Intelligent and accurate pushing method and system of securities information based on big data
By separating user interest characteristics from information content characteristics in a securities big data platform, and combining real-time user scenarios and historical transaction data, the push weight of securities information is calculated, solving the problem of inaccurate information push in existing technologies, and realizing personalized and efficient information push.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENZHEN JINHUI RONGZHI DATA SERVICE CO LTD
- Filing Date
- 2026-03-09
- Publication Date
- 2026-06-09
AI Technical Summary
Existing securities information push methods lack the ability to dynamically analyze and adapt multi-dimensional user behavior data and deep characteristics of information content, resulting in a low degree of matching between pushed content and user needs, making it difficult to achieve intelligent and accurate push.
By collecting multi-dimensional data sources from securities big data platforms, separating user interest feature vectors from information content feature vectors, calculating correlation matching degree, and combining data such as user real-time scenario attributes and historical transaction frequency, the priority of content adaptation and urgency of demand are calculated, the effective push weight of securities information is calculated, and personalized information arrangement and push are carried out.
It enables intelligent and precise push of securities information based on big data, improving the accuracy of push and personalized recommendation capabilities, and solving the problem of indiscriminate push in traditional push methods.
Smart Images
Figure CN122175695A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method and system for intelligent and precise delivery of securities information based on big data, belonging to the field of securities information processing technology. Background Technology
[0002] Securities information push is an important means for securities companies to serve their clients. Its accuracy and timeliness directly affect clients' investment decision-making experience and service quality. At present, securities information push mostly adopts a static push model based on content, relying on keyword filtering and hot news ranking, and lacks the ability to dynamically analyze and adapt to multi-dimensional user behavior data and deep characteristics of information content.
[0003] However, most of the pushed content is based on global popularity or simple tag matching, without fully considering user interest characteristics, real-time scenarios and differences in historical behavior. This results in a low degree of matching between the pushed content and the user's actual needs, making it difficult to achieve truly intelligent and accurate push. Therefore, a method is needed to improve the accuracy of intelligent and accurate push of securities information. Summary of the Invention
[0004] This invention provides a method and system for intelligent and precise push of securities information based on big data, the main purpose of which is to improve the accuracy of intelligent and precise push of securities information based on big data.
[0005] To achieve the above objectives, the present invention provides a method for intelligent and precise push of securities information based on big data, comprising:
[0006] Collect multi-dimensional data sources from a securities big data platform, separate user interest feature vectors and information content feature vectors from the multi-dimensional data sources, and calculate the correlation matching degree between the interest feature vectors and the information content feature vectors; Determine the user's real-time scenario attributes and, in conjunction with the correlation matching degree, calculate the content adaptation priority of securities information in the securities big data platform during the push process; The system acquires the user's historical transaction frequency, position size, and information interaction rate data to calculate the urgency of the user's need for securities information and determine the target information push accuracy for the user. By combining the content adaptation priority, the urgency of the demand, and the relevance matching degree, the effective push weight of the securities information is calculated. Combined with the target information push accuracy, the securities information is then processed for content arrangement and push to obtain the push result.
[0007] Optionally, separating the user's interest feature vector and information content feature vector from the multi-dimensional data source includes: Extract the structured data set from the multi-dimensional data source to extract the user's behavioral features and information text features, and obtain an initial feature set; Based on the initial feature set, the user's interest feature space boundary and information feature space boundary are constructed respectively, so as to separate the user's interest feature vector and information content feature vector from the structured data set.
[0008] Optionally, the step of constructing the user's interest feature space boundary and information feature space boundary based on the initial feature set includes: Core interest features and core information features are extracted from the initial feature set and mapped to obtain the interest feature matrix and information feature matrix, respectively. The interest feature matrix and the information feature matrix are respectively subjected to clustering and partitioning to obtain user interest feature clusters and information content feature clusters; Calculate the minimum distance threshold between the user interest feature cluster and the information content feature cluster respectively, so as to construct the user interest feature space boundary and the information feature space boundary.
[0009] Optionally, calculating the correlation matching degree between the interest feature vector and the information content feature vector includes: Calculate the interaction information entropy and correlation entropy between the interest feature vector and the information content feature vector to determine the static correlation strength and dynamic coupling coefficient between the interest feature vector and the information content feature vector; Calculate the cross-entropy gain factor between the interest feature vector and the information content feature vector; The correlation matching degree between the interest feature vector and the information content feature vector is calculated by combining the static correlation strength and the dynamic coupling coefficient.
[0010] Optionally, the step of calculating the content adaptation priority of securities information in the securities big data platform during the push process, based on the correlation matching degree, includes: Obtain the screen display specifications of the user's device to calculate the effective display area of the screen; Calculate the presentation complexity of the securities information on the currently used device to calculate the real-time understanding load corresponding to the securities information; By combining the real-time scene attributes, the correlation matching degree, the effective screen display area, and the instant understanding load, the content adaptation priority of the securities information during the push process is calculated.
[0011] Optionally, calculating the real-time understanding load corresponding to the information based on the presentation complexity includes: Identify the professional financial entities within the main text of the securities information; The parsing difficulty of the information text is evaluated, and the professional entity density of the professional financial entity is calculated to obtain the real-time comprehension load of the information.
[0012] Optionally, calculating the urgency of the user's need for securities information includes: Based on the historical transaction frequency, calculate the recent transaction activity of the user and the market value volatility corresponding to the holding size; Based on the information interaction rate data, the user's response sensitivity to information is calculated; By combining the recent trading activity, the market capitalization volatility, and the response sensitivity, the urgency of the user's need for securities information is calculated.
[0013] Optionally, calculating the user's recent transaction activity based on the historical transaction frequency includes: Based on the historical transaction frequency, the user's average daily transaction frequency is calculated; Query the average daily transaction frequency of the user's reference user group to obtain the group's benchmark level; Based on the average daily transaction frequency and the group's benchmark level, the user's recent transaction activity level is obtained.
[0014] Optionally, the step of calculating the effective push weight of the securities information by combining the content adaptation priority, the urgency of the demand, and the relevance matching degree includes: Obtain raw data on the dissemination behavior of the securities information to calculate the information dissemination popularity index; Based on the aforementioned dissemination popularity index, the intrinsic value coefficient corresponding to the securities information is determined; The intrinsic value coefficient is weighted by combining the content adaptation priority, the urgency of the demand, and the relevance matching degree to obtain the effective push weight of the securities information.
[0015] To address the above problems, this invention also provides a big data-based intelligent and precise securities information push system, the system comprising: The data matching and calculation module is used to collect multi-dimensional data sources from the securities big data platform, separate user interest feature vectors and information content feature vectors from the multi-dimensional data sources, and calculate the correlation matching degree between the interest feature vectors and the information content feature vectors. The scene adaptation calculation module is used to determine the user's real-time scene attributes and, in combination with the correlation matching degree, calculate the content adaptation priority of securities information in the securities big data platform during the push process. The push accuracy determination module is used to obtain the user's historical transaction frequency, holding size and information interaction rate data to calculate the urgency of the user's demand for securities information and determine the user's target information push accuracy. The push weight calculation module is used to calculate the effective push weight of the securities information by combining the content adaptation priority, the urgency of the demand, and the relevance matching degree. In combination with the target information push accuracy, the securities information is processed for content arrangement and push to obtain the push result.
[0016] Compared to the problems described in the background technology, this invention, by separating user interest feature vectors and information content feature vectors from multi-dimensional data sources of securities, can accurately capture the core correlation between user preferences and information value, providing a quantitative decision-making basis for personalized information recommendation in the securities field and effectively solving the problem of indiscriminate recommendation in traditional mass recommendation. Then, by determining the real-time scenario attributes of the user when receiving information and combining this with the correlation matching degree, this invention can quantify the actual value sequence of each piece of information to the user in a specific scenario, providing a direct basis for subsequent accurate push ranking. Next, by obtaining the user's historical trading frequency, holding size, and information interaction rate data, this invention can quantify the urgency of the user's demand for securities information at the current time. Then, by combining the content adaptation priority, the urgency of the demand, and the correlation matching degree to calculate the effective push weight, this invention can quantitatively evaluate the comprehensive push value of a single piece of information in a specific user and scenario, providing a precise numerical basis for information distribution ranking. Therefore, this invention can improve the accuracy of intelligent and accurate push of securities information based on big data. Attached Figure Description
[0017] Figure 1 A flowchart illustrating a method for intelligent and precise push of securities information based on big data, provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the data processing flow in a big data-based intelligent and precise securities information push method provided in an embodiment of the present invention; Figure 3 A schematic diagram of modules for implementing the intelligent and accurate securities information push method based on big data, provided in an embodiment of the present invention; Figure 4 A schematic diagram of a computer device for a big data-based intelligent and precise securities information push method provided in an embodiment of the present invention; The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0019] This application provides a method for intelligent and precise push of securities information based on big data. The executing entity of this method includes, but is not limited to, at least one electronic device that can be configured to execute the method provided in this application, such as a server or a terminal. In other words, the method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.
[0020] Reference Figure 1 The diagram shown is a flowchart illustrating a method for intelligent and precise push of securities information based on big data, according to an embodiment of the present invention. In this embodiment, the method for intelligent and precise push of securities information based on big data includes: S1. Collect multi-dimensional data sources from the securities big data platform, separate user interest feature vectors and information content feature vectors from the multi-dimensional data sources, and calculate the correlation matching degree between the interest feature vectors and the information content feature vectors.
[0021] This invention separates user interest feature vectors and information content feature vectors from multi-dimensional data sources in the securities industry, which can accurately capture the core correlation between user preferences and information value, providing a quantitative decision-making basis for personalized information recommendation in the securities field and effectively solving the problem of undifferentiated traditional mass recommendation.
[0022] The multi-dimensional data source refers to the comprehensive data set integrated in the securities big data platform that reflects user behavior and information attributes, such as user behavior data and securities information text data; the user interest feature vector refers to a multi-dimensional vector that quantifies user preferences obtained through behavioral analysis and feature extraction, such as the quantitative expression of "high-risk financial management preference + attention to the new energy sector + daily morning browsing"; the information content feature vector refers to a quantitative vector that represents the content theme and keywords extracted from the information through text mining, such as the feature combination of "new energy policy + industry research report + medium-risk rating"; furthermore, the multi-dimensional data sources in the securities big data platform can be collected and preliminarily processed through a data preprocessing module to ensure data validity. The data preprocessing module is compiled using the JAVA language.
[0023] Figure 2 This is a schematic diagram of the data processing flow in a big data-based intelligent and precise securities information push method provided in an embodiment of the present invention.
[0024] As an embodiment of the present invention, separating the user's interest feature vector and information content feature vector from the multi-dimensional data source includes: S201: Extract the structured data set from the multi-dimensional data source to extract the user's behavioral features and information text features, and obtain an initial feature set; S202: Based on the initial feature set, construct the user's interest feature space boundary and information feature space boundary respectively, so as to separate the user's interest feature vector and information content feature vector from the structured data set.
[0025] The structured data set refers to a regularized data set obtained through data cleaning and processing, suitable for modeling, such as a record table containing user ID, timestamp, operation type, object identifier, and numerical fields. The behavioral features are statistical or sequential features extracted from the structured data set that quantify operational habits and attention patterns, such as click-through rate, holding duration, and search keyword frequency. The information text features are linguistic or statistical features extracted from the structured data set that characterize its semantic content, such as bag-of-words model vectors, named entity lists, and sentiment polarity scores. The initial feature set refers to a hybrid feature matrix formed by encoding and concatenating the behavioral features and the information text features. The interest feature space boundary refers to a separating hyperplane or decision region defined in a high-dimensional feature space through supervised learning or clustering algorithms, capable of distinguishing different interest preference patterns. The information feature space boundary refers to a separating hyperplane or decision region defined in a high-dimensional feature space through topic modeling or classifiers, capable of distinguishing different information categories and topics.
[0026] Optionally, the structured data set from the multi-dimensional data source can be extracted through distributed SQL queries and NoSQL data connectors; user behavior features and information text features can be extracted from the structured data set through a feature engineering pipeline, which includes statistical aggregation, sequence modeling, and text vectorization; the initial features belonging to the user dimension and information dimension can be trained or fitted through gradient boosting decision trees, deep neural network classifiers, or Gaussian mixture models, respectively, thereby constructing the boundary of the interest feature space and the boundary of the information feature space, such as XGBoost; feature vector extraction based on boundary discrimination or multi-dimensional space projection techniques can be used to map and separate each record in the structured data set into independent feature vectors belonging to user interests or information content based on the constructed boundaries.
[0027] Optionally, the step of constructing the user's interest feature space boundary and information feature space boundary based on the initial feature set includes: Core interest features and core information features are extracted from the initial feature set and mapped to obtain the interest feature matrix and information feature matrix, respectively. The interest feature matrix and the information feature matrix are respectively subjected to clustering and partitioning to obtain user interest feature clusters and information content feature clusters; Calculate the minimum distance threshold between the user interest feature cluster and the information content feature cluster respectively, so as to construct the user interest feature space boundary and the information feature space boundary.
[0028] The core interest features are the subset of key features extracted from the initial feature set that contribute most to distinguishing user interest patterns, such as the main dimensions reflecting industry preferences, risk tolerance, and operational activity. The core information features are the subset of key features obtained from the initial feature set that characterize the core themes and attributes of information, such as core indicators of the sector, event type, and sentiment polarity. The interest feature matrix and the information feature matrix are numerical matrices formed by arranging the core interest features and core information features, serving as the input data structure for cluster analysis. The user interest feature cluster and the information content feature cluster refer to sample sets with high internal similarity and large inter-group differences obtained after grouping their respective feature matrices using a clustering algorithm, such as a "value investment user cluster" or a "macroeconomic policy information cluster." The minimum distance threshold is a quantitative boundary value used to determine sample affiliation, determined by calculating the distance between the nearest sample points of two feature clusters in the vector space and comprehensively considering the intra-cluster distribution density.
[0029] Optionally, when extracting core features, a random forest or XGBoost model is used to train initial features to predict user behavior labels, and the top N features by importance are selected as core interest features; an LDA topic model is used to analyze the news text, and the dimension with the strongest dominance in the topic probability distribution is selected as the core news feature. When constructing the feature matrix, the core interest feature vector of each user is used as the rows to form the interest feature matrix; the core news feature vector of each news item is used as the rows to form the news feature matrix. When performing clustering, the silhouette coefficient is used to determine the optimal number of clusters, and the clustering algorithm is applied to both matrices respectively, assigning cluster labels to each cluster and calculating the cluster centroid. When calculating the minimum distance threshold, firstly, the Euclidean distances between all sample pairs between any user interest feature cluster and an information content feature cluster are calculated, and the minimum value is taken as the "nearest distance" between the two clusters. Then, the average of the "nearest distances" between the user interest feature cluster and all information content feature clusters is calculated, and a weighted correction is made by combining the average radius of the user interest feature cluster itself. The average radius and the average distance from the centroid of the samples within the cluster are used to obtain the final value, which is the minimum distance threshold between the user interest feature space and the information content feature space. This threshold defines that in the vector space, if the distance between a sample point and the centroid of the user interest cluster is less than this threshold, it is more likely to be classified into the interest feature space; otherwise, it may belong to the information feature space or outside its boundary. The boundary constructed by this threshold directly serves the subsequent feature vector separation and association matching degree calculation.
[0030] This invention quantifies the correlation strength between users and information by calculating the association matching degree between the interest feature vector and the information content feature vector, thereby facilitating the optimization of subsequent personalized recommendation ranking and distribution strategies. The association matching degree refers to a measure of similarity and contextual association between the interest feature vector and the information content feature vector.
[0031] As an embodiment of the present invention, calculating the correlation matching degree between the interest feature vector and the information content feature vector includes: Calculate the interaction information entropy and correlation entropy between the interest feature vector and the information content feature vector to determine the static correlation strength and dynamic coupling coefficient between the interest feature vector and the information content feature vector; Calculate the cross-entropy gain factor between the interest feature vector and the information content feature vector; The correlation matching degree between the interest feature vector and the information content feature vector is calculated by combining the static correlation strength and the dynamic coupling coefficient.
[0032] Wherein, the interaction information entropy refers to a static correlation metric obtained by calculating the KL divergence between the joint probability distribution and the marginal probability distribution of two feature vectors, based on information theory principles, measuring the amount of shared information between them; the correlation entropy refers to a dynamic coupling metric obtained by calculating the inner product of two feature vectors in a high-dimensional reproducing kernel Hilbert space, based on the kernel function method, measuring the degree of nonlinear correlation between them; the static correlation strength refers to a quantitative index reflecting the global dependency between interest features and information content, calculated by interaction information entropy; the dynamic coupling coefficient refers to a quantitative index reflecting the local nonlinear correlation between interest features and information content, calculated by correlation entropy; and the cross-entropy gain factor is an adjustment parameter calculated based on the distribution difference between the interest feature vector and the information content feature vector, used to dynamically adjust the weight of the synergistic effect between the two in the matching degree formula.
[0033] Optionally, the interaction information entropy between the interest feature vector and the information content feature vector can be calculated using a nonparametric estimation method based on k-nearest neighbors. Specifically, this involves: first, normalizing the interest feature vector and the information content vector; then, using a KSG estimator to calculate the mutual information value between them, which is the interaction information entropy; based on the interaction information entropy, mapping its value range to the [0,1] interval and directly using it as an intensity index to determine the static association strength; the correlation entropy between the interest feature vector and the information content feature vector can be calculated using a Gaussian kernel function-based method. Specifically, this involves: first, mapping the interest feature vector and the information content vector... The process involves retrieving the feature vector from a high-dimensional space and then calculating the inner product of the two features in that high-dimensional space. This inner product value is the correlation entropy. Based on the correlation entropy, the dynamic coupling coefficient is determined by normalizing the ratio of the correlation entropy to the maximum possible correlation entropy. The process of calculating the cross-entropy gain factor between the interest feature vector and the information content feature vector is as follows: First, the Softmax function is applied to both feature vectors to convert them into probability distributions. Second, the cross-entropy value is calculated based on these two probability distributions to quantify their distribution differences. Finally, the cross-entropy value is mapped and transformed using a negative exponential function to obtain the gain factor used to adjust the collaborative weights. This factor is inversely proportional to the distribution difference.
[0034] Optionally, as another optional embodiment of the present invention, the correlation matching degree between the interest feature vector and the information content feature vector is calculated using the following formula, combining the static correlation strength, the dynamic coupling coefficient, and the cross-entropy gain factor:
[0035] Where M represents the association matching degree, I represents the static association strength, C represents the dynamic coupling coefficient, and λ represents the cross-entropy gain factor.
[0036] S2. Determine the user's real-time scenario attributes and, in conjunction with the correlation matching degree, calculate the content adaptation priority of securities information in the securities big data platform during the push process.
[0037] This invention, by determining the real-time scenario attributes of a user when receiving information and combining this with the correlation matching degree, can quantify the actual value sequence of each piece of information to the user in a specific scenario, providing a direct basis for subsequent accurate push ranking. Here, the user is the recipient of securities information, i.e., an investor or analyst who needs to obtain information through the platform; the real-time scenario attributes refer to the immediate environment and state that affect the effectiveness of the user's reception and processing of information, such as the user's device type, current network status, and the current market time; the content adaptation priority refers to the order in which information should be pushed to the user in a specific scenario. Furthermore, the user's real-time scenario attributes can be determined by reading the user's device system information and network status interface, such as the user's login terminal type, current network connection method, and the location corresponding to the device's location. This information can directly relate to the display format and content focus of the information.
[0038] As an embodiment of the present invention, the step of calculating the content adaptation priority of securities information in the securities big data platform during the push process, based on the correlation matching degree, includes: Obtain the screen display specifications of the user's device to calculate the effective display area of the screen; Calculate the presentation complexity of the securities information on the currently used device to calculate the real-time understanding load corresponding to the securities information; By combining the real-time scene attributes, the correlation matching degree, the effective screen display area, and the instant understanding load, the content adaptation priority of the securities information during the push process is calculated.
[0039] The screen display specifications refer to the combination of the display area and resolution of the user's current device, which determines the information density that a single screen can carry. For example, the screen size of a smartphone is 6.1 inches. The presentation complexity refers to the comprehensive evaluation value of the elements that affect the reading experience, such as the density of data tables, charts, and professional terms contained in the securities information itself. The real-time comprehension load refers to the estimated value of the cognitive effort required for a user to fully understand the core content of a piece of information in a specific scenario.
[0040] Optionally, the screen display specifications can be obtained by calling the display application programming interface of the mobile or desktop operating system; the presentation complexity can be obtained by statistical analysis of the text length, number of charts, and data point density of the information.
[0041] Optionally, as another optional embodiment of the present invention, the content adaptation priority of the securities information during the push process is calculated using the following formula, combining the real-time scene attributes, the correlation matching degree, the effective screen display area, and the instant understanding load:
[0042] Where δ represents the content adaptation priority of securities information during the push process, H represents the correlation matching degree, k represents the adjustment constant related to real-time scene attributes, L represents the real-time understanding load, and S represents the effective display area of the screen.
[0043] Wherein, k represents the adjustment constant related to real-time scene attributes. It is used to transform the qualitative description of the scene into a quantitative scaling factor that affects the user's information processing ability. It can be obtained by querying a predefined scene state and influence intensity value mapping table. Specifically, the system first identifies key scene labels such as "user is moving" and "market is in the opening call auction stage" through device sensors and network status. Then, according to the preset empirical mapping relationship, for example, "moving" corresponds to k=1.3 and "after market closure" corresponds to k=0.7, the corresponding adjustment constant value of the formula is directly assigned.
[0044] Optionally, the calculation of the real-time understanding load corresponding to the information based on the presentation complexity includes: Identify the professional financial entities within the main text of the securities information; The parsing difficulty of the information text is evaluated, and the professional entity density of the professional financial entity is calculated to obtain the real-time comprehension load of the information.
[0045] The term "specialized financial entity" refers to a set of technical terms and data indicators that appear in the text of the information and require a specific financial knowledge background to fully understand, such as stock codes of listed companies, macroeconomic indicator terms, and names of financial derivatives. The term "analysis difficulty" refers to the cognitive effort required for users to understand the logical structure, data relationships, and conclusion derivations of the text of the information. The term "specialized entity density" refers to the number of specialized financial entities contained within a unit of text length, used to quantify the concentration of specialized information in the text.
[0046] Optionally, when extracting and identifying specialized financial entities, a pre-defined financial terminology database and securities code database can be loaded first. Then, the information text is segmented, and the segmentation results are precisely matched with the terminology and code databases to select all successfully matched terms as the set of identified entities. When evaluating the difficulty of interpretation, the paragraph structure of the information text can be analyzed first, counting the number of paragraphs containing independent data conclusions. The frequency of causal and comparative conjunctions in the text, such as "therefore," "because," "however," and "compared to," can be combined to calculate a structural difficulty score using predefined scoring rules. Simultaneously, it is checked whether the information contains data statements requiring cross-referencing (such as "increased by X% compared to the previous quarter"), as such statements will increase the interpretation difficulty score. When calculating the density of specialized entities, the formula is used: Specialized Entity Density = (Total Number of Identified Entities / Total Number of Words in the Text) 100. When comprehensively calculating the immediate load, a weighted summation formula is used: Immediate Load = ω1 Professional entity density +ω2 The analysis difficulty is determined by ω1 and ω2, which are normalized weighting coefficients calibrated based on experimental data to ensure that the load value falls within the preset range.
[0047] S3. Obtain the user's historical transaction frequency, position size, and information interaction rate data to calculate the user's urgency of needing securities information and determine the user's target information push accuracy.
[0048] This invention quantifies the urgency of a user's demand for securities information at a given time by acquiring the user's historical transaction frequency, position size, and information interaction rate data. Specifically, the historical transaction frequency is the cumulative number of securities transaction orders placed by the user within a specific statistical period; the position size is the total real-time market value of all securities currently held by the user; the information interaction rate data is a statistical record set of the user's clicks, reads, and in-depth interactions with information pushed by the platform; the urgency of demand is a value reflecting the intensity of the user's demand for timely information, obtained through a demand calculation function after comprehensively considering the user's trading behavior, position changes, and information attention habits; and the target information push accuracy is the required accuracy of information filtering and classification to meet the user's personalized information content requirements. Furthermore, by analyzing the information filtering conditions and attention lists set by the user, a preset accuracy comparison table is used to determine the personalized push accuracy level.
[0049] As an embodiment of the present invention, calculating the urgency of the user's need for securities information includes: Based on the historical transaction frequency, calculate the recent transaction activity of the user and the market value volatility corresponding to the holding size; Based on the information interaction rate data, the user's response sensitivity to information is calculated; By combining the recent trading activity, the market capitalization volatility, and the response sensitivity, the urgency of the user's need for securities information is calculated.
[0050] The recent trading activity refers to a quantitative indicator reflecting the frequency of a user's trading behavior, calculated by analyzing the user's buy and sell order records within a preset time window; the market value volatility refers to a statistical indicator reflecting the degree of change in the total value of a user's holdings, calculated by analyzing the historical price data of each security in the user's portfolio; and the response sensitivity refers to a quantitative indicator reflecting the intensity of a user's attention to and timely feedback on information, calculated by analyzing the user's clicks, reading time, and secondary dissemination behavior on pushed information.
[0051] Optionally, when calculating recent transaction activity, the analysis time window can be set to the most recent 30 trading days. Then, the transaction system logs can be used to query all order records of users with the status of "completed" during this period, calculate the average number of completed orders per day, and compare this value with the average daily level of all users on the platform. The value is then normalized using the percentile ranking method to finally obtain a standardized activity value in the range of 0-1.
[0052] Optionally, when calculating response sensitivity, a recent observation window can be set, such as the past 7 days. The proportion of "effective reading" times of push information by users within this window is counted to the total number of pushes. The frequency of users' active search behavior is also taken into account. A sensitivity value in the range of 0-1 is obtained through linear mapping. Effective reading can be defined as reading time greater than 15 seconds.
[0053] Optionally, by combining the recent trading activity, the market capitalization volatility, and the response sensitivity, the urgency of the user's need for securities information can be calculated using the following formula:
[0054] Where U represents the urgency of a user's need for securities information, A represents recent trading activity, V represents market capitalization volatility, and Y represents responsiveness.
[0055] This formula reflects that trading activity and position risk together constitute the basis of demand, and is non-linearly adjusted through responsiveness: when users are sensitive to information, that is, when the responsiveness is close to 1, the urgency of demand is closer to the base value. When users are not sensitive to information responses, that is, when the response sensitivity is close to 0, the urgency of the need drops to half of the base value.
[0056] Optionally, calculating the user's recent transaction activity based on the historical transaction frequency includes: Based on the historical transaction frequency, the user's average daily transaction frequency is calculated; Query the average daily transaction frequency of the user's reference user group to obtain the group's benchmark level; Based on the average daily transaction frequency and the group's benchmark level, the user's recent transaction activity level is obtained.
[0057] The preset time window is the most recent 30 trading days; the reference user group refers to a set of other users with similar asset size, investment experience and trading preferences as the target user; the group benchmark level refers to the statistical average of the daily trading frequency of the reference user group within the same time window.
[0058] Optionally, when extracting the cumulative number of trades, the system can obtain all transaction order records of the user within the most recent 30 trading days through the securities trading system database interface, and calculate the total number of orders as the cumulative number of trades. When calculating the average daily trading frequency, the cumulative number of trades is divided by 30 trading days to obtain the user's average daily trading frequency. When querying the group benchmark level, reference user groups that meet the similarity criteria are first selected from the entire platform's user database based on three dimensions: user account asset size, account opening time, and main trading instrument types. Then, the average daily trading frequency of each user in this group within the most recent 30 trading days is calculated. Finally, the arithmetic mean of the average daily trading frequencies of all reference users is taken as the group benchmark level. When calculating the ratio to obtain the recent trading activity, the formula is: Recent Trading Activity = User Average Daily Trading Frequency / Group Benchmark Level.
[0059] S4. Calculate the effective push weight of the securities information by combining the content adaptation priority, the urgency of the demand, and the correlation matching degree. Combine this with the target information push accuracy to perform content arrangement and push processing on the securities information to obtain the push result.
[0060] This invention, by combining content adaptation priority, urgency of need, and relevance matching degree to calculate effective push weight, can quantitatively evaluate the comprehensive push value of a single piece of information in a specific user and scenario, providing a precise numerical basis for information distribution and ranking. The effective push weight refers to a comprehensive score calculated considering the user's immediate context, the urgency of information need, and content matching degree, used to determine the final ranking of information in the user's push queue.
[0061] As an embodiment of the present invention, the calculation of the effective push weight of the securities information by combining the content adaptation priority, the urgency of the demand, and the relevance matching degree includes: Obtain raw data on the dissemination behavior of the securities information to calculate the information dissemination popularity index; Based on the aforementioned dissemination popularity index, the intrinsic value coefficient corresponding to the securities information is determined; The intrinsic value coefficient is weighted by combining the content adaptation priority, the urgency of the demand, and the relevance matching degree to obtain the effective push weight of the securities information.
[0062] The original dissemination data refers to the set of records of the number of times the securities information was viewed, clicked, collected, and shared by users within a preset statistical period after its release. The dissemination popularity index is a comprehensive quantitative value representing the speed of information dissemination and the degree of audience attention, obtained by assigning corresponding scores to each behavior type in the original data according to the differences in the depth of user attention they reflect and by performing time decay processing. The intrinsic value coefficient is a core indicator reflecting the basic value of the information itself, obtained by adjusting the dissemination popularity index in combination with the authoritativeness of the information source and the originality of the content.
[0063] Optionally, when calculating the dissemination popularity index, the statistical period can be initially set to 24 hours after the information is published. Then, the total exposure, number of unique clicks, number of valid reads completed, and number of shares of the information within this period can be extracted from platform logs. Preset scores can be assigned to each type of behavioral data, such as an exposure score coefficient of 0.2, a click score coefficient of 0.3, a read completion score coefficient of 0.3, and a share score coefficient of 0.2. The result of the composite summation is then multiplied by a decay factor that decreases based on the publication duration, such as decay factor = 1 / (1+0.1). The number of hours of publication is used to obtain a dissemination popularity index in the range of 0-100.
[0064] Optionally, a basic range can be first divided based on the dissemination popularity index, such as an index ≥ 0.8 corresponding to a basic coefficient of 1.0, 0.5-0.8 corresponding to 0.8, and < 0.5 corresponding to 0.6. Then, the popularity decay characteristics of securities information can be adjusted, and the popularity decay characteristics of information type can be further adjusted to determine the intrinsic value coefficient corresponding to the securities information.
[0065] Optionally, the intrinsic value coefficient is weighted by combining the content adaptation priority, the urgency of demand, and the relevance matching degree to obtain the effective push weight of the securities information. First, the original values of content adaptation priority, urgency of demand, and relevance matching degree are normalized to the [0,1] interval. The three are then weighted and summed with weights of 0.4, 0.3, and 0.3 respectively to obtain a comprehensive adjustment coefficient. Then, this coefficient is multiplied by the intrinsic value coefficient, and the result is normalized to the [0,1] interval, which is the effective push weight of the securities information. Specifically, the weight classification can be carried out by combining expert scoring method with actual application scenarios.
[0066] This invention improves the accuracy of securities information delivery by arranging and pushing the information based on the effective push weight and the target information push accuracy.
[0067] Optionally, based on the effective push weight and the target information push accuracy, the securities information is processed for content arrangement and push to obtain the push result. The specific steps are as follows: First, according to the effective push weight value, information that meets the weight standard is selected to form a set of information to be pushed; Second, according to the upper limit of the number of items corresponding to the target information push accuracy, the corresponding number of information items with the highest ranking are selected from the set of information to be pushed to form the final push batch; Finally, according to the user device type and the current network conditions, the information in the push batch is subjected to lightweight packaging processing such as title refinement, summary generation, and image and text format adaptation, and delivered through a designated channel to output a complete push result that conforms to the user's receiving habits and the carrying capacity of the current scenario. Specifically, the selection criteria are set according to the actual business scenario.
[0068] like Figure 3 The diagram shown is a functional module diagram of the intelligent and precise securities information push system based on big data according to the present invention.
[0069] The big data-based intelligent and precise securities information push system 300 described in this invention can be installed in an electronic device. Depending on the functions implemented, the big data-based intelligent and precise securities information push system 300 may include a data matching calculation module 301, a scene adaptation calculation module 302, a push accuracy determination module 303, and a push weight calculation module 304. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.
[0070] In this embodiment of the invention, the functions of each module / unit are as follows: The data matching and calculation module 301 is used to collect multi-dimensional data sources from the securities big data platform, separate user interest feature vectors and information content feature vectors from the multi-dimensional data sources, and calculate the correlation matching degree between the interest feature vectors and the information content feature vectors. The scenario adaptation calculation module 302 is used to determine the user's real-time scenario attributes and, in conjunction with the correlation matching degree, calculate the content adaptation priority of securities information in the securities big data platform during the push process. The push accuracy determination module 303 is used to obtain the user's historical transaction frequency, holding size and information interaction rate data, so as to calculate the urgency of the user's demand for securities information and determine the user's target information push accuracy. The push weight calculation module 304 is used to calculate the effective push weight of the securities information by combining the content adaptation priority, the urgency of the demand and the correlation matching degree, so as to combine the target information push accuracy to perform content arrangement and push processing on the securities information to obtain the push result.
[0071] In detail, the modules in the big data-based intelligent and precise securities information push system 300 described in this embodiment of the invention employ the same methods as described above. Figure 1 The method uses the same technical means as the big data-based intelligent and accurate securities information push method described in the article, and can produce the same technical effect, so it will not be elaborated here.
[0072] In one embodiment, a computer device is provided, which may be a server or a client, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used for communication with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a big data-based intelligent and precise securities information push method on the server or client side.
[0073] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0074] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.
[0075] Finally, it should be noted that in the above embodiments, each embodiment can be combined with each other or independent. Deleting any one of them will not affect the technical implementation of other embodiments. The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for intelligent and precise delivery of securities information based on big data, characterized in that: The method includes: Collect multi-dimensional data sources from a securities big data platform, separate user interest feature vectors and information content feature vectors from the multi-dimensional data sources, and calculate the correlation matching degree between the interest feature vectors and the information content feature vectors; Determine the user's real-time scenario attributes and, in conjunction with the correlation matching degree, calculate the content adaptation priority of securities information in the securities big data platform during the push process; The system acquires the user's historical transaction frequency, position size, and information interaction rate data to calculate the urgency of the user's need for securities information and determine the target information push accuracy for the user. By combining the content adaptation priority, the urgency of the demand, and the relevance matching degree, the effective push weight of the securities information is calculated. Combined with the target information push accuracy, the securities information is then processed for content arrangement and push to obtain the push result.
2. The intelligent and precise securities information push method based on big data as described in claim 1, characterized in that, The process of separating user interest feature vectors and information content feature vectors from the multi-dimensional data source includes: Extract the structured data set from the multi-dimensional data source to extract the user's behavioral features and information text features, and obtain an initial feature set; Based on the initial feature set, the user's interest feature space boundary and information feature space boundary are constructed respectively, so as to separate the user's interest feature vector and information content feature vector from the structured data set.
3. The intelligent and precise securities information push method based on big data as described in claim 2, characterized in that, The step of constructing the user's interest feature space boundary and information feature space boundary based on the initial feature set includes: Core interest features and core information features are extracted from the initial feature set and mapped to obtain the interest feature matrix and information feature matrix, respectively. The interest feature matrix and the information feature matrix are respectively subjected to clustering and partitioning to obtain user interest feature clusters and information content feature clusters; Calculate the minimum distance threshold between the user interest feature cluster and the information content feature cluster respectively, so as to construct the user interest feature space boundary and the information feature space boundary.
4. The intelligent and precise securities information push method based on big data as described in claim 1, characterized in that, The calculation of the correlation matching degree between the interest feature vector and the information content feature vector includes: Calculate the interaction information entropy and correlation entropy between the interest feature vector and the information content feature vector to determine the static correlation strength and dynamic coupling coefficient between the interest feature vector and the information content feature vector; Calculate the cross-entropy gain factor between the interest feature vector and the information content feature vector; The correlation matching degree between the interest feature vector and the information content feature vector is calculated by combining the static correlation strength and the dynamic coupling coefficient.
5. The intelligent and precise securities information push method based on big data as described in claim 1, characterized in that, The calculation of the content adaptation priority of securities information in the securities big data platform during the push process, based on the aforementioned correlation matching degree, includes: Obtain the screen display specifications of the user's device to calculate the effective display area of the screen; Calculate the presentation complexity of the securities information on the currently used device to calculate the real-time understanding load corresponding to the securities information; By combining the real-time scene attributes, the correlation matching degree, the effective screen display area, and the instant understanding load, the content adaptation priority of the securities information during the push process is calculated.
6. The intelligent and precise securities information push method based on big data as described in claim 5, characterized in that, The calculation of the real-time understanding load corresponding to the information based on the presentation complexity includes: Identify the professional financial entities within the main text of the securities information; The parsing difficulty of the information text is evaluated, and the professional entity density of the professional financial entity is calculated to obtain the real-time comprehension load of the information.
7. The intelligent and precise securities information push method based on big data as described in claim 1, characterized in that, The calculation of the user's urgency regarding securities information includes: Based on the historical transaction frequency, calculate the recent transaction activity of the user and the market value volatility corresponding to the holding size; Based on the information interaction rate data, the user's response sensitivity to information is calculated; By combining the recent trading activity, the market capitalization volatility, and the response sensitivity, the urgency of the user's need for securities information is calculated.
8. The intelligent and precise securities information push method based on big data as described in claim 7, characterized in that, The calculation of the user's recent transaction activity based on the historical transaction frequency includes: Based on the historical transaction frequency, the user's average daily transaction frequency is calculated; Query the average daily transaction frequency of the user's reference user group to obtain the group's benchmark level; Based on the average daily transaction frequency and the group's benchmark level, the user's recent transaction activity level is obtained.
9. The intelligent and precise securities information push method based on big data as described in claim 1, characterized in that, The calculation of the effective push weight of the securities information, combining the content adaptation priority, the urgency of the demand, and the relevance matching degree, includes: Obtain raw data on the dissemination behavior of the securities information to calculate the information dissemination popularity index; Based on the aforementioned dissemination popularity index, the intrinsic value coefficient corresponding to the securities information is determined; The intrinsic value coefficient is weighted by combining the content adaptation priority, the urgency of the demand, and the relevance matching degree to obtain the effective push weight of the securities information.
10. A securities information intelligent and precise push system based on big data, characterized in that: The system includes: The data matching and calculation module is used to collect multi-dimensional data sources from the securities big data platform, separate user interest feature vectors and information content feature vectors from the multi-dimensional data sources, and calculate the correlation matching degree between the interest feature vectors and the information content feature vectors. The scene adaptation calculation module is used to determine the user's real-time scene attributes and, in combination with the correlation matching degree, calculate the content adaptation priority of securities information in the securities big data platform during the push process. The push accuracy determination module is used to obtain the user's historical transaction frequency, holding size and information interaction rate data to calculate the urgency of the user's demand for securities information and determine the user's target information push accuracy. The push weight calculation module is used to calculate the effective push weight of the securities information by combining the content adaptation priority, the urgency of the demand, and the relevance matching degree. In combination with the target information push accuracy, the securities information is processed for content arrangement and push to obtain the push result.