A fund sales ecological platform based on four-party collaborative operation

Abstract

This invention discloses a fund sales ecosystem platform based on four-party collaborative operation, belonging to the financial services field. It addresses the problem of information synchronization issues in existing fund sales ecosystem platforms. Through modules for custodians, institutions, financial advisors, clients, sales management, and dynamic management, this invention addresses the long-standing pain points of system fragmentation and operational inefficiency in the fund industry. By integrating resources in a "one-stop" manner, it achieves single sign-on, seamless end-to-end workflow, and seamless data transfer, completely changing the traditional model where users need to repeatedly switch between multiple systems and experience fragmented operational paths. The FSP platform periodically pulls data from multiple custodian institutions and multiple fund products for unified cleaning and aggregation, establishing standardized data asset management specifications. It supports structured and unstructured storage and processing, solving the data silo problem. Based on the data's value and accessibility, data is stored on different platform media, optimizing costs.

Description

Technical Field

[0001] This invention relates to the field of financial services technology, and in particular to a fund sales ecosystem platform based on four-party collaborative operation. Background Technology

[0002] The fund sales ecosystem platform integrates four core entities: fund custodians, sales / management institutions, financial advisors, and clients. It is a fund sales service system that achieves full-process collaboration through digital modules. Its core is to enable data interoperability, access control, and risk prevention, so as to make the sales process compliant, efficient, and accurate.

[0003] Currently, the fund sales ecosystem suffers from information silos, with fund managers, custodians, and sales agents each maintaining independent systems. This results in a lack of real-time synchronization of asset data, transaction instructions, and customer information. Furthermore, existing platforms mostly employ a post-event monitoring model, which cannot intercept abnormal transactions (such as related-party transactions or transfer of benefits) in real time, leading to delayed risk assessment. Finally, the sales side lacks custodian data support, making it difficult to provide investors with proof of asset security; and the management side cannot obtain accurate customer profiles from sales channels, resulting in inefficient product matching.

[0004] To address this, we propose a fund sales ecosystem platform based on collaborative operations among four parties. Summary of the Invention

[0005] The purpose of this invention is to provide a fund sales ecosystem platform based on four-party collaborative operation, which solves the problem of information synchronization in current fund sales ecosystem platforms in the background art.

[0006] To achieve the above objectives, this invention provides the following technical solution: a fund sales ecosystem platform based on four-party collaborative operation, comprising: a custodian module, used to access the standard OpenAPI of banks and securities custodian institutions, periodically extract fund net asset value and holding details data, adopt an incremental synchronization mechanism to obtain only changed data, perform data standardization processing on fields, and realize data interoperability among multiple custodian institutions; an institution module, used to receive applications from sales institutions and management institutions, review their business licenses and agency license qualification materials, and grant system permissions after approval, supporting financial advisors and investors to complete mobile account binding via mobile phone number + verification code; and a financial advisor module, used to manage information of sales personnel within the institutions and assign differentiated operations. The permissions module supports the tracking of operation records related to customers, products, and orders; the customer module is used to collect customer information, supporting manual entry and OCR recognition of ID card and bank card information, while establishing a customer tagging system to achieve automatic and manual tagging, and also conducting dynamic risk control rating and qualified investor certification, and anonymizing sensitive customer information; the sales management module is used to enter product information and authorize data, and can also generate and review pre-orders, as well as manage contracts; the dynamic management module is used to connect to the core databases of the custodian module, institution module, financial advisor module, customer module, and sales management module, supporting related queries and real-time data updates, and tracing the source of fund sales risks.

[0007] Furthermore, the sales management module includes a product module, an order module, and a contract module. The product module receives basic product information manually entered by the fund manager, supports the manager in selectively opening product data fields, and completes product data authorization and pre-listing review. The order module is used by financial advisors to generate pre-orders by associating customers, products, and subscription amounts on the client side. The system automatically verifies the matching between customer risk level and product risk level through the institutional risk control system. Large orders trigger a manual review process, and after approval, the order is pushed to the manager's end. The contract module stores sales contracts and distribution agreements related to the funds from the participating institutions, and supports contract classification query, version management, and operation log recording.

[0008] Furthermore, the dynamic management module includes a database construction module, a relational query module, a real-time update module, a personalized recommendation module, and a risk penetration module. The database construction module connects to the core databases of the custodian module, institution module, financial advisor module, customer module, and sales management module, while also accessing external data. It cleans heterogeneous data, standardizes field formats, hashes sensitive data, retains only publicly recognizable identifiers, and stores data in layers based on data value and accessibility, storing data on different platform media. Additionally, it isolates and partitions data, ensuring that each institution's data is only accessible to itself. The relational query module establishes queries... The index allows for querying data in the database, with secondary indexes created for frequently queried fields. Furthermore, the query process requires verification of the queryer's identity to ensure data security. The real-time update module updates the database when the custodian module updates fund net asset value or the customer module updates risk ratings, and also synchronizes low-frequency change data daily at midnight. The risk penetration module allows for rapid identification of relevant relationships and early warning when fund sales risks arise, through a correlation query module. The personalized recommendation module combines database data with dynamic display of suitable fund products based on investor risk preferences and historical trading behavior, enabling precise marketing.

[0009] Furthermore, the customer module introduces multi-dimensional dynamic risk factor modeling for customers. Based on the existing customer risk rating, and combined with customer interaction data, feature engineering and machine learning algorithms are used to generate a multi-dimensional dynamic risk preference vector for each customer. This vector includes the customer's sensitivity score to different risk factors and the maximum acceptable exposure threshold. This threshold will be dynamically adjusted according to the customer's wealth status, age, and market sentiment factors.

[0010] Furthermore, the custodian module introduces a multi-dimensional risk factor profile for the product. Utilizing the fund net asset value and holding details data extracted periodically by the custodian module, and combined with external data accessed by the database construction module, a multi-dimensional risk factor profile is created for each fund product. Using a factor analysis model, the risk of each fund is decomposed into individual risk factors corresponding to the customer's risk preference vector. Each product will obtain a multi-dimensional risk factor vector, and the product profile will be adjusted in real time as holdings change, market fluctuations occur, and macroeconomic data is updated.

[0011] Furthermore, the dynamic management module introduces a core matching and early warning engine. Combining the client's real-time holdings, it performs a thorough analysis of the client's entire investment portfolio, calculates its real-time multi-dimensional risk exposure vector, and then performs dynamic matching and deviation calculation. It compares the client's dynamic risk preference vector with its real-time multi-dimensional risk exposure vector in multiple dimensions. Subsequently, it introduces an intelligent correction strategy engine. When it finds that the risk exposure in a certain dimension deviates from the client's tolerable threshold, it immediately triggers an early warning. Combined with the personalized recommendation module, it intelligently analyzes the reasons for the client's deviation and generates personalized correction suggestions based on the client's risk tolerance and market conditions. The early warning and suggestions are pushed to the client and financial advisor through the real-time update module.

[0012] Furthermore, the dynamic risk preference vector of the customer ( This is considered as an expected risk distribution or an acceptable risk boundary, while the client's real-time portfolio multidimensional risk exposure vector ( This is considered as an actual risk distribution, and based on this, the Dynamic Risk Deviation Index (DRDI) is proposed. DRDI aims to quantify the degree of deviation between these two distributions across multidimensional risk factors, with a focus on the negative deviation where actual risk exceeds expected risk; and defines the client's dynamic risk preference vector. For the k-th customer, its dynamic risk preference vector This represents the upper limit of his risk exposure on N predefined risk factors; ;in Representing client k to the first The maximum acceptable exposure to each risk factor can be determined by a normalized threshold derived from customer questionnaires, historical behavior, wealth status, and AI analysis. This can be viewed as the client's target risk weight for each risk factor. To better align with the KL divergence, it can be considered as the proportion of the client's expected risk budget allocated to each risk factor, thus... Define the client's real-time portfolio multidimensional risk exposure vector. For k clients, their current investment portfolio is Actual exposure to each risk factor; in, Representing client k's portfolio in the 1st The actual risk exposure on each risk factor is normalized to make... Introducing the importance weight of risk factors ( ), indicating the first The importance of each risk factor to the overall risk assessment is dynamically adjusted based on macroeconomic conditions, industry changes, and the client's historical sensitivity to specific risk factors. Define an asymmetric penalty function. Used to measure actual risk Relative to expected risk The deviation, and when The punishment is relatively small when The punishment is relatively severe at times: ;in: Is the client's portfolio k in terms of risk factors The actual opening on; It is customer k pairs of risk factors The maximum allowable opening; It is a very small positive number used to avoid zero values ​​in the logarithmic function; This is a penalty coefficient used to amplify the punishment when actual risk exceeds expectations, and it is set according to business needs; the final formula is: Specifically, a smaller DRDI value indicates a better match between the client's real-time portfolio risk exposure and their dynamic risk appetite, resulting in a lower degree of risk deviation. Conversely, a larger DRDI value indicates a significant risk deviation, particularly when the actual exposure exceeds the acceptable limit. Therefore, a further DRDI threshold should be set. ,when When a risk warning is triggered, further analysis is conducted to determine which risk factors(s) are involved. Contributed the most This allows for targeted corrective suggestions.

[0013] Furthermore, the risk factors include underlying asset correlation risk factors and customer risk perception bias risk factors. Specifically, the underlying asset correlation risk factor is determined by extracting fund holding details through the custodian module, accessing external asset correlation matrix data through the database construction module, and defining the customer's acceptable exposure. The data was obtained through a decentralized customer preference questionnaire and AI analysis. Normalization was based on mapping 0-100% correlation to 0-1, satisfying... Define the actual average correlation exposure of the underlying assets in the client's current portfolio, i.e., the real-time portfolio exposure. The underlying asset details of all funds within the portfolio are extracted through the custody module. The database construction module is then connected to an external asset correlation matrix, and the Pearson coefficient is used to calculate the correlation of returns between assets. ,in For assets The weighting of holdings, For assets The correlation coefficient is normally taken as... The absolute value, After normalization, it satisfies Define factor weights Factor weights Dynamic adjustments are made based on the current macroeconomic industry chain concentration and the sensitivity of customers' historical losses due to asset linkages; an asymmetric penalty function is introduced. By introducing an exponential penalty term to amplify the negative deviation of high-correlation factors, the transformed asymmetric penalty function becomes: ;when Positive deviation, controllable correlation, mild penalty, when Negative deviation, excessive correlation, and index penalty; among them To avoid the logarithm being meaningless, The penalty coefficient is... The exponential coefficient, usually 10, is used to control the degree of nonlinearity penalty; factor components The risk penetration module identifies highly correlated assets, providing a corrective direction for the personalized recommendation module to replace uncorrelated assets. Factor parameters can be iterated in real time as the custodian module adjusts its holdings and external correlation data in the database is updated, perfectly adapting to the real-time update module's function. This ensures that risk assessment is synchronized with market changes and client holdings, making dynamic risk control ratings more granular and accurate.

[0014] Furthermore, the customer risk perception bias risk factor is used to measure the deviation risk between the customer's self-assessed risk tolerance and the actual risk tolerance calculated by AI. Specifically, it refers to the risk where the customer overestimates their own risk tolerance due to cognitive bias, causing the portfolio risk to exceed the actual acceptable range. This risk is assessed by collecting self-assessment questionnaire data and interaction behavior data through the customer module, and the actual risk tolerance can be calculated by AI. The customer's acceptable exposure is defined. It is obtained through the customer module, and the normalization basis is to map the 0-100% deviation rate to 0-1, which satisfies... Define real-time combined exposure. This refers to the actual value of the customer's current risk perception deviation, which is collected through the customer module to assess their self-assessed risk tolerance. Then, AI is used to calculate the customer's actual risk tolerance based on their wealth, age, historical loss response, and browsing behavior. ,Right now Define factor weights The value is derived from the dynamic management module and is dynamically adjusted based on the current market risk level and the frequency of position adjustments caused by historical cognitive biases of clients. The asymmetric penalty function is modified by introducing a piecewise penalty, resulting in the following asymmetric penalty function: ;when Positive deviation, controllable deviation, mild penalty, when Mild negative deviation, basic penalty, when Severe negative deviation will be punished more severely. To avoid the logarithm being meaningless, This is a mild penalty coefficient. The severe penalty coefficient is typically twice the mild penalty coefficient; therefore, the factor components... It can proactively mitigate behavioral risks by triggering warnings when customers plan to buy high-risk products due to cognitive biases, rather than intervening after the holdings exceed the risk level. After the warning, it can link with the customer module to push risk perception calibration questionnaires and historical loss cases, thereby dynamically reducing cognitive biases through customer education and achieving a closed loop of risk warning and perception correction.

[0015] Furthermore, the product module introduces a data authorization and review method, including the following steps: receiving basic product information entered by the fund manager; determining permissions for data fields based on a multi-dimensional permission calculation model; selecting the appropriate review path based on the permission determination result; executing a multi-level review process and applying a business rule engine for verification; and performing status transitions and permission configuration based on the review result. The multi-dimensional permission calculation model uses the following formula to calculate the permission score: Permission Score = Field Sensitivity Coefficient * Role Permission Coefficient * Business Scenario Coefficient * Time Sensitivity Coefficient, where the field sensitivity coefficient ranges from 1.0 to 3.0, the role permission coefficient ranges from 0.1 to 1.0, the business scenario coefficient ranges from 0.8 to 1.5, and the time sensitivity coefficient ranges from 1.0 to 2.0. Simultaneously, permission determination thresholds are set, where the automatic rejection threshold is set to 0.3 points, and the manual review threshold is set to 0.6 points. The automatic approval threshold is set to 0.9 points. The review path selection is based on the following conditions: when the product type is equity and the risk level is ≥R4, the mandatory risk control review path is selected; when the modification involves fee rates or investment scope, the compliance-focused review path is selected; when an urgent listing application is submitted and the historical review score is ≥4.5, the fast track review path is selected. The business engine includes the following verification rules: Risk disclosure consistency verification: when the risk level is R5 and the information disclosure completeness is <90%, it is automatically rejected; Fee structure rationality verification: when the total fee rate exceeds 3%, a warning is issued; Liquidity requirement compliance verification: when the liquidity ratio of the money market fund is <30%, it is automatically rejected; The status transition includes 9 statuses: draft, submitted for review, initial review in progress, secondary review in progress, final review in progress, approved, rejected, listed, and delisted. Each status transition has strict entry conditions and processing logic.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention proposes a fund sales ecosystem platform based on four-party collaborative operation. In existing technologies, information in fund sales ecosystem platforms cannot be synchronized. However, this invention addresses the long-standing pain points of system fragmentation and inefficient operation in the fund industry through custodian modules, institutional modules, financial advisor modules, customer modules, sales management modules, and dynamic management modules. It achieves single sign-on, end-to-end connectivity, and seamless data flow through "one-stop" resource integration, completely changing the traditional mode where users need to repeatedly switch between multiple systems and experience fragmented operation paths. The FSP platform periodically pulls data (net asset value, share, etc.) from multiple custodian institutions and multiple fund products for unified cleaning and aggregation, establishing standardized data asset management specifications. It supports the storage and processing of structured and unstructured data (such as agreements, contracts, and announcements), solving the data silo problem. Based on the value and popularity of the data, it stores data on different platform media (e.g., hot data is stored in a cache database, and cold data is stored in low-cost storage), optimizing costs. Attached Figure Description

[0017] Figure 1 This is a flowchart of the overall program of the fund sales ecosystem platform based on four-party collaborative operation, as described in this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] To address the technical issue of information synchronization problems on existing fund sales platforms, such as... Figure 1As shown, the following preferred technical solution is provided: a fund sales ecosystem platform based on four-party collaborative operation, including: a custodian module, used to access the standard OpenAPI of banks and securities custodian institutions, periodically extract fund net asset value and holding details data, adopt an incremental synchronization mechanism to only obtain changed data, perform data standardization processing on fields, and realize data interoperability among multiple custodian institutions; an institution module, used to receive onboarding applications from sales institutions and management institutions, review their business licenses and agency license qualification materials, and grant system permissions after approval, supporting financial advisors and investors to complete mobile account binding via mobile phone number + verification code; and a financial advisor module, used to manage information of sales personnel within onboarding institutions and assign differentiated operation permissions. It supports the traceability of operation records related to customers, products, and orders; the customer module is used to collect customer information, supporting manual entry and OCR recognition of ID card and bank card information, and establishing a customer tagging system to achieve automatic and manual tagging. In addition, it conducts dynamic risk control rating and qualified investor certification, and anonymizes sensitive customer information; the sales management module is used to enter product information and authorize data, and can also generate and review pre-orders, as well as manage contracts; the dynamic management module is used to connect with the core databases of the custodian module, institution module, financial advisor module, customer module, and sales management module, supporting related queries and real-time data updates, and tracing the source of fund sales risks. The sales management module includes a product module, an order module, and a contract module. The product module receives basic product information manually entered by fund managers, allows managers to selectively open product data fields, and completes product data authorization and pre-listing review. The order module allows financial advisors to generate pre-orders by associating clients, products, and subscription amounts on the client side. The system automatically verifies the matching between the client's risk level and the product's risk level through the institutional risk control system. Large orders trigger a manual review process, and after approval, the order is pushed to the manager's end. The contract module stores sales contracts and distribution agreements related to the funds from participating institutions, and supports contract category query, version management, and operation log recording.

[0020] The dynamic management module includes a database construction module, a relational query module, a real-time update module, a personalized recommendation module, and a risk penetration module. The database construction module connects to the core databases of the custodian, institutional, financial advisor, customer, and sales management modules, while also accessing external data. It cleans heterogeneous data, standardizes field formats, hashes sensitive data, retains only publicly related identifiers, and implements tiered data storage based on data value and accessibility, storing data on different platform media. Furthermore, it isolates and partitions data, ensuring that each institution's data is only accessible to itself. The relational query module establishes query indexes, enabling queries on data within the database. For frequently queried fields, secondary indexes are created. Additionally, during the query process, the identity of the queryer must be verified first to ensure data security. A real-time update module updates the database information when the custodian module updates the fund's net asset value or the client module updates its risk rating. It also synchronizes low-frequency change data daily at midnight, such as the remaining days of the institution's qualification validity period and adjustments to the industry classification of underlying assets. A risk penetration module allows for rapid identification of relevant relationships and early warning when fund sales risks arise, through a correlation query module. A personalized recommendation module combines data from the database accessed to dynamically display suitable fund products based on investors' risk preferences and historical trading behavior, achieving precise marketing.

[0021] The customer module introduces multi-dimensional dynamic risk factor modeling for customers. Based on the existing customer risk rating, it combines customer interaction data and uses feature engineering and machine learning algorithms to generate a multi-dimensional dynamic risk preference vector for each customer. This vector includes the customer's sensitivity score to different risk factors and the maximum acceptable exposure threshold. This threshold is dynamically adjusted according to the customer's wealth status, age, and market sentiment factors. The vector is also dynamically updated based on the customer's subsequent trading behavior, browsing behavior, and even communication with financial advisors.

[0022] The custodian module introduces a multi-dimensional risk factor profile for products. It uses fund net asset value and holding details data extracted periodically by the custodian module, combined with external data accessed by the database construction module, to create a multi-dimensional risk factor profile for each fund product. Using a factor analysis model, the risk of each fund is decomposed into individual risk factors corresponding to the customer's risk preference vector. Each product will obtain a multi-dimensional risk factor vector, and the product profile will be adjusted in real time as holdings change, market fluctuations, and macroeconomic data updates.

[0023] The dynamic management module introduces a core matching and early warning engine. By combining the client's real-time holdings, it performs a thorough analysis of the client's entire investment portfolio, calculates its real-time multi-dimensional risk exposure vector, and then performs dynamic matching and deviation calculation. It compares the client's dynamic risk preference vector with its real-time multi-dimensional risk exposure vector. Subsequently, it introduces an intelligent correction strategy engine. When it finds that the risk exposure in a certain dimension deviates from the client's tolerable threshold, it immediately triggers an early warning. Combined with the personalized recommendation module, it intelligently analyzes the reasons for the client's deviation and generates personalized correction suggestions based on the client's risk tolerance and market conditions. The early warning and suggestions are pushed to the client and financial advisor through the real-time update module.

[0024] The customer's dynamic risk preference vector ( This is considered as an expected risk distribution or an acceptable risk boundary, while the client's real-time portfolio multidimensional risk exposure vector ( This is considered as an actual risk distribution, and based on this, the Dynamic Risk Deviation Index (DRDI) is proposed. DRDI aims to quantify the degree of deviation between these two distributions across multidimensional risk factors, with a focus on the negative deviation where actual risk exceeds expected risk; and defines the client's dynamic risk preference vector. For the k-th customer, its dynamic risk preference vector This represents the upper limit of his risk exposure on N predefined risk factors; ;in Representing client k to the first The maximum acceptable exposure to each risk factor can be determined by a normalized threshold derived from customer questionnaires, historical behavior, wealth status, and AI analysis. This can be viewed as the client's target risk weight for each risk factor. To better align with the KL divergence, it can be considered as the proportion of the client's expected risk budget allocated to each risk factor, thus... Define the client's real-time portfolio multidimensional risk exposure vector. For k clients, their current investment portfolio is Actual exposure to each risk factor; ;in, Representing client k's portfolio in the 1st The actual risk exposure on each risk factor is normalized to make... Introducing the importance weight of risk factors ( ), indicating the first The importance of each risk factor to the overall risk assessment is dynamically adjusted based on macroeconomic conditions, industry changes, and the client's historical sensitivity to specific risk factors. Define an asymmetric penalty function. Used to measure actual risk Relative to expected risk The deviation, and when The punishment is relatively small when The punishment is relatively severe at times: ;in: Is the client's portfolio k in terms of risk factors The actual opening on; It is customer k pairs of risk factors The maximum allowable opening; It is a very small positive number used to avoid zero values ​​in the logarithmic function; This is a penalty coefficient used to amplify the punishment when actual risk exceeds expectations, and it is set according to business needs; the final formula is: Specifically, a smaller DRDI value indicates a better match between the client's real-time portfolio risk exposure and their dynamic risk appetite, resulting in a lower degree of risk deviation. Conversely, a larger DRDI value indicates a significant risk deviation, particularly when the actual exposure exceeds the acceptable limit. Therefore, a further DRDI threshold should be set. ,when When a risk warning is triggered, further analysis is conducted to determine which risk factors(s) are involved. Contributed the most This allows for targeted corrective suggestions.

[0025] Because traditional KL divergence is symmetric, DRDI, through its customized f(x,y) function, focuses more on the negative deviation of actual risk exceeding expectations, which better meets the business needs of financial risk management. DRDI can incorporate multi-dimensional risk factors of customers and products and consider the dynamic changes of these factors, solving the problem of the static nature and granularity mismatch of traditional risk assessment. The introduced weights allow the system to dynamically adjust the importance of different risk factors according to market environment, macroeconomic conditions, or customer characteristics, making risk assessment more flexible and adaptable.

[0026] For example: assuming the customer It exceeded the threshold, and the following was calculated: Technology Stock Concentration Risk Factor Contribution Bond duration risk factor Contribution Liquidity risk factor for small-cap stocks Contribution The contribution of the technology stock concentration risk factor is far greater than that of other factors, indicating that clients' actual exposure to technology stocks is significant. It significantly exceeded his tolerance limit. Furthermore, the concentration risk of technology stocks is a very important risk point in the current market or in the risk perception of clients. Therefore, the system will first suggest that clients appropriately reduce their allocation to technology funds.

[0027] Risk factors include underlying asset correlation risk factors and client risk perception bias risk factors. Specifically, the underlying asset correlation risk factor is determined by extracting fund holding details through the custodian module and accessing external asset correlation matrix data through the database construction module to define the client's acceptable exposure. The data was obtained through a decentralized customer preference questionnaire and AI analysis. Normalization was based on mapping 0-100% correlation to 0-1, satisfying... Define the actual average correlation exposure of the underlying assets in the client's current portfolio, i.e., the real-time portfolio exposure. The underlying asset details of all funds within the portfolio are extracted through the custody module. The database construction module is then connected to an external asset correlation matrix, and the Pearson coefficient is used to calculate the correlation of returns between assets. ,in For assets The weighting of holdings, For assets The correlation coefficient is normally taken as... The absolute value, After normalization, it satisfies Define factor weights Factor weights Dynamic adjustments are made based on the current macroeconomic industry chain concentration and the sensitivity of customers' historical losses due to asset linkages; an asymmetric penalty function is introduced. By introducing an exponential penalty term to amplify the negative deviation of high-correlation factors, the transformed asymmetric penalty function becomes: ;when Positive deviation, controllable correlation, mild penalty, when Negative deviation, excessive correlation, and index penalty; among them To avoid the logarithm being meaningless, The penalty coefficient is... The exponential coefficient, usually 10, is used to control the degree of nonlinearity penalty; factor components The risk penetration module identifies highly correlated assets, providing a corrective direction for the personalized recommendation module to replace uncorrelated assets. Factor parameters can be iterated in real time as the custodian module adjusts its holdings and external correlation data in the database is updated, perfectly adapting to the real-time update module's function. This ensures that risk assessment is synchronized with market changes and client holdings, making dynamic risk control ratings more granular and accurate.

[0028] The customer risk perception bias risk factor measures the discrepancy between a customer's self-assessed risk tolerance and the actual risk tolerance calculated by AI. Specifically, it refers to the risk of a customer overestimating their risk tolerance due to cognitive bias, leading to portfolio risk exceeding their actual tolerance level. This risk is assessed by collecting self-assessment questionnaire data and interaction behavior data through the customer module, and AI can then calculate the actual risk tolerance. The customer's acceptable exposure is defined. It is obtained through the customer module, and the normalization basis is to map the 0-100% deviation rate to 0-1, which satisfies... Define real-time combined exposure. This refers to the actual value of the customer's current risk perception deviation, which is collected through the customer module to assess their self-assessed risk tolerance. Then, AI is used to calculate the customer's actual risk tolerance based on their wealth, age, historical loss response, and browsing behavior. ,Right now Define factor weights The value is derived from the dynamic management module and is dynamically adjusted based on the current market risk level and the frequency of position adjustments caused by historical cognitive biases of clients. The asymmetric penalty function is modified by introducing a piecewise penalty, resulting in the following asymmetric penalty function: ;when Positive deviation, controllable deviation, mild penalty, when Mild negative deviation, basic penalty, when Severe negative deviation will be punished more severely. To avoid the logarithm being meaningless, This is a mild penalty coefficient. The severe penalty coefficient is typically twice the mild penalty coefficient; therefore, the factor components... It can proactively mitigate behavioral risks by triggering warnings when customers plan to buy high-risk products due to cognitive biases, rather than intervening after the holdings exceed the risk level. After the warning, it can link with the customer module to push risk perception calibration questionnaires and historical loss cases, thereby dynamically reducing cognitive biases through customer education and achieving a closed loop of risk warning and perception correction.

[0029] In the order module, the subscription and purchase of fund products are as follows: After registering and logging into the client (including but not limited to mini-programs, mobile applications, websites, etc.), investors complete risk rating and qualified investor authentication, and place an order after completing product matching; the salesperson logs into the system to review the investor's order (qualified investor, share, etc.), and pushes the order to the manager after confirming that the subscription qualifications are met; the manager logs into the system to conduct a second review of the order and investor, and pushes the order and investor information to the custody platform after confirming that the subscription conditions are met; after the manager confirms the share, the status is synchronized back to the manager, and the product order is completed.

[0030] In the product module, a data authorization and review method is introduced, including the following steps: the fund manager connects to the custody platform's OpenAPI to complete the collection of its own products; the manager discloses basic information about its own products; the sales agent collects fund products and completes the signing of a distribution agreement; the sales agent completes the internal product collection access approval process; finally, the product is listed for sale; investors view and purchase fund products after conducting risk control rating and qualified investor certification through the client provided by the sales agent; permissions are determined for data fields based on a multi-dimensional permission calculation model; the corresponding review path is selected based on the permission determination results; a multi-level review process is executed and business rules are applied. The engine then performs verification; based on the review results, it performs status transitions and permission configurations; the multi-dimensional permission calculation model uses the following formula to calculate the permission score: Permission Score = Field Sensitivity Coefficient * Role Permission Coefficient * Business Scenario Coefficient * Time Sensitivity Coefficient, where the field sensitivity coefficient ranges from 1.0 to 3.0, the role permission coefficient ranges from 0.1 to 1.0, the business scenario coefficient ranges from 0.8 to 1.5, and the time sensitivity coefficient ranges from 1.0 to 2.0; at the same time, permission judgment thresholds are set, where: the automatic rejection threshold is set to 0.3 points, the manual review threshold is set to 0.6 points, and the automatic approval threshold is set to 0.9 points.

[0031] The review path selection is based on the following conditions: when the product type is equity and the risk level is ≥R4, the mandatory risk control review path is selected; when the modification involves fee rates or investment scope, the compliance-focused review path is selected; when an urgent listing application is submitted and the historical review score is ≥4.5, the fast track review path is selected.

[0032] The business engine includes the following verification rules: Risk disclosure consistency verification: Automatic rejection when the risk level is R5 and the information disclosure completeness is <90%; Fee structure rationality verification: Warning is issued when the total fee rate exceeds 3%; Liquidity requirement compliance verification: Automatic rejection when the liquidity ratio of the money market fund is <30%; Status transition includes 9 statuses: draft, submitted for review, preliminary review in progress, secondary review in progress, final review in progress, approved, rejected, listed, and delisted. Each status transition has strict entry conditions and processing logic.

[0033] By calculating multi-dimensional coefficients, refined permission determination is achieved. At the same time, based on threshold settings, it can automatically make decisions to approve, reject, or require manual review, reducing subjective judgment and improving the objectivity and consistency of permission management. In addition, differentiated review paths are adopted for products with different risk levels, and mandatory risk control review is implemented for high-risk products.

[0034] Specifically, addressing the long-standing pain points of system fragmentation and operational inefficiency in the fund industry, the platform achieves single sign-on, seamless end-to-end workflow, and seamless data transfer through "one-stop" resource integration. This completely changes the traditional model where users have to repeatedly switch between multiple systems and experience fragmented operational paths. The FSP platform periodically pulls data (net asset value, share, etc.) from multiple custodian institutions and fund products for unified cleaning and aggregation, establishing standardized data asset management specifications. It supports the storage and processing of structured and unstructured data (such as agreements, contracts, and announcements), solving the data silo problem. Based on the value and accessibility of the data, it stores data on different platform media (e.g., hot data is stored in a cache database, and cold data is stored in low-cost storage), optimizing costs. By establishing a unified data tenant space, the FSP platform achieves data resource isolation among various institutions within the platform, ensuring data security, ensuring data is released on demand, and preventing the leakage of sensitive information.

[0035] Based on standardized data, the FSP platform provides capabilities such as data development, data integration, and data management to support the sales system in obtaining consistent product information (such as fund net asset value, holding ratio, risk indicators, etc.).

[0036] By combining a customer rating system and a qualified investor certification system, the FSP platform can dynamically display suitable fund products based on tags such as investors' risk preferences and historical trading behavior, thereby achieving precise marketing.

[0037] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A fund sales ecosystem platform based on four-party collaborative operation, characterized in that, include: The Custodian module is used to access the standard OpenAPI of banks and securities custodian institutions, extract fund net asset value and holding details data on a regular basis, and adopt an incremental synchronization mechanism to only obtain changed data. The data fields are processed in a standardized manner to achieve data interoperability among multiple custodian institutions. The Institution module is used to receive onboarding applications from sales institutions and management institutions, review their business licenses and agency license qualification materials, and grant system permissions after the review is approved. It supports financial advisors and investors to complete mobile account binding via mobile phone number + verification code. The financial advisor module is used to manage information on sales personnel within participating institutions, assign differentiated operation permissions, and support the traceability of operation records related to customers, products, and orders. The customer module is used to collect customer information, supporting manual entry and OCR recognition of ID card and bank card information. It also establishes a customer tagging system to achieve automatic and manual tagging, and conducts dynamic risk control rating and qualified investor certification, while de-identifying sensitive customer information. The sales management module is used to input product information, authorize data, generate and review pre-orders, and manage contracts. The dynamic management module connects to the core databases of the custodian, institution, financial advisor, and customer modules, supporting related queries and real-time data updates, as well as tracing the source of fund sales risks. The customer module introduces multi-dimensional dynamic risk factor modeling for customers. Based on existing customer risk ratings and combined with customer interaction data, feature engineering and machine learning algorithms are used to generate a multi-dimensional dynamic risk preference vector for each customer. This vector includes the customer's sensitivity score to different risk factors and the maximum acceptable exposure threshold, which is dynamically adjusted according to the customer's wealth, age, and market sentiment.

2. The fund sales ecosystem platform based on four-party collaborative operation as described in claim 1, characterized in that: The sales management module includes a product module, an order module, and a contract module. The product module is used to receive basic product information manually entered by the fund manager, and supports the manager to selectively open product data fields to complete product data authorization and pre-listing review. The order module is used by financial advisors to generate pre-orders by associating clients, products, and subscription amounts on the client side. The institutional risk control system automatically verifies the matching between the client's risk level and the product's risk level. Large orders trigger a manual review process, and after approval, the order is pushed to the manager's end. The contract module is used to store sales contracts and distribution agreements related to funds between the participating institutions and the fund. It supports contract category query, version management, and operation log recording.

3. A fund sales ecosystem platform based on four-party collaborative operation as described in claim 2, characterized in that: The dynamic management module includes a database construction module, a relational query module, a real-time update module, a personalized recommendation module, and a risk penetration module. The database construction module connects to the core databases of the custodian module, institution module, financial advisor module, customer module, and sales management module, while also accessing external data. It cleans heterogeneous data, standardizes field formats, hashes sensitive data, retains only publicly related identifiers, and stores data in layers based on data value and accessibility, storing data on different platform media. Furthermore, it isolates and partitions data, ensuring that each institution's data is only accessible to itself. The relational query module establishes query indexes to query data in the database. It creates secondary indexes for frequently queried fields and verifies the queryer's identity beforehand to ensure data security. The real-time update module updates the database information when the custodian module updates fund net asset value or the customer module updates risk ratings, and also synchronizes low-frequency changed data daily at midnight. The risk penetration module allows for quick identification of relevant relationships and early warning when fund sales risks arise, through the correlation query module. The personalized recommendation module combines data from the database to dynamically display suitable fund products based on investors' risk preferences and historical trading behavior, achieving precise marketing.

4. A fund sales ecosystem platform based on four-party collaborative operation as described in claim 3, characterized in that: The custodian module introduces a multi-dimensional risk factor profile for each product. It uses fund net asset value and holding details data extracted periodically by the custodian module, combined with external data accessed by the database construction module, to create a multi-dimensional risk factor profile for each fund product. Using a factor analysis model, the risk of each fund is decomposed into individual risk factors corresponding to the customer's risk preference vector. Each product will obtain a multi-dimensional risk factor vector, and the product profile will be adjusted in real time as holdings change, market fluctuations, and macroeconomic data updates occur.

5. A fund sales ecosystem platform based on four-party collaborative operation as described in claim 4, characterized in that: The dynamic management module incorporates a core matching and early warning engine. Combined with the client's real-time holdings, it performs a thorough analysis of the client's entire investment portfolio, calculating its real-time multi-dimensional risk exposure vector. Next, it performs dynamic matching and deviation calculation, comparing the client's dynamic risk preference vector with its real-time multi-dimensional risk exposure vector. Then, it introduces an intelligent correction strategy engine. When a risk exposure in a certain dimension deviates from the client's acceptable threshold, an early warning is immediately triggered. Combined with a personalized recommendation module, it intelligently analyzes the reasons for the client's deviation and generates personalized correction suggestions based on the client's risk tolerance and market conditions. The warnings and suggestions are pushed to the client and financial advisor through a real-time update module.

6. A fund sales ecosystem platform based on four-party collaborative operation as described in claim 5, characterized in that: The dynamic risk preference vector of a client is regarded as an expected risk distribution or an acceptable risk boundary, while the client's real-time portfolio multidimensional risk exposure vector is regarded as an actual risk distribution. Based on this, the Dynamic Risk Deviation Index (DRDI) is proposed. The DRDI aims to quantify the degree of deviation between these two distributions on multidimensional risk factors, and focuses on the negative deviation of actual risk from expected risk. Define the customer's dynamic risk preference vector For the k-th customer, its dynamic risk preference vector This represents the upper limit of his risk exposure on N predefined risk factors; ;in Representing client k to the first The maximum exposure to each risk factor can be determined by a normalized threshold derived from customer questionnaires, historical behavior, wealth status, and AI analysis. This is considered as the client's target risk weight for each risk factor. To better align with the KL divergence, it is viewed as the proportion of the client's desired risk budget allocated to each risk factor. Define the client's real-time portfolio multidimensional risk exposure vector. For k clients, their current investment portfolio is Actual exposure to each risk factor; ;in, Representing client k's portfolio in the 1st The actual risk exposure on each risk factor is normalized to make... Introducing the importance weight of risk factors, representing the importance of the first risk factor. The importance of each risk factor to the overall risk assessment is dynamically adjusted based on macroeconomic conditions, industry changes, and the client's historical sensitivity to specific risk factors. Define an asymmetric penalty function. Used to measure actual risk Relative to expected risk The deviation, and when The punishment is relatively small when The punishment is relatively severe at times: ;in: Is the client's portfolio k in terms of risk factors The actual opening on; It is customer k pairs of risk factors The maximum allowable opening; It is a very small positive number used to avoid zero values ​​in the logarithmic function; This is a penalty coefficient used to amplify the punishment when actual risk exceeds expectations; it is set according to business needs. The final formula is: Specifically, a smaller DRDI value indicates a closer match between the client's real-time portfolio risk exposure and their dynamic risk appetite, resulting in a lower degree of risk deviation. Conversely, a larger DRDI value indicates a significant risk deviation, and when the actual exposure exceeds the acceptable upper limit, a further DRDI threshold is set. ,when When a risk warning is triggered, further analysis is conducted to determine which risk factors(s) are involved. Contributed the most This allows for targeted corrective suggestions.

7. A fund sales ecosystem platform based on four-party collaborative operation as described in claim 6, characterized in that: The risk factors include underlying asset correlation risk factors and client risk perception bias risk factors. Specifically, the underlying asset correlation risk factor is determined by extracting fund holding details through the custodian module, accessing external asset correlation matrix data through the database construction module, and defining the client's acceptable exposure. The data was obtained through a decentralized customer preference questionnaire and AI analysis. Normalization was based on mapping 0-100% correlation to 0-1, satisfying... Define the actual average correlation exposure of the underlying assets in the client's current portfolio, i.e., the real-time portfolio exposure. The underlying asset details of all funds within the portfolio are extracted through the custody module. The database construction module is then connected to an external asset correlation matrix, and the Pearson coefficient is used to calculate the correlation of returns between assets. ,in For assets The weighting of holdings, For assets The correlation coefficient is normally taken as... The absolute value, After normalization, it satisfies Define factor weights Factor weights Dynamic adjustments are made based on the current macroeconomic industry chain concentration and the sensitivity of customers' historical losses due to asset linkages; an asymmetric penalty function is introduced. By introducing an exponential penalty term to amplify the negative deviation of high-correlation factors, the transformed asymmetric penalty function becomes: ;when Positive deviation, controllable correlation, mild penalty, when Negative deviation, excessive correlation, and index penalty; among them To avoid the logarithm being meaningless, The penalty coefficient is... The exponential coefficient, usually 10, is used to control the degree of nonlinearity penalty; factor components .

8. A fund sales ecosystem platform based on four-party collaborative operation as described in claim 7, characterized in that: The customer risk perception bias risk factor is used to measure the deviation risk between the customer's self-assessed risk tolerance and the actual risk tolerance calculated by AI. That is, the customer overestimates their own risk tolerance due to cognitive bias, resulting in the risk of the investment portfolio exceeding the actual tolerable range. The actual risk tolerance can be calculated by AI through the collection of self-assessment questionnaire data and interaction behavior data through the customer module. Define the customer's acceptable exposure It is obtained through the customer module, and the normalization basis is to map the 0-100% deviation rate to 0-1, which satisfies... Define real-time combined exposure. This refers to the actual value of the customer's current risk perception deviation, which is collected through the customer module to assess their self-assessed risk tolerance. Then, AI is used to calculate the customer's actual risk tolerance based on their wealth, age, historical loss response, and browsing behavior. ,Right now Define factor weights The value is derived from the dynamic management module and is dynamically adjusted based on the current market risk level and the frequency of position adjustments caused by historical cognitive biases of clients. The asymmetric penalty function is modified by introducing a piecewise penalty, resulting in the following asymmetric penalty function: ;when Positive deviation, controllable deviation, mild penalty, when Mild negative deviation, basic penalty, when Severe negative deviation will be punished more severely. To avoid the logarithm being meaningless, This is a mild penalty coefficient. The severe penalty coefficient is typically twice the mild penalty coefficient; therefore, the factor components... .

9. A fund sales ecosystem platform based on four-party collaborative operation as described in claim 8, characterized in that: In the product module, a data authorization and review method is introduced, including the following steps: receiving basic product information entered by the fund manager; determining permissions for data fields based on a multi-dimensional permission calculation model; selecting the appropriate review path based on the permission determination result; executing a multi-level review process and applying a business rule engine for verification; and configuring status transitions and permissions based on the review results. The multi-dimensional permission calculation model uses the following formula to calculate the permission score: Permission Score = Field Sensitivity Coefficient * Role Permission Coefficient * Business Scenario Coefficient * Time Sensitivity Coefficient, where the field sensitivity coefficient ranges from 1.0 to 3.0, the role permission coefficient ranges from 0.1 to 1.0, the business scenario coefficient ranges from 0.8 to 1.5, and the time sensitivity coefficient ranges from 1.0 to 2.

0. Simultaneously, permission judgment thresholds are set, where the automatic rejection threshold is set to 0.3 points, the manual review threshold is set to 0.6 points, and the automatic... The approval threshold is set at 0.9 points. The review path selection is based on the following conditions: when the product type is equity and the risk level is ≥R4, the mandatory risk control review path is selected; when the modification involves fee rates or investment scope, the compliance-focused review path is selected; when an urgent listing application is submitted and the historical review score is ≥4.5, the fast track review path is selected. The business engine includes the following verification rules: Risk disclosure consistency verification: when the risk level is R5 and the information disclosure completeness is <90%, it is automatically rejected; Fee structure rationality verification: when the total fee rate exceeds 3%, a warning is issued; Liquidity requirement compliance verification: when the liquidity ratio of the money market fund is <30%, it is automatically rejected; The status transition includes 9 statuses: draft, submitted for review, initial review in progress, secondary review in progress, final review in progress, approved, rejected, listed, and delisted. Each status transition has strict entry conditions and processing logic.

Images