, where A1 is the data of investment and trading characteristic 1 of business type 1 (which can be called investment and trading characteristic data 1), A2 is the data of investment and trading characteristic 2 of business type 1 (which can be called investment and trading characteristic data 2), ..., An is the data of investment and trading characteristic n of business type 2 (which can be called investment and trading characteristic data n).

[0116] The asset valuation percentage of each business type of each investment and wealth management product is used to construct a second dataset; for example, if the investment and wealth management products include business types A1 to An, the asset estimated value percentage of business types A1 to An is used to construct a second dataset B.

[0117] The product type of each investment and wealth management product and its corresponding regulatory data are constructed into a third dataset consisting of a two-dimensional array. For example, if the product type is used as the horizontal data of the two-dimensional array and the regulatory data corresponding to the product type is used as the vertical data, then the third dataset C[m'][n'] can be constructed, where m' is the number of rows of the two-dimensional array and n' is the number of columns of the two-dimensional array.

[0118] The asset ratio data released by peers for each investment and wealth management product is used as a reference influencing factor for the prediction model, and a fourth dataset (let's call it D) is constructed.

[0119] For each investment and financial product, normalization is performed on the first dataset (A), the second dataset (B), the third dataset (C), and the fourth dataset (D).

[0120] The normalized first dataset (denoted as A'), the normalized second dataset (denoted as B'), the normalized third dataset (denoted as C'), and the normalized fourth dataset (denoted as D') are multiplied by their respective preset reference weights to obtain the normalized first dataset, the normalized second dataset, the normalized third dataset, and the normalized fourth dataset.

[0121] For example, the preset weights for the first normalized dataset (denoted as A'), the second normalized dataset (denoted as B'), and the third normalized dataset (denoted as C') are all internal weights of the production management institution (denoted as w1). The preset weights for the fourth normalized dataset (denoted as D') are peer reference weights (denoted as w2). Usually, w1 > w2. For example, w1 can be set to 0.8 and w2 can be set to 0.2.

[0122] Multiplying A', B', and C' by w1 respectively yields the normalized first dataset (A'×w1), the normalized second dataset (B'×w1), and the normalized third dataset (C'×w1). Multiplying D' by w2 yields the normalized fourth dataset (D'×w2).

[0123] Alternatively, A, B, and C can be multiplied by w1, and D can be multiplied by w2. Then, the results of each multiplication can be normalized to obtain the first normalized dataset, the second normalized dataset, the third normalized dataset, and the fourth normalized dataset. These datasets constitute the normalized associated datasets of the corresponding investment and financial products.

[0124] After obtaining the normalized associated dataset for the corresponding business product, the prediction model previously used for the corresponding business product can be used to predict the first warning threshold for the investment concentration of each business in the current corresponding investment product based on the normalized dataset for the corresponding investment and wealth management product.

[0125] The forecasting model for each of the above business categories can be obtained in the following ways:

[0126] Obtain the historical correlation dataset for each business product; normalize the historical correlation dataset for each business product; divide the normalized historical correlation dataset into a training set and a validation set according to a preset ratio; train the classification business early warning model constructed by the ensemble learning voting algorithm using the training set, and validate the trained classification business early warning model using the validation set, until the accuracy of the trained classification business early warning model in predicting the concentration warning threshold of the corresponding business product reaches the set value, thus obtaining the trained prediction model for the corresponding business product (i.e., the prediction model for the business product).

[0127] For example, with a preset ratio of 9:1 and a set value of 90%, the historical associated dataset corresponding to business type 1 at a specified historical time is obtained using the aforementioned method for acquiring associated datasets. This dataset is then normalized using the same method as the aforementioned associated dataset, resulting in a normalized historical associated dataset. This normalized dataset is then divided into a training set and a validation set at a 9:1 ratio. The training set is used to train a classification business early warning model constructed using an ensemble learning voting algorithm, resulting in a trained classification business early warning model. The validation set is used to validate the trained classification business early warning model until the accuracy of the trained model in predicting the concentration warning threshold for the corresponding business type reaches 90%, thus obtaining a trained prediction model for business type 1. Using the trained prediction model for business type 1, the first warning value for business type 1 is predicted based on the normalized associated data. The training method for prediction models corresponding to other business types included in the asset management institution is similar and will not be elaborated further.

[0128] In other embodiments, for non-first-time predictions, the previously used prediction model can be used to predict the current first warning value for the corresponding business type based on normalized correlation data. This can improve the accuracy of the prediction. The previously used prediction model can be the prediction model obtained after the initial training, or it can be the prediction model used previously. The specific method for obtaining the iterative prediction model will be introduced later.

[0129] After obtaining the first warning threshold for each business in each investment and wealth management product, step 103 can be executed.

[0130] Step 103: Based on the first warning threshold of all businesses in each investment and wealth management product and the weights corresponding to different investment levels in the asset management institution, construct a multi-level investment warning system for the asset management system; wherein, the first warning threshold corresponds to the bottom warning of the multi-level investment warning system.

[0131] Step 104: Use a multi-tiered investment early warning system to provide early warnings on the investment concentration at different levels within the asset management institution.

[0132] In step 103, based on the first early warning threshold of all businesses in each investment and wealth management product and the weights corresponding to different investment levels in the asset management institution, a multi-layered investment early warning system for the asset management system is constructed, which can be achieved in the following ways:

[0133] Based on the first warning threshold for each business type in each investment and wealth management product and the frequency weight of each business type in the investment of each investment and wealth management product, the second warning threshold for the investment concentration of each investment and wealth management product is determined; based on the second warning threshold for each investment and wealth management product and the size weight of each investment and wealth management product in the asset management institution, the third warning threshold for the investment concentration of the asset management institution is determined; based on the first warning threshold, the second warning threshold, and the third warning threshold, a multi-layered investment warning system is constructed.

[0134] Please see Figure 2 This is a schematic diagram of the structure of a multi-layered early warning system provided in an embodiment of the present invention.

[0135] like Figure 2 As shown, an asset management institution manages multiple investment and wealth management products, which include multiple businesses (Business 1, Business 2, etc.). All businesses within these products can be categorized into n business types (Business Type 1 to Business Type n). Each business type has a corresponding prediction model, trained using an ensemble learning voting algorithm on the associated dataset of the corresponding business type. Based on the associated data of the business types within the investment and wealth management products, the prediction model can predict the first warning threshold for the investment concentration of the corresponding business. After predicting the first warning threshold for each business within each investment and wealth management product, the prediction model is combined with the data of each business type... By weighting the frequency of each investment product, the second warning threshold for investment concentration can be obtained. After obtaining the second warning threshold for each investment product managed by the asset management institution, the third warning threshold for investment concentration of the asset management institution can be obtained by weighting the scale of each investment product. In this way, the first warning threshold for investment concentration of each business (belonging to the first level of warning threshold, i.e., the bottom warning), the second warning threshold for investment concentration of each investment product (belonging to the second level of warning threshold), and the third warning threshold for investment concentration of the asset management institution (belonging to the third level of warning threshold) can be constructed into a multi-level warning system.

[0136] For example, suppose asset management institution A manages investment and wealth management products including Investment and Wealth Management Product 1, Investment and Wealth Management Product 2, and Investment and Wealth Management Product 3. Investment and Wealth Management Product 1 includes Business 1, Business 2, and Business 3, which belong to the business categories of stocks, bonds, and repurchase agreements, respectively. Investment and Wealth Management Product 2 includes Business 4, Business 5, and Business 6, which belong to the business categories of futures, options, and cash deposits, respectively. Investment and Wealth Management Product 3 includes Business 7, Business 8, and Business 9, which belong to the business categories of funds, cash deposits, and bonds, respectively. Using the prediction model previously used for stocks, based on the associated dataset of Business 1, the first warning threshold k1 for investment concentration of Business 1 is predicted. Using the prediction model previously used for bonds, based on the associated dataset of Business 2, the first warning threshold k2 for investment concentration of Business 2 is predicted. Using the prediction model previously used for repurchase agreements, based on the associated dataset of Business 3, the first warning threshold k3 for investment concentration of Business 3 is predicted. Similarly, the first warning thresholds for investment concentration of the businesses included in the other investment and wealth management products can be obtained, which will not be elaborated here.

[0137] If the frequency weights of stocks, bonds, and repurchase agreements are m1, m2, and m3 respectively, the above frequency weights are weighted and calculated with the first warning threshold corresponding to the business included in investment and wealth management product 1 to obtain the second warning threshold for investment concentration of investment and wealth management product 1 (denoted as p1, p1=k1×m1+k2×m2+k3×m3); similarly, the second warning thresholds for investment concentration of other investment and wealth management products can be obtained, which will not be elaborated here. The second warning threshold for investment concentration of investment and wealth management product 2 is denoted as p2, and the second warning threshold for investment concentration of investment and wealth management product 3 is denoted as p3.

[0138] If the size weights of investment and wealth management products 1, 2, and 3 are n1, n2, and n3 respectively, then by combining the weighted calculation of the second warning threshold of investment concentration of each investment and wealth management product, the third warning threshold of investment concentration of asset management institution A can be obtained (p1×n1+p2×n2+p3×n3).

[0139] In this way, by using the first early warning threshold corresponding to each of the above-mentioned businesses, the second early warning threshold corresponding to each investment and wealth management product, and the third early warning threshold corresponding to the asset management institution, a multi-layered investment early warning system for asset management institution A is constructed.

[0140] After constructing the aforementioned multi-layered investment early warning system, it can be used to issue early warnings for investment concentration at different levels within the asset management institution. If business exceeds the corresponding first early warning threshold, an alarm message can be issued to prompt relevant personnel to handle the situation promptly. Alarm messages may include pop-up notifications on the display screen, audible and visual alarms, and text messages sent to the relevant management personnel; the specific form of the alarm message is not limited here.

[0141] After using a multi-tiered early warning system to provide early warnings on investment concentration at different levels within an asset management institution, the previously used prediction model can be iterated based on the early warning information, and the multi-tiered early warning system can be updated. This can be achieved in the following ways:

[0142] Acquire the correlation data generated during the use of the multi-layer early warning system; perform correlation analysis on the correlation data generated during the use of the multi-layer early warning system to obtain positive correlation data of alarms that trigger alarms and non-alarm correlation data that do not trigger alarms; use the alarm correlation data and non-alarm correlation data to iteratively train the prediction model used last time to obtain the iterated prediction model, and update the first early warning threshold, the second early warning threshold, and the third early warning threshold.

[0143] In some embodiments, correlation analysis is performed on the data triggered by the multi-layered early warning system to obtain non-alarm correlated data that is positively correlated with the alarm correlated data triggered by the multi-layered early warning system but has not triggered an alarm. This can be achieved in the following ways:

[0144] The associated data generated during the use of the multi-level early warning system are divided into alarmed associated datasets and non-alarmed datasets; the cosine similarity between alarmed associated data in the alarmed associated dataset and non-alarmed associated data in the non-alarmed dataset is calculated; the non-alarmed associated data with a cosine similarity greater than a set threshold is taken as the positively correlated data of the corresponding alarmed associated data; the set of all positively correlated data is determined as the non-alarmed associated dataset.

[0145] In some embodiments, calculating the cosine similarity between alarm-related data in the alarm-related dataset and non-alarm-related data in the non-alarm dataset can be achieved in the following ways:

[0146] Extract the investment transaction feature data from both the alerted and non-alarmed related data. Calculate the cosine similarity between the investment transaction feature data corresponding to each alerted related data and the investment transaction feature data corresponding to each non-alarmed related data in the non-alarmed dataset. The formula for calculating the cosine similarity is as follows:

[0147] ;

[0148] Where cosθ is the cosine similarity, with a value range of [0,1], A is the investment transaction feature data corresponding to the non-alarm related data, B is the investment transaction feature data corresponding to the alarmed related data, and n is the total number of investment transaction features contained in the non-alarm related data and the alarmed related data. i For the i-th investment transaction feature in the investment transaction feature data of the non-alarm related data, B iThis refers to the data of the i-th investment transaction feature in the alarm-related data, where the value of i ranges from [1, n].

[0149] Please see Figure 3 This is a schematic diagram of the iterative prediction model and the updated multi-layer early warning system provided in the embodiments of the present invention.

[0150] Step 301: Obtain a multi-layered early warning system.

[0151] Step 302: Collect real-time related data.

[0152] Step 303: Use the warning thresholds corresponding to each level in the multi-level warning system to determine whether the business corresponding to the real-time related data is alarmed.

[0153] Step 304a: If yes, then obtain the investment transaction feature data of the alarm-related data.

[0154] The real-time related data corresponding to the alarmed business is determined by using a multi-layered investment system as the alarmed related data. The set of all alarmed related data is the alarmed related dataset.

[0155] Step 304b: If not, obtain the investment transaction characteristic data of the non-alarm related data.

[0156] The real-time related data corresponding to the non-alarm business is determined by using a multi-layer investment system as the non-alarm related data, and the set of all non-alarm related data is the non-alarm dataset.

[0157] Step 305: Perform correlation analysis.

[0158] For each alarm-related data in the alarm-related dataset, the investment transaction characteristic data is compared with each non-alarm-related data in the non-alarm-related dataset. Correlation analysis is performed, such as calculating cosine similarity between them. Non-alarm-related data with calculation results greater than a set threshold are identified as related to the corresponding alarm-related data (at this time, the corresponding non-alarm-related data can be called positively correlated data), otherwise they are not related (at this time, the corresponding non-alarm-related data can be called negatively correlated data).

[0159] Step 306: If irrelevant, then end.

[0160] Step 307: If relevant, obtain the investment transaction feature dataset of the alarmed related dataset and the investment transaction feature data of the non-alarmed related dataset.

[0161] The set of the aforementioned alarm-related data is taken as the alarm-related dataset, and the investment and transaction feature data of the alarm-related dataset is extracted. The set of the aforementioned positively correlated data is taken as the non-alarm-related dataset, and the investment and transaction feature data of the non-alarm-related dataset is extracted.

[0162] Step 308: Iterate over the prediction model used last time.

[0163] The prediction model used previously is iteratively trained using the investment transaction feature datasets of both alarm-related datasets and non-alarm-related datasets to obtain the iterative prediction model. The iterative prediction model is then used to update the corresponding business product.

[0164] For example, taking the aforementioned asset management institution A as an example, after obtaining the multi-level early warning system, the associated dataset generated during the use of the multi-level early warning system is acquired. Based on whether the associated data in the associated dataset triggered an alarm under the multi-level early warning system, the associated data that triggered an alarm is designated as alarmed associated data, and the associated data that did not trigger an alarm is designated as non-alarmed associated data. The set of all alarmed associated data is called the alarmed associated dataset, and the set of all non-alarmed associated data is called the non-alarmed dataset. Then, the cosine similarity between each alarmed associated data in the alarmed associated dataset and each non-alarmed associated data in the non-alarmed dataset is calculated. Non-alarmed associated data with a cosine similarity greater than a set threshold is designated as positively correlated data for the corresponding alarmed associated data. The set of all positively correlated data is designated as the non-alarmed associated dataset. Afterwards, the non-alarmed associated dataset and the alarmed associated dataset are normalized. The specific normalization process is similar to the normalization process described above and will not be repeated here. The normalized non-alarm associated dataset and the normalized alarm associated dataset are input into the previously used prediction model for iterative training, resulting in an iterative prediction model. This iterative prediction model updates the first warning threshold. Combined with the previously described methods for obtaining the second and third warning thresholds, these thresholds are also updated. This not only allows for timely iteration of the prediction model but also enables adaptive adjustment of the multi-layered warning system. As the prediction model is trained with more and more sample data over time, its prediction accuracy continuously improves. Consequently, since the first, second, and third warning thresholds are continuously updated with the iterative prediction model, the accuracy of the multi-layered warning system constructed from these thresholds in predicting concentration risk at different levels is improved. This reduces the probability of triggering concentration risk at each level, increases investment success rate, and enhances the liquidity of investment products.

[0165] In the embodiments provided by this invention, a correlation dataset related to investment concentration is obtained for each business product managed by the asset management institution; the correlation dataset is normalized, and the prediction model previously used for each business product in each investment and wealth management product is used to predict the first warning threshold of investment concentration for each business in the current corresponding investment and wealth management product based on the normalized correlation dataset; wherein, the prediction model is trained based on an ensemble learning voting algorithm, so that a corresponding prediction model can be constructed according to different business products using the ensemble learning voting algorithm, and the first warning threshold of investment concentration for the corresponding business is predicted using the latest iteration of the prediction model; then, based on the first warning threshold of all businesses in the investment and wealth management product and the weights corresponding to different investment levels in the asset management institution, a multi-layer investment warning system for the asset management system is constructed; wherein, the first warning threshold corresponds to the bottom warning of the multi-layer investment warning system; the multi-layer investment warning system is used to issue warnings for investment concentration corresponding to different levels in the asset management institution. This adaptively adjusts the warning thresholds of investment concentration at different levels in the multi-layer investment warning system, improves the accuracy of concentration risk warnings, reduces the probability of triggering concentration risk at each level in the asset management institution, and improves investment success rate, making the investment and wealth management products managed by the asset management institution have better liquidity. Meanwhile, since the aforementioned multi-layered investment early warning system does not rely on human experience and post-event compensation for concentration control, nor is it limited to a single business type or product, it can comprehensively control concentration at different levels of asset management institutions, thereby improving the efficiency of concentration risk management.

[0166] Based on the same inventive concept, one embodiment of the present invention provides a data processing apparatus. Specific implementation methods of the data processing method of this apparatus can be found in the description of the method embodiments section; repeated details will not be repeated here. Figure 4 The device includes:

[0167] Acquisition unit 401 is used to acquire the associated dataset related to investment concentration in each investment and wealth management product managed by the asset management institution.

[0168] Processing unit 402 is used to normalize the associated dataset and predict the first warning threshold of investment concentration of each business in the corresponding investment and wealth management product based on the prediction model used last time for each business in each investment and wealth management product after normalization of the associated dataset; wherein, the prediction model is trained based on the ensemble learning voting algorithm.

[0169] Construction unit 403 is used to construct a multi-level investment early warning system for the asset management system based on the first early warning threshold of all businesses in each investment and wealth management product and the weights corresponding to different investment levels in the asset management institution; wherein, the first early warning threshold corresponds to the bottom early warning of the multi-level investment early warning system;

[0170] The early warning unit 404 is used to provide early warning of investment concentration at different levels within the asset management institution using the multi-layer investment early warning system.

[0171] In one possible implementation, the acquisition unit 401 is further configured to:

[0172] Collect investment transaction flow data for each investment and wealth management product in the asset management institution; and extract investment transaction feature data for each business type from the investment transaction flow data according to the business type.

[0173] Collect basic product information for each investment and wealth management product;

[0174] Collect the concentration ratio data of each single business type in each investment and wealth management product;

[0175] Collect regulatory data for each business product;

[0176] Collect asset ratio data released by peers;

[0177] The investment transaction characteristic data, the regulatory clause data, the product basic information, the concentration ratio data, and the asset ratio data released by peers are used to construct the associated dataset.

[0178] In one possible implementation, the processing unit 402 is further configured to:

[0179] Based on the investment and transaction characteristics included in the business type, a one-dimensional array consisting of data of all investment and transaction characteristics included in the business type is constructed and used as the first dataset;

[0180] The asset valuation percentage of each business segment of each investment and wealth management product is used to construct a second dataset;

[0181] The product type of each investment and wealth management product and its corresponding regulatory data are constructed into a third dataset consisting of a two-dimensional array;

[0182] The asset ratio data released by peers for each investment and wealth management product is used as a reference influencing factor for the prediction model of the corresponding investment and wealth management product, and a fourth dataset is constructed.

[0183] Normalization is performed on the first dataset, the second dataset, the third dataset, and the fourth dataset respectively;

[0184] The normalized first dataset, normalized second dataset, normalized third dataset, and normalized fourth dataset are multiplied by their respective preset reference weights to obtain the normalized first dataset, normalized second dataset, normalized third dataset, and normalized fourth dataset.

[0185] In one possible implementation, the building unit 403 is further configured to:

[0186] Obtain the historical associated dataset of the aforementioned business type;

[0187] The historical associated dataset is normalized.

[0188] The normalized historical association dataset is divided into a training set and a validation set according to a preset ratio;

[0189] The training set is used to train the classification business early warning model constructed by the ensemble learning voting algorithm, and the training set is used to verify the trained classification business early warning model until the accuracy of the trained classification business early warning model in predicting the concentration early warning threshold of the corresponding business type reaches the set value, thus obtaining the trained prediction model of the business type.

[0190] In one possible implementation, the building unit 403 is further configured to:

[0191] Based on the first warning threshold of each business in each investment and wealth management product and the frequency weight of each business type in each investment and wealth management product, the second warning threshold of investment concentration for each investment and wealth management product is determined.

[0192] Based on the second warning threshold for each investment and wealth management product and the size weight of each investment and wealth management product in the asset management institution, the third warning threshold for the investment concentration of the asset management institution is determined.

[0193] The multi-layered investment early warning system is constructed based on the first early warning threshold, the second early warning threshold, and the third early warning threshold.

[0194] In one possible implementation, the building unit 403 is further configured to:

[0195] After using the multi-layered investment early warning system to issue early warnings on the investment concentration corresponding to different levels in the asset management institution, the associated dataset generated during the use of the multi-layered early warning system is obtained.

[0196] A correlation analysis is performed on the associated datasets generated after using the multi-layer early warning system to obtain non-alarm associated datasets that are positively correlated with the alarm associated datasets that have triggered alarms by the multi-layer early warning system but have not triggered alarms.

[0197] Using the alarm-associated dataset and the alarm-untriggered dataset, the previously used prediction model is iteratively trained to obtain the iterative prediction model, and the first warning threshold, the second warning threshold, and the third warning threshold are updated.

[0198] In one possible implementation, the building unit 403 is further configured to:

[0199] The associated datasets generated during the use of the multi-layered early warning system will be divided into alarmed associated datasets and non-alarmed datasets.

[0200] Calculate the cosine similarity between the alarmed associated data in the alarmed dataset and the non-alarmed associated data in the non-alarmed dataset;

[0201] Non-alarm related data with a cosine similarity greater than a set threshold are used as positively correlated data of the corresponding alarm related data.

[0202] The set of all positively correlated data is defined as the non-alarm correlated dataset.

[0203] Based on the same inventive concept, this embodiment of the invention provides an early warning device for investment concentration, comprising: at least one processor, and

[0204] Memory connected to the at least one processor;

[0205] The memory stores instructions that can be executed by the at least one processor, and the at least one processor executes the data processing method described above by executing the instructions stored in the memory.

[0206] Based on the same inventive concept, embodiments of the present invention also provide a readable storage medium, comprising:

[0207] memory,

[0208] The memory is used to store instructions that, when executed by a processor, cause a device including the readable storage medium to perform the data processing method described above.

[0209] Based on the same inventive concept, embodiments of the present invention also provide a computer program product comprising computer-executable instructions, wherein the computer-executable instructions are used to cause a computer to perform the data processing method described above.

[0210] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0211] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0212] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0213] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0214] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method of data processing, characterized by, The method comprises the following steps: obtaining a set of correlation data related to investment concentration in each investment product managed by an asset management institution; normalizing the set of correlation data, and predicting a first early warning threshold of investment concentration in each business of each investment product using a prediction model last used for each business of each investment product according to the normalized set of correlation data of the corresponding investment product, wherein the prediction model is trained based on an ensemble learning voting algorithm; constructing a multi-layer investment early warning system of the asset management system according to the first early warning threshold of all businesses in each investment product and the weight corresponding to different investment levels in the asset management institution, wherein the first early warning threshold corresponds to the bottom layer of the multi-layer investment early warning system; using the multi-layer investment early warning system to perform early warning on the investment concentration corresponding to different levels in the asset management institution; wherein, according to the first early warning threshold of all businesses in each investment product and the weight corresponding to different investment levels in the asset management institution, the multi-layer investment early warning system of the asset management system is constructed, which comprises: determining a second early warning threshold of investment concentration of each investment product according to the first early warning threshold of each business in each investment product and the frequency weight of each business category in each investment product; determining a third early warning threshold of investment concentration of the asset management institution according to the second early warning threshold of each investment product and the size weight of each investment product in the asset management institution; and constructing the multi-layer investment early warning system according to the first, second and third early warning thresholds.

2. The method of claim 1, wherein, obtaining a set of correlation data related to investment concentration in each investment product managed by an asset management institution, comprising: collecting investment transaction flow data corresponding to each investment product in the asset management institution; and extracting investment transaction feature data corresponding to each business category from the investment transaction flow data according to the business category; collecting product basic information of each investment product; collecting concentration ratio data of single business category in each investment product; collecting regulatory clause data of each business category; collecting asset proportion data published by the industry; constructing the investment transaction feature data, the regulatory clause data, the product basic information, the concentration ratio data, and the asset proportion data published by the industry into the set of correlation data.

3. The method of claim 2, wherein, normalizing the set of correlation data, comprising: constructing a one-dimensional array composed of all investment transaction features contained in the business category according to the investment transaction features contained in the business category, and taking it as a first data set; constructing the asset valuation proportion of each business category of each investment product into a second data set; constructing the product type of each investment product and the corresponding regulatory data into a third data set composed of two-dimensional arrays; constructing the asset proportion data published by the industry corresponding to each investment product as a reference influence factor of the prediction model of the corresponding investment product into a fourth data set; The first data set, the second data set, the third data set and the fourth data set are normalized respectively; The normalized first data set, the normalized second data set, the normalized third data set and the normalized fourth data set are multiplied by the corresponding preset reference weight respectively to obtain the normalized first data set, the normalized second data set, the normalized third data set and the normalized fourth data set.

4. The method of claim 1, wherein, The prediction model is obtained in the following manner: Obtain the historical correlation data set of the business variety; The historical correlation data set is normalized; The normalized historical correlation data set is divided into a training set and a validation set according to a preset ratio; The classification business early warning model constructed by the ensemble learning voting algorithm is trained using the training set, and the trained classification business early warning model is verified using the validation set until the accuracy of the trained classification business early warning model in predicting the concentration early warning threshold of the corresponding business variety reaches a set value, and the trained prediction model of the business variety is obtained.

5. The method according to any one of claims 1 to 4, characterized in that, After using the multi-layer investment early warning system to early warn the investment concentration corresponding to different levels in the asset management institution, the following steps are further included: Obtain the correlation data set generated during the use of the multi-layer early warning system; Perform correlation analysis on the correlation data set generated after using the multi-layer early warning system to obtain the alarm correlation data set positively correlated with the alarm correlation data set triggered by the multi-layer early warning system and the non-alarm correlation data set not triggered by the alarm; Iteratively train the last used prediction model using the already alarmed correlation data set and the non-alarm correlation data set to obtain an iteratively trained prediction model and update the first early warning threshold, the second early warning threshold and the third early warning threshold.

6. The method of claim 5, wherein, The correlation analysis on the correlation data set generated during the use of the multi-layer early warning system includes: The correlation data set generated during the use of the multi-layer early warning system is divided into an already alarmed correlation data set and a non-alarm correlation data set; Calculate the cosine similarity between the already alarmed correlation data in the already alarmed data set and the non-alarm correlation data in the non-alarm data set; The non-alarm correlation data corresponding to the cosine similarity greater than a set threshold is determined as the positively correlated data of the corresponding already alarmed correlation data; The set of all positively correlated data is determined as the non-alarm correlation data set.

7. An apparatus for data processing, characterized by It includes: The acquisition unit is configured to acquire a correlation data set related to investment concentration in each investment and financial product managed by the asset management institution; The processing unit is configured to normalize the correlation data set and use the last used prediction model of each business belonging to a business variety in each investment and financial product to predict the first early warning threshold of the investment concentration of each business in the corresponding investment and financial product according to the normalized correlation data set of the corresponding investment and financial product; wherein the prediction model is trained based on the ensemble learning voting algorithm; The constructing unit is configured to construct a multi-layer investment early warning system of the asset management system according to the first early warning threshold of each business in each investment product and the weight corresponding to different investment levels in the asset management institution, wherein the first early warning threshold corresponds to a bottom layer early warning of the multi-layer investment early warning system. The early warning unit is configured to use the multi-layer investment early warning system to perform early warning on the investment concentration corresponding to different levels in the asset management institution. The processing unit is further configured to determine a second early warning threshold of the investment concentration of each investment product according to the first early warning threshold of each business in each investment product and the frequency weight of each business variety in each investment product, determine a third early warning threshold of the investment concentration of the asset management institution according to the second early warning threshold of each investment product and the size weight of each investment product in the asset management institution, and construct the multi-layer investment early warning system according to the first early warning threshold, the second early warning threshold and the third early warning threshold.

8. An apparatus for data processing, characterized by It comprises: at least one processor, and a memory connected with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the at least one processor executes the instructions stored in the memory to perform the method of any one of claims 1-6.

9. A readable storage medium, characterized by, The memory is configured to store instructions, and when the instructions are executed by the processor, the apparatus comprising the readable storage medium completes the method of any one of claims 1-6. The computer executable instructions are used to make the computer execute the method of any one of claims 1-6.

10. A computer program product, characterised in that, The computer executable instructions are used to make the computer execute the method of any one of claims 1-6.

Images