A message push method and system based on grid data

Abstract

The invention discloses a message push method and system based on grid data, which relates to the technical field of message push. The feedback degree is constructed by a user feedback data set. If the obtained feedback degree is lower than expected, the user and related data are collected; a corresponding grid division scheme is selected to divide the message push area into a plurality of sub-areas, and after the user predicted position enters the selected target sub-area, an information push instruction is issued to the outside; after matching the corresponding message push strategy for the user, the message content is generated and pushed to the user, and the push frequency of the message content is constrained according to the feedback data; the message push strategy is optimized according to key factors, and the grid division in the grid area is dynamically adjusted according to the constructed uniformity. The push frequency to the user is constrained and controlled, and by adjusting the message push frequency to the user, it can be avoided that too many push messages harass the user.

Description

Technical Field

[0001] The present invention relates to the technical field of message push, and in particular to a message push method and system based on grid data. Background Art

[0002] Advertising message push is a technical means to analyze and mine data such as user interests, geographic location, search history and user behavior, and then push advertising content that matches the user's interests to the user. It allows companies or individuals to send customized advertising content to users through mobile devices, web pages, social media platforms and other channels, using big data analysis and user behavior tracking to achieve accurate advertising delivery and improve advertising effectiveness and conversion rates.

[0003] In a Chinese invention patent with application publication number CN110335081A, a method, system and medium for delivering message push type advertisements are disclosed. The delivery method of the present invention obtains a third party's advertisement push request through an external message push SDK; forwards the advertisement push request to an internal advertisement SDK through an external message push SDK; initiates an advertisement material request to an internal advertisement material server according to the advertisement push request through the internal advertisement SDK, and obtains the advertisement material corresponding to the reviewed advertisement push request; displays the obtained advertisement material through the internal advertisement SDK, synchronizes the advertisement display message to the external message push SDK, and returns the message to the third party that initiated the advertisement push request through the external message push SDK, which can effectively control the standardization of the advertisement content of the external message push SDK and realize effective control of message type advertisements.

[0004] Combined with the above application and the contents of the prior art:

[0005] When pushing messages to users based on grid data, usually after collecting various preference data of users, a recommendation algorithm generates message content based on the user's location data, and pushes the message to the user. As the user's location changes, the push message received by the user may change. However, in the existing message push method based on grid data, after completing the initial message push, the user's feedback data will be collected, and then the current message push strategy will be adjusted and optimized based on the obtained feedback data to make the pushed message content more matched with the user's interests. Although this recommendation method is highly targeted, it is often due to the high frequency of pushes and the random timing of pushes. It is easy for the user to no longer be interested after the push is completed, resulting in the difficulty in achieving the expected effect when pushing messages.

[0006] Therefore, the present invention provides a message push method and system based on grid data. Summary of the invention

[0007] 1. Technical issues to be resolved

[0008] In view of the deficiencies of the prior art, the present invention provides a message push method and system based on grid data, which divides the message push area into several sub-areas, and sends an information push instruction to the outside after the user's predicted position enters the selected target sub-area; after matching the corresponding message push strategy for the user, the message content is generated and pushed to the user, and the push frequency of the message content is constrained according to the feedback data; the message push strategy is optimized according to key factors, and the grid division within the grid area is dynamically adjusted according to the constructed uniformity. The push frequency to the user is constrained and controlled, and by adjusting the push frequency of the user's message, it is possible to avoid harassment to the user by too many pushed messages, thereby solving the technical problems raised in the background technology.

[0009] (II) Technical solution

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0011] A message push method based on grid data includes collecting user feedback data after message push, aggregating and constructing a user feedback data set, constructing a feedback degree Fts from the user feedback data set, and if the obtained feedback degree Fts is lower than expected, collecting user and related data; wherein the feedback degree Fts is constructed in the following manner: linear normalization is performed on the click rate Np and the reading rate Zp, and the corresponding data values ​​are mapped to the interval [0, 1] according to the following method:

[0012]

[0013] Weight coefficient: 0≤Q1≤1, 0≤Q2≤1, and Q2+Q1=1; i=1, 2, …, n, n is the total number of pushed messages, Np i is the click rate of the i-th push message, Np max and Np min is the maximum and minimum value of the click rate; Zp i is the reading rate of the i-th push message, Zp max and Zp min is the maximum and minimum value of the reading rate;

[0014] According to the status in the message push area, select the corresponding grid division scheme to divide the message push area into several sub-areas. After the user's predicted position enters the selected target sub-area, send an information push instruction to the outside;

[0015] Collect data in the target sub-area and extract push features. After matching the corresponding message push strategy for the user, generate message content and push it to the user, and constrain the push frequency of the message content based on the feedback data;

[0016] The key factors are selected from several push condition parameters, the message push strategy is optimized based on the key factors, the message is pushed to the user again and feedback data is obtained, and the grid division in the grid area is dynamically adjusted based on the constructed uniformity Yts.

[0017] Furthermore, after the message is pushed to the user, the feedback data of each user on the pushed message is collected within the feedback cycle, and the feedback data is aggregated to construct a user feedback data set; the feedback degree Fts is constructed from the user feedback data set. If the obtained feedback degree Fts is lower than the feedback threshold, the user and related data are collected, including: identity information data, user's basic information, behavior data, geographic information data; after the above data is aggregated, a push condition data set is constructed.

[0018] Furthermore, the data in the push condition data set is preprocessed to obtain the preprocessed data, the push area is determined as a grid area, and after obtaining the user location data and user-related data in the grid area, they are summarized as grid-related data; feature extraction is performed on the grid-related data to obtain corresponding grid data features; a grid division scheme library is constructed, and according to the correspondence between the grid data features and the grid division schemes, the corresponding grid division schemes are matched in the grid division scheme library, the grid division schemes are executed, the grid division is implemented, and a number of sub-areas are obtained.

[0019] Furthermore, an electronic map covering the grid area is constructed, and each sub-area is displayed on the electronic map. Based on the user's historical location data, a user location prediction model is constructed by convolutional neural network training. After obtaining the user's location information in real time, the location prediction model predicts the user's next location data and uses it as the predicted location. If the predicted location enters the selected target sub-area, an information push instruction is issued to the outside.

[0020] Furthermore, after receiving the information push instruction, the marketing information, weather and traffic data in the target sub-area are collected and used as push environment data; feature extraction is performed on the acquired push environment data, identity information data, behavior data and preference data to obtain corresponding push features.

[0021] Furthermore, based on the push features, the pre-built message push strategy knowledge graph matches the corresponding message push strategy for the user; the message recommendation model is obtained by training the recommendation algorithm, and based on the message push strategy, the message recommendation model selects message content that meets the user's interests, sends the message content to the user, and issues push frequency adjustment instructions to the outside.

[0022] Furthermore, after receiving the push frequency adjustment instruction, the push frequency of the message content is constrained according to the feedback data. If used in the target sub-area, the user message push frequency is constrained in the target sub-area according to the constraint conditions, and the message content push is terminated after the user leaves the target sub-area; the constraint method is as follows:

[0023]

[0024] Where n is the number of times the message is pushed, V(i, j) is the time interval from the i-th push to the j-th push of the message content, and V avg is the mean time interval, Fts is the feedback degree when the message is pushed in the target sub-area, and Vt is the average moving speed of users in the target sub-area.

[0025] Furthermore, after receiving the feedback data within the analysis period, the corresponding feedback degree Fts is constructed, and multiple linear regression analysis is performed with the user's push condition parameters as independent variables and the feedback degree Fts as dependent variables, and the corresponding linear regression equation is obtained;

[0026] The regression coefficient corresponding to each independent variable is obtained from the linear regression equation, and the regression coefficient is used as the influence, and the independent variables with the largest influence are used as key factors; the improvement of the feedback degree Fts is used as the optimization goal, and the message push strategy is optimized by the optimization algorithm; the optimized message push strategy is executed, and the corresponding message content is re-obtained by the message recommendation model.

[0027] Furthermore, after obtaining the number, position and feedback degree Fts of the target sub-regions, the uniformity Yts between the target sub-regions in the grid region is constructed, and the grid division strategy in the grid region is dynamically adjusted according to the uniformity Yts, and the coarseness of the grid division is adjusted so that the coarseness is inversely proportional to the uniformity Yts; the uniformity Yts is constructed as follows, under dimensionless conditions according to the following method:

[0028]

[0029] Weight coefficient: 0≤β≤1, 0≤α≤1, and α+β=1. The weight coefficient can be obtained by referring to the hierarchical analysis method; Fts i is the feedback degree in the ith sub-region, Fts avg is the mean value of feedback degree, d ij is the distance from the i-th target sub-region to the j-th target sub-region; i = 1, 2, ..., p, p is the number of target sub-regions; Po i is the uniform mean value in the i-th target sub-region, Po avg is the average value of uniformity, It is the qualified standard value of the mean value in the uniformity.

[0030] A message push system based on grid data, comprising: a feedback evaluation unit, which collects user feedback data after message push, summarizes and constructs a user feedback data set, constructs a feedback degree Fts from the user feedback data set, and collects user and related data if the obtained feedback degree Fts is lower than expected;

[0031] The position prediction unit selects a corresponding grid division scheme according to the state in the message push area, divides the message push area into several sub-areas, and sends an information push instruction to the outside after the user's predicted position enters the selected target sub-area;

[0032] The recommendation unit collects data in the target sub-area and extracts push features. After matching the corresponding message push strategy for the user, it generates message content and pushes it to the user, and constrains the push frequency of the message content based on the feedback data.

[0033] The area division unit selects key factors from several push condition parameters, optimizes the message push strategy based on the key factors, re-pushes messages to users and obtains feedback data, and dynamically adjusts the grid division within the grid area based on the constructed uniformity Yts.

[0034] (III) Beneficial effects

[0035] The present invention provides a message push method and system based on grid data, which has the following beneficial effects:

[0036] 1. Determine whether the previous message push effect is qualified based on the feedback degree Fts. If it is qualified, the original message push strategy can be maintained to continue pushing. If it is unqualified, the push strategy and method need to be adjusted to ensure the effectiveness of the message push after a comprehensive evaluation.

[0037] 2. The corresponding grid division scheme is matched by the pre-built grid division scheme library, which makes the message push more targeted and improves the efficiency of grid division. The user's location data is predicted and used as a trigger point to push messages to the user. After the user's location data is obtained, the message is pushed again, which can make the pushed content more consistent with the actual environment scene, thereby obtaining better feedback effects.

[0038] 4. The message push strategy knowledge graph matches the corresponding message push strategy for the user. Based on the given message push strategy, the recommendation algorithm is used to extract and generate content of interest to the user. The matching between various data is higher. When pushing messages, the various data around the user's real-time location are taken into consideration, which can also increase the user's interest in the pushed content. By constraining and controlling the push frequency to users and adjusting the message push frequency, it is possible to avoid harassing users by pushing too many messages and reduce user disgust.

[0039] 5. Determine the degree of influence of each push condition parameter as an independent variable on the feedback degree Fts, select key parameters from several push condition parameters according to the degree of influence, and obtain the optimized message push strategy, so as to have better feedback effect when recommending messages to users.

[0040] 6. Evaluate the effect of grid area division based on uniformity Yts. When there are large differences in user density and feedback data in each target sub-area, redivide the grid area to save costs and reduce resource waste when pushing messages. It can also reallocate investment in message push, improve the efficiency of message push, and accelerate the iteration of message push content and effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the flow of the message push method based on grid data of the present invention;

[0042] Figure 2 It is a schematic diagram of the message push method system based on grid data of the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0044] See also Figure 1 The present invention provides a message push method based on grid data, comprising:

[0045] Step 1: Collect user feedback data after the message is pushed, summarize and construct a user feedback data set, and construct a feedback degree Fts from the user feedback data set. If the obtained feedback degree Fts is lower than expected, collect user and related data;

[0046] The step 1 includes the following contents:

[0047] Step 101: After the message is pushed to the user, a feedback cycle is set to collect feedback data from each user on the pushed message during the feedback cycle, including the click rate, reading rate, conversion rate, bounce rate and dwell time of each user on the pushed message, and the feedback data is aggregated to form a user feedback data set;

[0048] Step 102: construct the feedback degree Fts from the user feedback data set, wherein the click rate Np and the reading rate Zp are linearly normalized and the corresponding data values ​​are mapped to the interval [0, 1] according to the following method:

[0049]

[0050] Weight coefficient: 0≤Q1≤1, 0≤Q2≤1, and Q2+Q1=1; i=1, 2, …, n, where n is the total number of pushed messages. The weight coefficient can be obtained by referring to the hierarchical analysis method; Np i is the click rate of the i-th push message, Np max and Np min is the maximum and minimum value of the click rate; Zp i is the reading rate of the i-th push message, Zp max and Zp min is the maximum and minimum value of the reading rate;

[0051] Based on historical data and management expectations for message push, a feedback threshold is pre-set; if the obtained feedback degree Fts is lower than the feedback threshold, it means that the current message push effect is poor and fails to meet expectations, and timely adjustments are required. At this time, a data collection instruction is issued externally;

[0052] When in use, before sending a message push to the user, the corresponding feedback degree Fts is constructed after collecting relevant feedback data. The feedback degree Fts is used to judge whether the previous message push effect is qualified. If it is qualified, the original message push strategy can be maintained and the push can continue. If it is unqualified, the push strategy and method need to be adjusted to ensure the effectiveness of the message push after a comprehensive evaluation.

[0053] Step 103: after receiving the data collection instruction, collect the user and related data, including: identity information data, basic information of the user, such as age, gender, hobbies, behavior data, such as browsing history, purchase history, click behavior, etc.; geographic information data, such as map data, administrative division data, etc.; after aggregating the above data, construct a push condition data set;

[0054] When using, combine the contents in steps 101 to 103:

[0055] When it is necessary to push a message to the user again, data is collected again based on the user's location, and features are extracted from the collected data. Based on the extracted features, the message push is more targeted, thereby improving the current feedback effect;

[0056] Combined with the above application and the contents of the prior art:

[0057] When pushing messages to users based on grid data, usually after collecting various preference data of users, a recommendation algorithm generates message content based on the user's location data, and pushes the message to the user. As the user's location changes, the push message received by the user may change. However, in the existing message push method based on grid data, after completing the initial message push, the user's feedback data will be collected, and then the current message push strategy will be adjusted and optimized based on the obtained feedback data to make the pushed message content more matched with the user's interests. Although this recommendation method is highly targeted, it is often due to the high frequency of pushes and the random timing of pushes. It is easy for the user to no longer be interested after the push is completed, resulting in the difficulty in achieving the expected effect when pushing messages.

[0058] Step 2: According to the status in the message push area, select the corresponding grid division scheme, divide the message push area into several sub-areas, and send an information push instruction to the outside after the user's predicted position enters the selected target sub-area;

[0059] The step 2 includes the following contents:

[0060] Step 201: pre-process the data in the push condition data set to obtain pre-processed data, where the data pre-processing includes missing value processing and normalization, etc.; after determining the message push area, determine the push area as a grid area, obtain user location data and user-related data in the grid area, and aggregate them as grid-related data; perform feature extraction on the grid-related data to obtain corresponding grid data features;

[0061] Step 202: after searching or formulating several grid division schemes online, a grid division scheme library is constructed, and according to the correspondence between the grid data characteristics and the grid division schemes, a corresponding grid division scheme is matched in the grid division scheme library, and the grid division scheme is executed in the grid area, and the grid division is performed to obtain several sub-areas;

[0062] When in use, after determining the message push area, collect various data in the message push area, and after feature extraction, match the corresponding grid division scheme from the pre-built grid division scheme library, so that the message push is more targeted and the grid division efficiency can be improved;

[0063] Step 203: construct an electronic map covering the grid area, display each sub-area on the electronic map, and build a user location prediction model based on the user's historical location data by training a convolutional neural network. After obtaining the user's location information in real time, the location prediction model predicts the user's next location data and uses it as the predicted location. If the predicted location enters the selected target sub-area, an information push instruction is issued to the outside.

[0064] When using, combine the contents in steps 201 to 203:

[0065] After completing the grid division and obtaining the corresponding sub-areas, the user's location data is predicted, and the user's location data is used as a trigger point to push messages to the user. Pushing messages after obtaining the user's location data can make the pushed content more closely match the actual environment scene, thereby obtaining better feedback effects.

[0066] Step 3: Collect data in the target sub-area and extract push features. After matching the corresponding message push strategy for the user, generate message content and push it to the user, and constrain the push frequency of the message content based on the feedback data;

[0067] The step three includes the following contents:

[0068] Step 301: After receiving the information push instruction, collect the marketing information, weather and traffic data in the target sub-area and use them as push environment data; extract features from the acquired push environment data, identity information data, behavior data and preference data to obtain corresponding push features; and match the corresponding message push strategy for the user based on the push features using the pre-built message push strategy knowledge graph;

[0069] The message recommendation model is obtained by training the recommendation algorithm. According to the message push strategy, the message recommendation model selects the message content that meets the user's interests, sends the message content to the user, and issues a push frequency adjustment instruction to the outside. Among them, the recommendation algorithm includes content-based recommendation, collaborative filtering and graph-based recommendation, etc.

[0070] When in use, after collecting various data in the target sub-area, the message push strategy knowledge graph is used to match the corresponding message push strategy for the user through feature extraction and matching. Based on the given message push strategy, the recommendation algorithm is used to extract and generate content of interest to the user. The matching between various data is higher. When pushing messages, the data around the user's real-time location is taken into account, which can also increase the user's interest in the pushed content.

[0071] Step 302: After receiving the push frequency adjustment instruction, the push frequency of the message content is restricted according to the feedback data, in the following manner:

[0072]

[0073] Where n is the number of times the message is pushed, V(i, j) is the time interval from the i-th push to the j-th push of the message content, and V avg is the mean time interval, Fts is the feedback degree when the message is pushed in the target sub-area, and Vt is the average moving speed of users in the target sub-area;

[0074] If the user is still in the target sub-area, the message push frequency of the user is restricted in the target sub-area according to the constraint conditions, and the message content push is terminated after the user leaves the target sub-area;

[0075] When used, combine the contents in steps 301 and 302:

[0076] After completing the push content generation for the user, the recommendation is first sent to the user, and then after obtaining the feedback data in the target sub-area, the push frequency to the user is constrained and controlled. By adjusting the frequency of message push to the user, it is possible to avoid harassing the user by pushing too many messages and reduce user disgust.

[0077] Step 4: Filter out key factors from a number of push condition parameters, optimize the message push strategy based on the key factors, push messages to users again and obtain feedback data, and dynamically adjust the grid division within the grid area based on the constructed uniformity Yts;

[0078] The step 4 includes the following contents:

[0079] Step 401: Set an analysis period, construct a corresponding feedback degree Fts after receiving feedback data within the analysis period, use various user data, such as push condition parameters such as behavior preference and consumption capacity, as independent variables, and use the feedback degree Fts as a dependent variable to perform multiple linear regression analysis, and obtain the corresponding linear regression equation;

[0080] Obtain the regression coefficient corresponding to each independent variable from the linear regression equation, use the regression coefficient as the influence, and use the independent variables with the largest influence as the key factors; use the improvement of the feedback degree Fts as the optimization goal, adjust the weights of the key factors, and use the optimization algorithm to optimize the message push strategy; execute the optimized message push strategy, and use the message recommendation model to re-obtain the corresponding message content;

[0081] When in use, after obtaining the feedback data of each user, a multiple linear regression analysis is performed to determine the degree of influence of each push condition parameter as an independent variable on the feedback degree Fts, and key parameters are screened out from several push condition parameters based on the degree of influence. By adjusting the weights of key parameters, the optimized message push strategy is obtained, so that when recommending messages to users, better feedback effects can be achieved.

[0082] Step 402: After obtaining the number, position and feedback degree Fts of the target sub-regions, construct the uniformity Yts between the target sub-regions in the grid region according to the following method under dimensionless conditions:

[0083]

[0084] Weight coefficient: 0≤β≤1, 0≤α≤1, and α+β=1. The weight coefficient can be obtained by referring to the hierarchical analysis method; Fts i is the feedback degree in the ith sub-region, Fts avg is the mean value of feedback degree, d ij is the distance from the i-th target sub-region to the j-th target sub-region; i = 1, 2, ..., p, p is the number of target sub-regions; Po i is the uniform mean value in the i-th target sub-region, Po avg is the average value of uniformity, It is the qualified standard value of the mean value in uniformity;

[0085] Dynamically adjust the grid division strategy in the grid area according to the uniformity Yts, and adjust the coarseness of the grid division so that the coarseness is inversely proportional to the uniformity Yts;

[0086] When using, combine the contents in steps 401 and 402:

[0087] After obtaining the feedback of message push again, combined with the number and location of users and the uniformity Yts in the target sub-area, after completing the construction of uniformity Yts, the grid area division effect is evaluated according to uniformity Yts. Taking user feedback as the starting point, when there are large differences in user density and feedback data in each target sub-area, the grid area is re-divided. In this way, costs can be saved when pushing messages, resource waste can be reduced, the investment in message push can be reallocated, the efficiency of message push can be improved, and the iteration of message push content and effects can be accelerated.

[0088] The Analytic Hierarchy Process (AHP) is a decision-making method that decomposes decision-related elements into levels such as goals, criteria, and plans, and conducts qualitative and quantitative analysis on this basis. It is particularly suitable for dealing with target systems with hierarchical and staggered evaluation indicators, and when the target value is difficult to describe quantitatively, the AHP is an effective decision-making tool.

[0089] The core of the hierarchical analysis method is to decompose the decision problem into multiple levels to form a hierarchical structure, which usually includes the target level, the criterion level, the sub-criterion level and the solution level. By solving the eigenvector of the judgment matrix, the priority weight of each element of each level to an element of the previous level is obtained, and the final weight of each alternative solution to the total goal is recursively merged by the weighted sum method, so as to find the optimal solution.

[0090] The construction method of the message push strategy knowledge graph can refer to the following:

[0091] Determine the scope and objectives of the knowledge graph: clarify the types of message push strategies that the knowledge graph will cover (such as content-based push, user behavior-based push, etc.). Determine the main entities and relationships of the knowledge graph, such as users, messages, push strategies, effect evaluation, etc.

[0092] Data collection: Collect data related to the message push strategy, which may include user behavior data, historical push data, user feedback, etc. The data source can be diverse, including databases, log files, API interfaces, etc.

[0093] Entity extraction: Use natural language processing (NLP) technology to extract entities from text data, such as user ID, message ID, push time, etc. Named entity recognition (NER) technology can be used to automate or assist this process.

[0094] Relationship extraction: Identify and extract relationships between entities, such as the subscription relationship between users and messages, the correspondence between messages and push strategies, etc. Relationship extraction can use rule-based methods or machine learning models, such as BERT and other models for relationship extraction.

[0095] Knowledge fusion: Integrate knowledge from different data sources to eliminate duplicate or inconsistent entities and relationships. Perform entity alignment to ensure that the same or similar entities have a unified representation in the knowledge graph.

[0096] Knowledge processing: abstract the extracted entities and relationships to form the model layer of the knowledge graph. Construct an ontology to define the types, attributes, and constraints of entities and relationships.

[0097] Quality assessment: Conduct quality assessment on the constructed knowledge graph, including accuracy, completeness, consistency, etc. The quality of the knowledge graph can be assessed through expert review, data verification, etc.

[0098] Knowledge graph update and maintenance: As data changes and business develops, the knowledge graph is regularly updated and maintained. New entities, relationships, and attributes are added, and incorrect information is corrected.

[0099] Build the application interface of the knowledge graph: Provide the query and reasoning interface of the knowledge graph for the message push system. Allow the system to formulate more accurate message push strategies based on the information in the knowledge graph.

[0100] Evaluation and optimization of knowledge graphs: Evaluate the application effect of knowledge graphs in message push strategies, such as push accuracy, user satisfaction, and other indicators. Optimize the construction method and application strategy of knowledge graphs based on the evaluation results.

[0101] See also Figure 2 The present invention provides a message push system based on grid data, comprising:

[0102] The feedback evaluation unit collects user feedback data after the message is pushed, summarizes and constructs a user feedback data set, and constructs a feedback degree Fts from the user feedback data set. If the obtained feedback degree Fts is lower than expected, the user and related data are collected;

[0103] The position prediction unit selects a corresponding grid division scheme according to the state in the message push area, divides the message push area into several sub-areas, and sends an information push instruction to the outside after the user's predicted position enters the selected target sub-area;

[0104] The recommendation unit collects data in the target sub-area and extracts push features. After matching the corresponding message push strategy for the user, it generates message content and pushes it to the user, and constrains the push frequency of the message content based on the feedback data.

[0105] The area division unit selects key factors from several push condition parameters, optimizes the message push strategy based on the key factors, re-pushes messages to users and obtains feedback data, and dynamically adjusts the grid division within the grid area based on the constructed uniformity Yts.

[0106] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0107] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0108] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0109] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0110] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0111] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0112] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0113] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A message push method based on grid data, characterized in that: include, After the message is pushed, user feedback data is collected and aggregated to construct a user feedback data set. The feedback degree Fts is constructed from the user feedback data set. If the obtained feedback degree Fts is lower than expected, the user and related data are collected; the feedback degree Fts is constructed as follows: linear normalization is performed on the click rate Np and the reading rate Zp, and the corresponding data values ​​are mapped to the interval [0,1] according to the following method: Weight coefficient: 0≤Q1≤1, 0≤Q2≤1, and Q2+Q1=1; i=1, 2,…,n, n is the total number of pushed messages, Np i is the click rate of the i-th push message, Np max and Np min is the maximum and minimum value of the click rate; Zp i is the reading rate of the i-th push message, Zp max and Zp min is the maximum and minimum value of the reading rate; According to the status in the message push area, select the corresponding grid division scheme to divide the message push area into several sub-areas. After the user's predicted position enters the selected target sub-area, send an information push instruction to the outside; Collect data in the target sub-area and extract push features. After matching the corresponding message push strategy for the user, generate message content and push it to the user, and constrain the push frequency of the message content based on the feedback data; The key factors are selected from several push condition parameters, the message push strategy is optimized based on the key factors, the message is pushed to the user again and feedback data is obtained, and the grid division in the grid area is dynamically adjusted based on the constructed uniformity Yts.

2. A message push method based on grid data according to claim 1, characterized in that: After completing the message push to the user, the feedback data of each user on the push message is collected within the feedback cycle, and the feedback data is aggregated to construct a user feedback data set; the feedback degree Fts is constructed from the user feedback data set. If the obtained feedback degree Fts is lower than the feedback threshold, the user and related data are collected, including: identity information data, user basic information, behavior data, geographic information data; the above data is aggregated to construct a push condition data set.

3. A message push method based on grid data according to claim 2, characterized in that: Determine the push area as a grid area, obtain user location data and user-related data in the grid area, and aggregate them as grid-related data; perform feature extraction on the grid-related data to obtain corresponding grid data features; A grid partitioning scheme library is constructed, and according to the correspondence between the grid data characteristics and the grid partitioning scheme, the corresponding grid partitioning scheme is matched in the grid partitioning scheme library, and the grid partitioning scheme is executed to obtain several sub-areas.

4. The message push method based on grid data according to claim 1, characterized in that: Construct an electronic map covering the grid area and display each sub-area on the electronic map. Based on the user's historical location data, a user location prediction model is constructed by convolutional neural network training. After obtaining the user's location information in real time, the location prediction model predicts the user's next location data and uses it as the predicted location. If the predicted location enters the selected target sub-area, an information push instruction is issued to the outside.

5. A message push method based on grid data according to claim 4, characterized in that: After receiving the information push instruction, the marketing information, weather and traffic data in the target sub-area are collected and used as the push environment data; the acquired push environment data, identity information data, behavior data and preference data are subjected to feature extraction to obtain the corresponding push features.

6. A message push method based on grid data according to claim 5, characterized in that: Based on the push features, the pre-built message push strategy knowledge graph matches the corresponding message push strategy for the user; the message recommendation model is obtained by recommendation algorithm training, and based on the message push strategy, the message recommendation model selects message content that meets the user's interests, sends the message content to the user, and issues push frequency adjustment instructions to the outside.

7. A message push method based on grid data according to claim 6, characterized in that: After receiving the push frequency adjustment instruction, the push frequency of the message content is constrained according to the feedback data. If it is used in the target sub-area, the user message push frequency is constrained in the target sub-area according to the constraint conditions, and the message content push is terminated after the user leaves the target sub-area; the constraint method is as follows: Where n is the number of times the message is pushed, V(i,j) is the time interval from the i-th push to the j-th push of the message content, and V avg is the mean time interval, Fts is the feedback degree when the message is pushed in the target sub-area, and Vt is the average moving speed of users in the target sub-area.

8. A message push method based on grid data according to claim 7, characterized in that: After receiving the feedback data within the analysis period, the corresponding feedback degree Fts is constructed, and multiple linear regression analysis is performed with the user's push condition parameters as independent variables and the feedback degree Fts as the dependent variable, and the corresponding linear regression equation is obtained; Obtain the regression coefficient corresponding to each independent variable from the linear regression equation, use the regression coefficient as the influence, and use the independent variables with the largest influence as the key factors; The optimization algorithm optimizes the message push strategy by taking improving the feedback degree Fts as the optimization goal; the optimized message push strategy is executed, and the message recommendation model re-acquires the corresponding message content.

9. A message push method based on grid data according to claim 8, characterized in that: After obtaining the number, position and feedback degree Fts of the target sub-regions, the uniformity Yts between the target sub-regions in the grid region is constructed. The grid division strategy in the grid region is dynamically adjusted according to the uniformity Yts, and the coarseness of the grid division is adjusted so that the coarseness is inversely proportional to the uniformity Yts. Under dimensionless conditions, the uniformity Yts is constructed according to the following method: Weight coefficient: 0≤β≤1, 0≤α≤1, and α+β=1. The weight coefficient can be obtained by referring to the hierarchical analysis method; Fts i is the feedback degree in the ith sub-region, Fts avg is the mean value of feedback degree, d ij is the distance from the i-th target sub-region to the j-th target sub-region; i = 1, 2, ..., p, p is the number of target sub-regions; Po i is the uniform mean value in the i-th target sub-region, Po avg is the average value of uniformity, It is the qualified standard value of the mean value in the uniformity.

10. A message push system based on grid data, applying the message push method according to any one of claims 1 to 9, characterized in that: include: The feedback evaluation unit collects user feedback data after the message is pushed, summarizes and constructs a user feedback data set, and constructs a feedback degree Fts from the user feedback data set. If the obtained feedback degree Fts is lower than expected, the user and related data are collected; The position prediction unit selects a corresponding grid division scheme according to the state in the message push area, divides the message push area into several sub-areas, and sends an information push instruction to the outside after the user's predicted position enters the selected target sub-area; The recommendation unit collects data in the target sub-area and extracts push features. After matching the corresponding message push strategy for the user, it generates message content and pushes it to the user, and constrains the push frequency of the message content based on the feedback data. The area division unit selects key factors from several push condition parameters, optimizes the message push strategy based on the key factors, re-pushes messages to users and obtains feedback data, and dynamically adjusts the grid division within the grid area based on the constructed uniformity Yts.

Images