Financial knowledge base-based digital intelligent ecological group monitoring model optimization method and system

Abstract

The invention discloses a digital intelligent ecological group monitoring model optimization method and system based on a financial knowledge base, and belongs to the technical field of monitoring model optimization. Calling the key behavior data from the financial data platform, and constructing a transaction data pair; calculating a transaction behavior characteristic value of the user, and constructing a financial knowledge label library in combination with the structured knowledge label and a behavior finance theoretical index; through matching of financial knowledge tags, a tag-driven user group is constructed, and a personalized monitoring threshold interval is calculated; and on the basis of the personalized monitoring threshold interval, monitoring whether the transaction behavior characteristic values of all the users in the label-driven user group at the next transaction time exist in the personalized monitoring threshold interval in real time, and if not, performing early warning and dynamically updating the label-driven user group and the personalized monitoring threshold interval. According to the method, the real-time performance and accuracy of financial behavior monitoring are improved, and dynamic risk identification and intelligent regulation and control from individuals to groups can be realized.

Description

Technical Field

[0001] The present invention relates to the field of monitoring model optimization technology, and specifically to a method and system for optimizing a digital intelligence ecological group monitoring model based on a financial knowledge base. Background Art

[0002] With the rapid development of the financial market and the continuous improvement of its intelligence level, the digital financial ecosystem has gradually become one of the mainstream research directions. In this context, intelligent monitoring methods that integrate artificial intelligence, knowledge graphs, and behavioral finance theory have gradually attracted attention, especially in the areas of group identification and dynamic risk monitoring driven by financial knowledge bases. In recent years, some technologies have attempted to classify and model the behavioral patterns of financial users through machine learning methods, but most methods still focus on static rule matching and unstructured data mining, lacking in-depth modeling and explanatory support for the cognitive mechanisms of behavioral finance. In addition, traditional financial monitoring models often use all users as units, ignoring the significant heterogeneity of user groups in cognitive preferences, behavioral tendencies, and transaction characteristics. This causes the description of group characteristics to deviate from the actual behavior of individuals, thereby affecting the accuracy of early warning effects and the timeliness of responses.

[0003] Existing technologies still have limitations in terms of group segmentation criteria, user behavior modeling, and dynamic threshold setting mechanisms. First, current models often rely on single-dimensional indicators such as transaction frequency or asset size to segment user groups. They lack the combined modeling of structured knowledge labels and behavioral finance theory, making it difficult to accurately characterize users' psychological characteristics and decision-making mechanisms. Second, the monitoring thresholds for abnormal personalized trading behavior generally rely on fixed parameters and lack an adaptive update mechanism based on group distribution statistics, making it difficult to cope with behavioral pattern changes caused by market fluctuations. Finally, in terms of risk warning mechanisms, existing methods are mostly based on anomaly detection or outlier analysis, lacking dynamic feedback and threshold correction strategies based on the evolution of user groups. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for optimizing a digital ecological group monitoring model based on a financial knowledge base to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] The optimization method of the digital ecological group monitoring model based on the financial knowledge base includes the following steps: Step S1: after the user's authorization, the user's position change data, transaction amount data, transaction frequency data, transaction type data and transaction time are retrieved from the financial data platform; based on the transaction time, the position change data, transaction amount data, transaction frequency data and transaction type data are time-series aligned to construct transaction data pairs; Step S2: based on the financial database, the structured knowledge labels and behavioral finance theory indicators in the financial field are extracted; based on the transaction type data, the corresponding structured knowledge labels are extracted; the user's transaction behavior feature values ​​are calculated to Based on the behavioral characteristic values, the corresponding behavioral finance theory indicators are extracted; the extracted structured knowledge labels and behavioral finance theory indicators are used as the user's financial knowledge label library; step S3: obtain the financial knowledge label library of all users and build a label-driven user group set; calculate the personalized monitoring threshold interval of all users in the label-driven user group; step S4: based on the personalized monitoring threshold interval, real-time monitoring is performed to determine whether the transaction behavior characteristic values ​​of all users in the label-driven user group at the next transaction time are within the personalized monitoring threshold interval. If not, an early warning is issued and the label-driven user group and the personalized monitoring threshold interval are dynamically updated.

[0007] As a preferred solution of the method for optimizing the digital ecological group monitoring model based on the financial knowledge base described in the present invention, after user authorization, the user's financial behavior log is retrieved from the financial data platform. The financial behavior log includes the user's position change data, transaction amount data, transaction frequency data, transaction type data, and transaction time on the financial data platform; the position change data, transaction amount data, transaction frequency data, and transaction type data are standardized and noise-cleaned;

[0008] Based on the transaction time, the position change data, transaction amount data, transaction frequency data and transaction type data are aligned in time series, and transaction data pairs are constructed. The transaction data pair corresponding to the i-th transaction time is recorded as {PC i ,TA i ,TF i ,TT i}, where PC i Indicates the position change data corresponding to the i-th trading time, TA i Indicates the transaction amount data corresponding to the i-th transaction time, TF i Indicates the transaction frequency data corresponding to the i-th transaction time, TT i Indicates the transaction type data corresponding to the i-th transaction time.

[0009] As a preferred solution of the digital intelligence ecological group monitoring model optimization method based on the financial knowledge base of the present invention, based on the financial database, the structured knowledge labels and behavioral finance theory indicators in the financial field are extracted from the financial database, and a set of structured knowledge labels and a set of behavioral finance theory indicators are constructed. i ,TA i ,TF i ,TT i}, build a financial knowledge label library, in which one type of transaction type data corresponds to at least one behavioral finance theory indicator and at least one structured knowledge label, (structured knowledge labels include typical transaction feature labels, user behavior pattern labels, risk level classification labels and psychological behavior labels, etc. Behavioral finance theory indicators include: high-frequency short-term trading type → overconfidence indicator, anchoring bias indicator, instant gratification preference indicator; panic selling type → loss aversion indicator, herd behavior indicator; long-term stable holding type → mental account indicator, habit bias indicator, etc.).

[0010] As a preferred solution of the method for optimizing the digital intelligence ecological group monitoring model based on the financial knowledge base described in the present invention, the specific implementation process of constructing the financial knowledge tag library includes:

[0011] Based on transaction data i ,TA i ,TF i ,TT i}, with transaction type data TT i Based on the structured knowledge set, the transaction type data TT is extracted i Corresponding structured knowledge labels;

[0012] Based on position change data PC i , transaction amount dataTA i and transaction frequency data TF i , calculate the characteristic value of the user's transaction behavior at the i-th transaction time. The calculation formula is as follows:

[0013]

[0014] Among them, TBC i represents the characteristic value of the user’s transaction behavior at the i-th transaction time, represents the user's average transaction frequency data, ε represents the preset error term, α represents the preset selling impact factor, and β represents the preset position change data PC i Impact factor;

[0015] It should be noted that in this formula, Reflects the degree of deviation of the user's trading frequency at the i-th trading time from the historical mean. A positive value indicates an increase in trading frequency (possibly corresponding to high-frequency trading), while a negative value indicates a decrease (possibly corresponding to stable holdings). Reflects the correlation between position changes and amount. If PC i is negative (reduction in holdings), this part is positive, and the reduction ratio The larger the value, the more prominent the panic selling behavior. α amplifies the impact of selling behavior, reflecting the "loss aversion" theory in behavioral finance. By smoothing the absolute value of position changes, when |PC i When | is large (large increase or decrease in holdings), the denominator approaches 1+β, suppressing excessive fluctuations in the eigenvalue. i When | is small (trading is smooth), the denominator approaches 1, retaining the normal fluctuation characteristics; β adjusts the overall sensitivity of position changes to the characteristic value, reflecting the "anchoring bias" psychology (the intensity of users' reaction to position changes).

[0016] Preset transaction behavior feature threshold interval, if the user's transaction behavior feature value TBC at the i-th transaction time i If the value of the transaction behavior characteristic value TBC is less than the minimum value within the threshold range of the transaction behavior characteristic, the user is judged to have a long-term stable position at the i-th transaction time. i If the user's trading behavior characteristic value TBC at the i-th trading time falls within the trading behavior characteristic threshold range, then the user is judged to be panic selling at the i-th trading time. i If the value is greater than the maximum value within the threshold range of the trading behavior characteristics, the user is judged to be a high-frequency short-term trader at the i-th trading time;

[0017] The user's transaction behavior characteristic value TBC at the i-th transaction time i Based on this, corresponding behavioral finance theory indicators are extracted from the behavioral finance theory indicator set; the extracted structured knowledge labels and behavioral finance theory indicators are used as the financial knowledge label library of the user at the i-th transaction time.

[0018] As a preferred solution of the digital intelligence ecological group monitoring model optimization method based on the financial knowledge base of the present invention, the financial knowledge tag library of all users at the i-th transaction time is obtained, and users with the same structured knowledge tags and behavioral finance theory indicators in the financial knowledge tag library are divided into the same group, and a tag-driven user group set is constructed, which is recorded as TDG = {GP g |g∈[1,G]}, where GP g represents the gth tag-driven user group, and G represents the total number of tag-driven user groups;

[0019] Drive user group GP based on the gth tag g , get tag-driven user group GP g The transaction behavior feature values ​​of all users in the i-th transaction time are used to calculate the GP of the tag-driven user group g The mean and standard deviation of the trading behavior characteristics of all users in the i-th trading time are used to calculate the GP of the tag-driven user group. g The personalized monitoring threshold interval of all users in the ith transaction time is calculated as follows:

[0020] IMT i (GP g )=[μ i (GP g )-δ×σ i (GP g ),μ i (GP g )+δ×σ i (GP g )];

[0021] Among them, IMT i (GP g ) indicates tag-driven user group GP g The personalized monitoring threshold interval of all users in the ith transaction time, μ i (GP g ) represents the mean of trading behavior characteristics, σ i (GP g ) represents the standard deviation of transaction behavior characteristics, and δ represents the preset influencing factor.

[0022] As a preferred solution of the digital intelligence ecological group monitoring model optimization method based on the financial knowledge base of the present invention, the tag-driven user group GP g The personalized monitoring threshold interval IMT of all users at the i-th transaction time i (GP g ), real-time monitoring of tag-driven user groups GP g The transaction behavior characteristic values ​​of all users in the next transaction time. If the transaction behavior characteristic values ​​of the user in the next transaction time do not exist in the personalized monitoring threshold interval IMT i (GP g ), an early warning will be issued to relevant staff, and the tag-driven user group and personalized monitoring threshold range will be dynamically updated.

[0023] A digital ecological group monitoring model optimization system based on a financial knowledge base. This system includes: data acquisition and data pair construction modules, label extraction and feature value calculation modules, group set construction and threshold interval calculation modules, and analysis and early warning modules.

[0024] The data acquisition and data pair construction module: after user authorization, retrieves the user's position change data, transaction amount data, transaction frequency data, transaction type data and transaction time from the financial data platform; based on the transaction time, aligns the position change data, transaction amount data, transaction frequency data and transaction type data in time sequence to construct a transaction data pair;

[0025] The label extraction and feature value calculation module: extracts structured knowledge labels and behavioral finance theory indicators in the financial field based on the financial database; extracts corresponding structured knowledge labels based on transaction type data; calculates the user's transaction behavior feature values, and extracts corresponding behavioral finance theory indicators based on the transaction behavior feature values; and uses the extracted structured knowledge labels and behavioral finance theory indicators as the user's financial knowledge label library;

[0026] The group set construction and threshold interval calculation module: obtains the financial knowledge tag library of all users, constructs a tag-driven user group set; calculates the personalized monitoring threshold interval of all users in the tag-driven user group;

[0027] The analysis and early warning module: based on the personalized monitoring threshold interval, monitors in real time whether the transaction behavior characteristic values ​​of all users in the tag-driven user group at the next transaction time are within the personalized monitoring threshold interval. If not, an early warning is issued and the tag-driven user group and the personalized monitoring threshold interval are dynamically updated.

[0028] Furthermore, the label extraction and feature value calculation module includes a label extraction unit and a feature value calculation unit;

[0029] The label extraction unit: based on a financial database, extracts structured knowledge labels and behavioral finance theory indicators in the financial field from the financial database, constructs a structured knowledge label set and a behavioral finance theory indicator set, and constructs a financial knowledge label library based on transaction data pairs, wherein one type of transaction type data corresponds to at least one behavioral finance theory indicator and at least one structured knowledge label; based on the transaction data pairs and the transaction type data, extracts structured knowledge labels corresponding to the transaction type data from the structured knowledge set;

[0030] The characteristic value calculation unit: calculates the trading behavior characteristic value of the user at the i-th trading time based on the position change data, the transaction amount data and the transaction frequency data; presets the trading behavior characteristic threshold interval, if the trading behavior characteristic value of the user at the i-th trading time is less than the minimum value within the trading behavior characteristic threshold interval, then determines that the user at the i-th trading time is a long-term stable position; if the trading behavior characteristic value of the user at the i-th trading time falls within the trading behavior characteristic threshold interval, then determines that the user at the i-th trading time is a panic sell; if the trading behavior characteristic value of the user at the i-th trading time is greater than the maximum value within the trading behavior characteristic threshold interval, then determines that the user at the i-th trading time is a high-frequency short-term trade; based on the trading behavior characteristic value of the user at the i-th trading time, extracts the corresponding behavioral finance theory indicator from the behavioral finance theory indicator set; and uses the extracted structured knowledge label and behavioral finance theory indicator as the financial knowledge label library of the user at the i-th trading time.

[0031] Furthermore, the group set construction and threshold interval calculation module includes a group set construction unit and a threshold interval calculation unit;

[0032] The group set construction unit: obtains the financial knowledge tag library of all users at the i-th transaction time, divides users with the same structured knowledge tags and behavioral finance theory indicators in the financial knowledge tag library into the same group, and constructs a tag-driven user group set;

[0033] The threshold interval calculation unit: based on the tag-driven user group, obtains the transaction behavior feature values ​​of all users in the tag-driven user group at the i-th transaction time, calculates the transaction behavior feature mean and transaction behavior feature standard deviation of all users in the tag-driven user group at the i-th transaction time, and calculates the personalized monitoring threshold interval of all users in the tag-driven user group at the i-th transaction time based on the transaction behavior feature mean and transaction behavior feature standard deviation.

[0034] Furthermore, the analysis and warning module includes an analysis and warning unit;

[0035] The analysis and early warning unit: based on the personalized monitoring threshold interval of all users in the tag-driven user group at the i-th transaction time, monitors the transaction behavior characteristic values ​​of all users in the tag-driven user group at the next transaction time in real time. If the transaction behavior characteristic values ​​of the users at the next transaction time do not fall within the personalized monitoring threshold interval, an early warning is issued to relevant staff, and the tag-driven user group and the personalized monitoring threshold interval are dynamically updated.

[0036] Compared with the existing technology, the beneficial effects achieved by the present invention are: in the digital ecological group monitoring model optimization method and system based on the financial knowledge base provided by the present invention, by retrieving key behavioral data from the financial data platform and performing standardization, cleaning and time alignment, high-quality transaction data pairs that can be used for subsequent analysis are constructed, providing an accurate and consistent data foundation for the entire model; combining structured knowledge labels with behavioral finance theoretical indicators, the user's transaction behavior is multi-dimensionally characterized, and a financial knowledge label library is constructed, thereby realizing theoretical modeling of user behavior characteristics and enhancing the model's ability to identify potential psychological motivations; through matching financial knowledge labels, a label-driven user group is constructed, and based on the mean and standard deviation of the group's internal characteristics, the personalized monitoring threshold interval is calculated, effectively realizing the accurate characterization of the differentiated characteristics of different user groups, and improving the sensitivity and robustness of dynamic monitoring at the group level; real-time monitoring of whether the user's behavior changes at the next transaction time cross the limit, and triggering an early warning once it deviates from its personalized threshold range, and updating the group structure and threshold interval, so that the model has adaptive optimization and dynamic iteration capabilities. The present invention not only improves the real-time and accuracy of financial behavior monitoring, but also enables dynamic risk identification and intelligent regulation from individuals to groups. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0038] Figure 1 This is a schematic diagram of the steps of the method for optimizing the digital intelligence ecological group monitoring model based on the financial knowledge base of the present invention;

[0039] Figure 2 It is a structural diagram of the digital intelligence ecological group monitoring model optimization system based on the financial knowledge base of the present invention. DETAILED DESCRIPTION

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0041] See also Figure 1 In the first embodiment of the present invention, a method for optimizing a digital intelligence ecological group monitoring model based on a financial knowledge base is provided. The method includes the following steps:

[0042] Step S1: After user authorization, retrieve the user's position change data, transaction amount data, transaction frequency data, transaction type data and transaction time from the financial data platform; based on the transaction time, align the position change data, transaction amount data, transaction frequency data and transaction type data in time sequence to construct a transaction data pair.

[0043] Specifically, after authorization by the user, the user's financial behavior log is retrieved from the financial data platform. The financial behavior log includes the user's position change data, transaction amount data, transaction frequency data, transaction type data, and transaction time on the financial data platform; the position change data, transaction amount data, transaction frequency data, and transaction type data are normalized and noise-cleaned;

[0044] Furthermore, based on the transaction time, the position change data, transaction amount data, transaction frequency data and transaction type data are time-series aligned, and transaction data pairs are constructed. The transaction data pair corresponding to the i-th transaction time is recorded as {PC i ,TA i ,TF i ,TT i}, where PC i Indicates the position change data corresponding to the i-th trading time, TA i Indicates the transaction amount data corresponding to the i-th transaction time, TF i Indicates the transaction frequency data corresponding to the i-th transaction time, TT i Indicates the transaction type data corresponding to the i-th transaction time.

[0045] Step S2: Based on the financial database, extract structured knowledge labels and behavioral finance theory indicators in the financial field; based on the transaction type data, extract the corresponding structured knowledge labels; calculate the user's transaction behavior feature values, and based on the transaction behavior feature values, extract the corresponding behavioral finance theory indicators; use the extracted structured knowledge labels and behavioral finance theory indicators as the user's financial knowledge label library.

[0046] Specifically, based on the financial database, the structured knowledge labels and behavioral finance theory indicators in the financial field are extracted from the financial database, a set of structured knowledge labels and a set of behavioral finance theory indicators are constructed, and {PC i ,TA i ,TF i ,TT i}, build a financial knowledge label library, in which one type of transaction type data corresponds to at least one behavioral finance theory indicator and at least one structured knowledge label, (structured knowledge labels include typical transaction feature labels, user behavior pattern labels, risk level classification labels and psychological behavior labels, etc. Behavioral finance theory indicators include: high-frequency short-term trading type → overconfidence indicator, anchoring bias indicator, instant gratification preference indicator; panic selling type → loss aversion indicator, herd behavior indicator; long-term stable holding type → mental account indicator, habit bias indicator, etc.).

[0047] Furthermore, the specific implementation process of constructing the financial knowledge tag library includes:

[0048] Based on transaction data i ,TA i ,TF i ,TT i}, with transaction type data TT i Based on the structured knowledge set, the transaction type data TT is extracted i Corresponding structured knowledge labels;

[0049] Based on position change data PC i , transaction amount dataTA i and transaction frequency data TF i , calculate the characteristic value of the user's transaction behavior at the i-th transaction time. The calculation formula is as follows:

[0050]

[0051] Among them, TBC i represents the characteristic value of the user’s transaction behavior at the i-th transaction time, represents the user's average transaction frequency data, ε represents the preset error term, α represents the preset selling impact factor, and β represents the preset position change data PC i Impact factor;

[0052] In the present invention, by (frequency deviation) and (reduction ratio), combining "trading activity" with "asset adjustment direction", for example: high-frequency trading users TF i much higher than This may correspond to the "overconfidence" mentality, and the formula pushes up TBC through high frequency deviation i , accurately matching the "overtrading" theory in behavioral finance. i The negative value is large and TA iHigh) will significantly increase the numerator value, which is directly related to the panic selling caused by "loss aversion", avoiding the one-sidedness of a single indicator (such as only looking at the transaction amount).

[0053] Preset transaction behavior feature threshold interval, if the user's transaction behavior feature value TBC at the i-th transaction time i If the value of the transaction behavior characteristic value TBC is less than the minimum value within the threshold range of the transaction behavior characteristic, the user is judged to have a long-term stable position at the i-th transaction time. i If the user's trading behavior characteristic value TBC at the i-th trading time falls within the trading behavior characteristic threshold range, then the user is judged to be panic selling at the i-th trading time. i If the value is greater than the maximum value within the threshold range of the trading behavior characteristics, the user is judged to be a high-frequency short-term trader at the i-th trading time;

[0054] Furthermore, the user's transaction behavior characteristic value TBC at the i-th transaction time i Based on this, corresponding behavioral finance theory indicators are extracted from the behavioral finance theory indicator set; the extracted structured knowledge labels and behavioral finance theory indicators are used as the financial knowledge label library of the user at the i-th transaction time.

[0055] Step S3: Obtain the financial knowledge tag library of all users, build a tag-driven user group set; and calculate the personalized monitoring threshold interval of all users in the tag-driven user group.

[0056] Specifically, the financial knowledge tag library of all users at the i-th transaction time is obtained, and users with the same structured knowledge tags and behavioral finance theory indicators in the financial knowledge tag library are divided into the same group, and a tag-driven user group set is constructed, which is recorded as TDG = {GP g |g∈[1,G]}, where GP g represents the gth tag-driven user group, and G represents the total number of tag-driven user groups;

[0057] Furthermore, based on the g-th tag, the user group GP is driven g , get tag-driven user group GP g The transaction behavior feature values ​​of all users in the i-th transaction time are used to calculate the GP of the tag-driven user group g The mean and standard deviation of the trading behavior characteristics of all users in the i-th trading time are used to calculate the GP of the tag-driven user group. g The personalized monitoring threshold interval of all users in the ith transaction time is calculated as follows:

[0058] IMTi GP g )=[μ i (GP g )-δ×σ i (GP g ),μ i (GP g )+δ×σ i (GP g )];

[0059] Among them, IMT i (GP g ) indicates tag-driven user group GP g The personalized monitoring threshold interval of all users in the ith transaction time, μ i (GP g ) represents the mean of trading behavior characteristics, σ i (GP g ) represents the standard deviation of transaction behavior characteristics, and δ represents the preset influencing factor.

[0060] In the present invention, by μ i (GP g ) and σ i (GP g ) Construct a group-specific interval, for example: Institutional investor group: transaction frequency mean μ i (GP g ) is high and the standard deviation σ i (GP g ) is small (strong operational discipline), the threshold range is narrow, and its abnormal position reduction behavior can be accurately identified. Retail investor group: mean μ i (GP g ) is low but the standard deviation σ i (GP g ) is large (behavior dispersion), and the threshold range is wide, which avoids false warnings caused by emotional trading of individual retail investors. It can solve the problem of "ignoring the heterogeneity of user groups and causing inaccurate warnings" in the background technology, for example, avoiding misjudging the normal position adjustment of institutions as "panic selling"; and the threshold range is calculated in real time as the group members change (such as the change of user tags leading to group reorganization). For example, when a user moves from the "long-term stable holding" group to the "high-frequency trading" group, his historical data is included in the μ of the new group. i (GP g ) and σ i (GP g ) calculation, and the threshold interval is adjusted accordingly to ensure that the monitoring standard matches the current behavior.

[0061] Step S4: Based on the personalized monitoring threshold interval, real-time monitoring is performed to determine whether the transaction behavior characteristic values ​​of all users in the tag-driven user group at the next transaction time are within the personalized monitoring threshold interval. If not, an early warning is issued and the tag-driven user group and the personalized monitoring threshold interval are dynamically updated.

[0062] Specifically, based on the tag-driven user group GP g The personalized monitoring threshold interval IMT of all users at the i-th transaction time i (GP g ), real-time monitoring of tag-driven user groups GP g The transaction behavior characteristic values ​​of all users in the next transaction time. If the transaction behavior characteristic values ​​of the user in the next transaction time do not exist in the personalized monitoring threshold interval IMT i (GP g ), an early warning will be issued to relevant staff, and the tag-driven user group and personalized monitoring threshold range will be dynamically updated.

[0063] See also Figure 2 In the second embodiment, a digital ecological group monitoring model optimization system based on a financial knowledge base is provided, which includes: a data acquisition and data pair construction module, a label extraction and feature value calculation module, a group set construction and threshold interval calculation module, and an analysis and early warning module;

[0064] The data acquisition and data pair construction module: after user authorization, retrieves the user's position change data, transaction amount data, transaction frequency data, transaction type data and transaction time from the financial data platform; based on the transaction time, aligns the position change data, transaction amount data, transaction frequency data and transaction type data in time sequence to construct a transaction data pair;

[0065] The label extraction and feature value calculation module: extracts structured knowledge labels and behavioral finance theory indicators in the financial field based on the financial database; extracts corresponding structured knowledge labels based on transaction type data; calculates the user's transaction behavior feature values, and extracts corresponding behavioral finance theory indicators based on the transaction behavior feature values; and uses the extracted structured knowledge labels and behavioral finance theory indicators as the user's financial knowledge label library;

[0066] The group set construction and threshold interval calculation module: obtains the financial knowledge tag library of all users, constructs a tag-driven user group set; calculates the personalized monitoring threshold interval of all users in the tag-driven user group;

[0067] The analysis and early warning module: based on the personalized monitoring threshold interval, monitors in real time whether the transaction behavior characteristic values ​​of all users in the tag-driven user group at the next transaction time are within the personalized monitoring threshold interval. If not, an early warning is issued and the tag-driven user group and the personalized monitoring threshold interval are dynamically updated.

[0068] Furthermore, the label extraction and feature value calculation module includes a label extraction unit and a feature value calculation unit;

[0069] The label extraction unit: based on a financial database, extracts structured knowledge labels and behavioral finance theory indicators in the financial field from the financial database, constructs a structured knowledge label set and a behavioral finance theory indicator set, and constructs a financial knowledge label library based on transaction data pairs, wherein one type of transaction type data corresponds to at least one behavioral finance theory indicator and at least one structured knowledge label; based on the transaction data pairs and the transaction type data, extracts structured knowledge labels corresponding to the transaction type data from the structured knowledge set;

[0070] The characteristic value calculation unit: calculates the trading behavior characteristic value of the user at the i-th trading time based on the position change data, the transaction amount data and the transaction frequency data; presets the trading behavior characteristic threshold interval, if the trading behavior characteristic value of the user at the i-th trading time is less than the minimum value within the trading behavior characteristic threshold interval, then determines that the user at the i-th trading time is a long-term stable position; if the trading behavior characteristic value of the user at the i-th trading time falls within the trading behavior characteristic threshold interval, then determines that the user at the i-th trading time is a panic sell; if the trading behavior characteristic value of the user at the i-th trading time is greater than the maximum value within the trading behavior characteristic threshold interval, then determines that the user at the i-th trading time is a high-frequency short-term trade; based on the trading behavior characteristic value of the user at the i-th trading time, extracts the corresponding behavioral finance theory indicator from the behavioral finance theory indicator set; and uses the extracted structured knowledge label and behavioral finance theory indicator as the financial knowledge label library of the user at the i-th trading time.

[0071] Furthermore, the group set construction and threshold interval calculation module includes a group set construction unit and a threshold interval calculation unit;

[0072] The group set construction unit: obtains the financial knowledge tag library of all users at the i-th transaction time, divides users with the same structured knowledge tags and behavioral finance theory indicators in the financial knowledge tag library into the same group, and constructs a tag-driven user group set;

[0073] The threshold interval calculation unit: based on the tag-driven user group, obtains the transaction behavior feature values ​​of all users in the tag-driven user group at the i-th transaction time, calculates the transaction behavior feature mean and transaction behavior feature standard deviation of all users in the tag-driven user group at the i-th transaction time, and calculates the personalized monitoring threshold interval of all users in the tag-driven user group at the i-th transaction time based on the transaction behavior feature mean and transaction behavior feature standard deviation.

[0074] Furthermore, the analysis and warning module includes an analysis and warning unit;

[0075] The analysis and early warning unit: based on the personalized monitoring threshold interval of all users in the tag-driven user group at the i-th transaction time, monitors the transaction behavior characteristic values ​​of all users in the tag-driven user group at the next transaction time in real time. If the transaction behavior characteristic values ​​of the users at the next transaction time do not fall within the personalized monitoring threshold interval, an early warning is issued to relevant staff, and the tag-driven user group and the personalized monitoring threshold interval are dynamically updated.

[0076] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0077] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for optimizing a digital intelligence ecological group monitoring model based on a financial knowledge base, characterized by: The method comprises the following steps: Step S1: After user authorization, retrieve the user's position change data, transaction amount data, transaction frequency data, transaction type data, and transaction time from the financial data platform; based on the transaction time, perform time series alignment on the position change data, transaction amount data, transaction frequency data, and transaction type data to construct a transaction data pair; Step S2: Based on the financial database, extract structured knowledge labels and behavioral finance theory indicators in the financial field; based on the transaction type data, extract the corresponding structured knowledge labels; calculate the user's transaction behavior feature values, and based on the transaction behavior feature values, extract the corresponding behavioral finance theory indicators; use the extracted structured knowledge labels and behavioral finance theory indicators as the user's financial knowledge label library; Step S3: Obtain the financial knowledge tag library of all users and construct a tag-driven user group set; calculate the personalized monitoring threshold interval for all users in the tag-driven user group; Step S4: Based on the personalized monitoring threshold interval, real-time monitoring is performed to determine whether the transaction behavior characteristic values ​​of all users in the tag-driven user group at the next transaction time are within the personalized monitoring threshold interval. If not, an early warning is issued and the tag-driven user group and the personalized monitoring threshold interval are dynamically updated.

2. The method for optimizing a digital intelligence ecological group monitoring model based on a financial knowledge base according to claim 1 is characterized in that: The specific implementation process of step S1 includes: After the user's authorization, the user's financial behavior log is retrieved from the financial data platform. The financial behavior log includes the user's position change data, transaction amount data, transaction frequency data, transaction type data, and transaction time on the financial data platform; the position change data, transaction amount data, transaction frequency data, and transaction type data are normalized and noise-cleaned; Based on the transaction time, the position change data, transaction amount data, transaction frequency data and transaction type data are aligned in time series, and transaction data pairs are constructed. The transaction data pair corresponding to the i-th transaction time is recorded as {PC i ,TA i ,TF i ,TT i }, where PC i Indicates the position change data corresponding to the i-th trading time, TA i Indicates the transaction amount data corresponding to the i-th transaction time, TF i Indicates the transaction frequency data corresponding to the i-th transaction time, TT i Indicates the transaction type data corresponding to the i-th transaction time.

3. The method for optimizing a digital intelligence ecological group monitoring model based on a financial knowledge base according to claim 2 is characterized in that: The specific implementation process of step S2 includes: Based on the financial database, the structured knowledge labels and behavioral finance theory indicators in the financial field are extracted from the financial database, and a set of structured knowledge labels and a set of behavioral finance theory indicators are constructed. i ,TA i ,TF i ,TT i }, build a financial knowledge tag library, where one type of transaction data corresponds to at least one behavioral finance theory indicator and at least one structured knowledge tag.

4. The method for optimizing a digital intelligence ecological group monitoring model based on a financial knowledge base according to claim 3 is characterized in that: The specific implementation process of building the financial knowledge tag library includes: Based on transaction data i ,TA i ,TF i ,TT i }, with transaction type data TT i Based on the structured knowledge set, the transaction type data TT is extracted i Corresponding structured knowledge labels; Based on position change data PC i , transaction amount dataTA i and transaction frequency data TF i , calculate the transaction behavior characteristic value of the user at the i-th transaction time, the calculation formula is as follows: Among them, TBC i represents the characteristic value of the user’s transaction behavior at the i-th transaction time, represents the user's average transaction frequency data, ε represents the preset error term, α represents the preset selling impact factor, and β represents the preset position change data PC i Impact factor; Preset transaction behavior feature threshold interval, if the user's transaction behavior feature value TBC at the i-th transaction time i If the value of the transaction behavior characteristic value TBC is less than the minimum value within the threshold range of the transaction behavior characteristic, the user is judged to have a long-term stable position at the i-th transaction time. i If the user's trading behavior characteristic value TBC at the i-th trading time is within the trading behavior characteristic threshold range, then the user is judged to be panic selling at the i-th trading time. i If the value is greater than the maximum value within the threshold range of the trading behavior characteristics, the user is judged to be a high-frequency short-term trader at the i-th trading time; The user's transaction behavior characteristic value TBC at the i-th transaction time i Based on this, corresponding behavioral finance theory indicators are extracted from the behavioral finance theory indicator set; the extracted structured knowledge labels and behavioral finance theory indicators are used as the financial knowledge label library of the user at the i-th transaction time.

5. The method for optimizing a digital intelligence ecological group monitoring model based on a financial knowledge base according to claim 4 is characterized in that: The specific implementation process of step S3 includes: Obtain the financial knowledge tag library of all users at the i-th transaction time, divide the users with the same structured knowledge tags and behavioral finance theory indicators in the financial knowledge tag library into the same group, and construct a tag-driven user group set, denoted as TDG = {GP g |g∈[1,G]}, where GP g represents the gth tag-driven user group, and G represents the total number of tag-driven user groups; Drive user group GP based on the gth tag g , get tag-driven user group GP g The transaction behavior feature values ​​of all users in the i-th transaction time are used to calculate the GP of the tag-driven user group g The mean and standard deviation of the trading behavior characteristics of all users in the i-th trading time are used to calculate the GP of the tag-driven user group. g The personalized monitoring threshold interval of all users at the i-th transaction time is calculated as follows: IMT i (GP g )=[μ i (GP g )-δ×σ i (GP g ),μ i (GP g )+δ×σ i (GP g )]; Among them, IMT i (GP g ) indicates tag-driven user group GP g The personalized monitoring threshold interval of all users in the ith transaction time, μ i (GP g ) represents the mean of trading behavior characteristics, σ i (GP g ) represents the standard deviation of transaction behavior characteristics, and δ represents the preset influencing factor.

6. The method for optimizing a digital intelligence ecological group monitoring model based on a financial knowledge base according to claim 5 is characterized in that: The specific implementation process of step S4 includes: Tag-driven user group GP g The personalized monitoring threshold interval IMT of all users at the i-th transaction time i (GP g ), real-time monitoring of tag-driven user groups GP g The transaction behavior characteristic values ​​of all users in the next transaction time. If the transaction behavior characteristic values ​​of the user in the next transaction time do not exist in the personalized monitoring threshold interval IMT i (GP g ), an early warning will be issued to relevant staff, and the tag-driven user group and personalized monitoring threshold range will be dynamically updated.

7. A digital intelligence ecological group monitoring model optimization system based on a financial knowledge base, which executes the digital intelligence ecological group monitoring model optimization method based on a financial knowledge base as described in any one of claims 1-6, characterized in that: The system includes: a data acquisition and data pair construction module, a label extraction and feature value calculation module, a group set construction and threshold interval calculation module and an analysis and early warning module; The data acquisition and data pair construction module: after user authorization, retrieves the user's position change data, transaction amount data, transaction frequency data, transaction type data and transaction time from the financial data platform; based on the transaction time, aligns the position change data, transaction amount data, transaction frequency data and transaction type data in time sequence to construct a transaction data pair; The label extraction and feature value calculation module: extracts structured knowledge labels and behavioral finance theory indicators in the financial field based on the financial database; extracts corresponding structured knowledge labels based on transaction type data; calculates the user's transaction behavior feature values, and extracts corresponding behavioral finance theory indicators based on the transaction behavior feature values; and uses the extracted structured knowledge labels and behavioral finance theory indicators as the user's financial knowledge label library; The group set construction and threshold interval calculation module: obtains the financial knowledge tag library of all users, constructs a tag-driven user group set; calculates the personalized monitoring threshold interval of all users in the tag-driven user group; The analysis and early warning module: based on the personalized monitoring threshold interval, monitors in real time whether the transaction behavior characteristic values ​​of all users in the tag-driven user group at the next transaction time are within the personalized monitoring threshold interval. If not, an early warning is issued and the tag-driven user group and the personalized monitoring threshold interval are dynamically updated.

8. The digital intelligence ecological group monitoring model optimization system based on the financial knowledge base according to claim 7 is characterized by: The label extraction and feature value calculation module includes a label extraction unit and a feature value calculation unit; The label extraction unit: based on a financial database, extracts structured knowledge labels and behavioral finance theory indicators in the financial field from the financial database, constructs a structured knowledge label set and a behavioral finance theory indicator set, and constructs a financial knowledge label library based on transaction data pairs, wherein one type of transaction type data corresponds to at least one behavioral finance theory indicator and at least one structured knowledge label; based on the transaction data pairs and the transaction type data, extracts structured knowledge labels corresponding to the transaction type data from the structured knowledge set; The characteristic value calculation unit: calculates the trading behavior characteristic value of the user at the i-th trading time based on the position change data, the transaction amount data and the transaction frequency data; presets the trading behavior characteristic threshold interval, if the trading behavior characteristic value of the user at the i-th trading time is less than the minimum value within the trading behavior characteristic threshold interval, then determines that the user at the i-th trading time is a long-term stable position; if the trading behavior characteristic value of the user at the i-th trading time falls within the trading behavior characteristic threshold interval, then determines that the user at the i-th trading time is a panic sell; if the trading behavior characteristic value of the user at the i-th trading time is greater than the maximum value within the trading behavior characteristic threshold interval, then determines that the user at the i-th trading time is a high-frequency short-term trade; based on the trading behavior characteristic value of the user at the i-th trading time, extracts the corresponding behavioral finance theory indicator from the behavioral finance theory indicator set; and uses the extracted structured knowledge label and behavioral finance theory indicator as the financial knowledge label library of the user at the i-th trading time.

9. The digital intelligence ecological group monitoring model optimization system based on the financial knowledge base according to claim 8 is characterized by: The group set construction and threshold interval calculation module includes a group set construction unit and a threshold interval calculation unit; The group set construction unit: obtains the financial knowledge tag library of all users at the i-th transaction time, divides users with the same structured knowledge tags and behavioral finance theory indicators in the financial knowledge tag library into the same group, and constructs a tag-driven user group set; The threshold interval calculation unit: based on the tag-driven user group, obtains the transaction behavior feature values ​​of all users in the tag-driven user group at the i-th transaction time, calculates the transaction behavior feature mean and transaction behavior feature standard deviation of all users in the tag-driven user group at the i-th transaction time, and calculates the personalized monitoring threshold interval of all users in the tag-driven user group at the i-th transaction time based on the transaction behavior feature mean and transaction behavior feature standard deviation.

10. The digital intelligence ecological group monitoring model optimization system based on the financial knowledge base according to claim 9 is characterized by: The analysis and early warning module includes an analysis and early warning unit; The analysis and early warning unit: based on the personalized monitoring threshold interval of all users in the tag-driven user group at the i-th transaction time, monitors the transaction behavior characteristic values ​​of all users in the tag-driven user group at the next transaction time in real time. If the transaction behavior characteristic values ​​of the users at the next transaction time do not fall within the personalized monitoring threshold interval, an early warning is issued to relevant staff, and the tag-driven user group and the personalized monitoring threshold interval are dynamically updated.

Images