Financial service product marketing method based on manus

Abstract

The invention relates to the technical field of marketing, in particular to a financial service product marketing method based on manus, which comprises the following steps: analyzing behavior data, calculating product attribute response frequency, optimizing portrait label calling sequence, identifying label confidence, screening marketing content, calculating pushing frequency and adjusting a pushing time window. Pushing parameters are adjusted in combination with user feedback, and a product marketing pushing record is obtained. According to the method, the user cross-platform account behavior is identified, the time period behavior density is combined, the timeliness and representativeness of the behavior data are improved, the sensitivity matching between the product attribute and the portrait label is adopted, the accurate adaptation between the user feature and the financial product is realized, the priority of the pushed content is adjusted by using the label confidence, and the user experience is improved. According to the method, the content correlation and the pushing effectiveness are improved, the pushing window is adjusted according to the interaction frequency, the pushing rhythm is optimized, the pushing parameters are optimized in combination with the feedback content clicking and staying characteristics, and the marketing accuracy and the user response rate are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of marketing, and particularly to a marketing method for manus financial service products. Background Art

[0002] The technical field of marketing includes activities such as assisting enterprises in product promotion, customer management, sales improvement, and user relationship maintenance through information technology and data analysis means. It involves identifying, classifying, and modeling potential customers using multi-source data such as user behavior data, transaction information, and demographic data, constructing customer portraits and formulating precise marketing strategies through model deduction, and organically integrating system processes such as data collection, behavior analysis, user label construction, and personalized content push. It covers multiple sub-directions such as customer relationship management, market segmentation, and advertising push strategy optimization. Combining the development of digital finance and mobile Internet, it transforms from traditional population coverage-based dissemination to data-driven individual precise interaction, and realizes user response behavior prediction and marketing strategy iteration.

[0003] Among them, the marketing method for manus financial service products refers to carrying out label modeling and behavior recognition on the audience characteristics, service scenarios, and transaction processes of financial products through the business processing processes embedded in the manus system platform, and constructing a product adaptation model and a marketing push logic system with the help of customer segmentation data, financial management preference information, and risk tolerance records. Matching user portraits with product characteristics through a rule screening engine, formulating a trigger-based push strategy through a task scheduling link, and forming an automated marketing execution method by combining transaction log generation, account interaction frequency tracking, and script template configuration. Making combined judgments according to the classification attributes of financial products, customer authorization behaviors, and time series preference rules in data interaction, and realizing multi-path push condition judgment and dynamic combination of reach content.

[0004] Traditional marketing technologies for financial service products lack in-depth analysis of user behavior intensity and product attribute sensitivity, resulting in actual deviations between marketing content and user needs. When constructing customer portraits, the label sensitivity and stability are not fully considered, leading to deviations between portrait labels and users' true preferences, reducing the effect of precise marketing, resulting in frequent deviation of push content from users' interests, reducing users' participation enthusiasm, lacking dynamic analysis of the correlation between user interaction frequency and push response rate, and often adopting a static fixed mode in push frequency and push timing, being unable to respond in a timely manner to changes in users' behavior rhythms, resulting in improper selection of marketing timing and reduced user response, and reducing the overall conversion effect of marketing activities. Summary of the Invention

[0005] In order to solve the technical problems existing in the prior art, an embodiment of the present invention provides a marketing method for manus financial service products. The technical solution is as follows: To achieve the above object, the present invention adopts the following technical solutions. Based on the manus financial service product marketing method, it includes the following steps: S1: Use manus to identify the account operation behavior data of users on multiple platforms, establish a user behavior data sequence according to the time sequence, identify the behavior density of users in multiple time periods, and obtain behavior response section labels; S2: Call the behavior response section labels, analyze the interaction behaviors of users with various financial service products, calculate the behavior concentration degree and response frequency of users under each product attribute, use manus to match portrait labels, compare the association strength between each portrait label and the corresponding product attribute, optimize the label call order, and generate a label sensitivity sorting sequence; S3: According to the label sensitivity sorting sequence, by analyzing the stability of customer portrait labels in user interaction behaviors, identify the label confidence level, combine with the label call order, adjust the push priorities of various marketing contents, and obtain a push content screening record; S4: Call the push content screening record, analyze the change trend of the time interval between adjacent operation behaviors of users, calculate the correlation between the user behavior interaction frequency and the push response rate, identify the optimal push frequency, combine with the interaction frequency data of users in multiple time periods, identify the matching degree of multiple push time periods, and generate a push time window sequence.

[0006] As a further solution of the present invention, the behavior response section labels include behavior intensive sections, behavior sparse sections, and behavior switching boundaries. The label sensitivity sorting sequence is specifically the sorting label name, sorting level identifier, and call priority number. The push content screening record includes candidate content identifier, trigger strategy number, and push level assignment. The push time window sequence specifically refers to the candidate time period index, recommended trigger time point, and window priority number.

[0007] As a further solution of the present invention, the specific steps for obtaining the behavior response section labels are as follows: S101: Use manus to identify the account operation behavior data of users on multiple platforms, including clicks, accesses, browsing, and transactions, and establish a user behavior data sequence according to the operation time and behavior type; S102: Based on the user behavior sequence values, analyze the distribution of click, access, browsing, and transaction behaviors in each time period, calculate the occurrence times and distribution ratios of each behavior in multiple time periods, and obtain behavior distribution information; S103: According to the behavior distribution information, identify the behavior density of users in multiple time periods, detect the distribution density of each behavior in multiple time periods, and establish behavior response section labels.

[0008] As a further solution of the present invention, the steps for obtaining the tag-sensitive sorting sequence are specifically as follows: S201: Invoke the behavior response section tags, analyze the user behavior distribution corresponding to each product attribute according to the click, access, browsing, and transaction behavior data of the user on various financial service products in each time period, identify the concentration degree and response frequency of the behavior under each product attribute, and generate a product attribute behavior concentration coefficient; S202: According to the product attribute behavior concentration coefficient, identify the behavior pattern characteristics of the target account, use manus to match the portrait tags, construct a set of customer portrait tags for the account, and analyze the association strength between each portrait tag and the corresponding product attribute to obtain a portrait attribute correlation coefficient; S203: According to the portrait attribute correlation coefficient, compare the difference in the association strength between each customer portrait tag and the product attribute, adjust the priority order of the customer portrait tags, optimize the calling order of the tags, and establish a tag-sensitive sorting sequence.

[0009] As a further solution of the present invention, the specific formula for analyzing the association strength between each portrait tag and the corresponding product attribute is: ; Calculate the portrait attribute correlation coefficient to obtain the portrait attribute correlation coefficient; Wherein, represents the normalized value of the association strength between the j-th portrait tag and the corresponding financial product attribute in the i-th account, represents the normalized value of the interaction frequency of the j-th tag matching the financial product attribute in the i-th account during the time period t, represents the normalized value of the access frequency of the non-matching product attribute under the j-th tag in the i-th account during the time period t, represents the interaction weight coefficient of the financial product attribute during the time period t, represents the weight coefficient of the tag feature call during the time period t, represents the normalized value of the number of portrait tags in the i-th account, i represents the index number of the target account, j represents the index number of the customer portrait tag, T represents the total number of time periods divided within the portrait tag association period, and t represents the t-th specific time period in the portrait tag association period.

[0010] As a further solution of the present invention, the steps for obtaining the push content screening record are specifically as follows: S301: Invoke the tag-sensitive sorting sequence, analyze the stability of the customer portrait tags in the user interaction behavior according to the response frequency and the change range, and generate a tag behavior stability coefficient; S302: Calculate the accuracy of each tag's current expression of the user's interests and preferences based on the tag behavior stability coefficient, evaluate the confidence of each tag, and obtain tag confidence information; S303: Based on the tag confidence information and in combination with the tag calling order, the push priority of each type of marketing content is screened and adjusted, and a push content screening record is obtained.

[0011] As a further solution of the present invention, the specific formula for calculating the accuracy of each tag's current corresponding user interests and preferences is: ; Calculate label expression accuracy parameters; Among them, s represents the current user, k represents the label number, and n represents the time window number. represents the behavioral stability coefficient of the kth tag to the current user s, represents the number of forward operations of the k-th label in the n-th time window, Indicates the number of negative operations of the k-th label in the n-th time window, represents the normalized value of the standard deviation of the operation frequency of the k-th tag in the n-th time window, represents the normalized value of the standard deviation of the operation duration of the kth label in the nth time window, m represents the total number of time windows, represents the expression accuracy parameter of the k-th tag for user s.

[0012] As a further solution of the present invention, the step of acquiring the push time window sequence is specifically as follows: S401: Calling the push content screening record, analyzing the change direction and fluctuation trend of the operation intervals according to the time intervals between adjacent user operation behaviors, and generating change trend information; S402: Calculate the correlation between the interaction frequency of each user behavior and the number of push responses based on the change trend information, identify the optimal push frequency for the target user, and obtain a push frequency coefficient; S403: In combination with the push frequency coefficient, by analyzing the user's interaction frequency in multiple time periods, comparing the adaptability of each push time period to the user's active behavior, adjusting the time interval of each push task, and generating a push time window sequence.

[0013] As a further embodiment of the present invention, the method further comprises: S5: Based on the push time window sequence, collect and analyze the user's click and stay behavior characteristics for each marketing content category, calculate the actual user attraction and response trend of each content type, adjust the push parameters of the marketing content, and obtain product marketing push records; The product marketing push record includes a content type number, a push cycle configuration, and a response behavior summary.

[0014] As a further solution of the present invention, the steps for obtaining the product marketing push record are specifically as follows: S501: According to the push time window sequence, collect the click quantity and page stay duration of users for different marketing content categories within each time window, analyze the user behavior data of each content category, and generate behavior feedback data; S502: Based on the behavior feedback data, calculate the actual attraction degree of each type of marketing content to users, combine the change trends of the click quantity and stay duration, analyze the response trend of users to the content, and obtain content response degree information; S503: According to the content response degree information, adjust the push parameters of the financial service product marketing content, including the push frequency, push time point, and push content category, to obtain the product marketing push record.

[0015] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include: By identifying the cross-platform account behavior of users, combining the behavior density in time periods, improving the timeliness and representativeness of behavior data, using the sensitive matching of product attributes and portrait tags to achieve the precise adaptation between user characteristics and financial products, adjusting the push content priority using tag confidence, improving the content relevance and push effectiveness, adjusting the push window according to the interaction frequency, optimizing the push rhythm, combining the feedback content click and stay characteristics, optimizing the push parameters, and improving the marketing accuracy and user response rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a schematic diagram of the working process of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The following will describe the technical solutions in the present invention with reference to the drawings.

[0019] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to give examples, illustrations or explanations. Any embodiment or design described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or designs. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0020] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0021] In the embodiments of the present invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0022] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0023] Please refer to Figure 1 , the present invention provides a technical solution based on the manus financial service product marketing method, including the following steps: S1: Use manus to identify the account operation behavior data of users on multiple platforms, establish a user behavior data sequence according to the time sequence, identify the behavior density of users in multiple time periods, and obtain behavior response section labels; S2: Call the behavior response section labels, analyze the interaction behaviors of users with various financial service products, calculate the degree of behavior concentration and response frequency of users under each product attribute, use manus to match portrait labels, compare the association strength between each portrait label and the corresponding product attribute, optimize the label call order, and generate a label sensitivity sorting sequence; S3: According to the label sensitivity sorting sequence, by analyzing the stability of customer portrait labels in user interaction behaviors, identify the label confidence level, combine the call order of the labels, adjust the push priorities of various marketing contents, and obtain push content screening records; S4: Call the push content screening records, analyze the change trend of the time interval between adjacent operation behaviors of users, calculate the correlation between the user behavior interaction frequency and the push response rate, identify the optimal push frequency, combine the interaction frequency data of users in multiple time periods, identify the matching degree of multiple push time periods, and generate a push time window sequence; S5: According to the push time window sequence, collect and analyze the click and stay behavior characteristics of users for each marketing content category, calculate the actual attraction degree and response trend of each content type to users, adjust the push parameters of the marketing content, and obtain the product marketing push record.

[0024] The behavior response section tags include the behavior dense section, the behavior sparse section, and the behavior switching boundary. The tag sensitive sorting sequence is specifically the sorting tag name, the sorting level identifier, and the call priority number. The push content screening record includes the candidate content identifier, the trigger policy number, and the push level allocation. The push time window sequence specifically refers to the candidate time period index, the recommended trigger time point, and the window priority number. The product marketing push record includes the content type number, the push cycle configuration, and the response behavior summary.

[0025] The specific steps for obtaining the behavior response section tags are as follows: S101: Use manus to identify the user's account operation behavior data on multiple platforms, including clicks, accesses, browsing, and transactions. According to the operation time and behavior type, establish a user behavior data sequence. Use manus to identify the user's account operation behavior data on multiple platforms, including clicks, accesses, browsing, and transactions. Collect the operation logs of user A on three financial platforms. Each log records the operation type and the specific operation time. For example, user A completed a product browsing at 9:32 on July 5, 2024, a click on a financial product at 9:34, and a transaction operation at 9:45. By arranging these logs in the order of the specific time of operation, a complete user behavior data sequence is constructed. When traversing each log, extract the three elements of the platform, operation type, and operation time. The specific operation times of click behavior, access behavior, browsing behavior, and transaction behavior are respectively classified into independent sequences. Correspondingly, set the operation type numbers to 1, 2, 3, and 4. Assume that a total of 10 operation logs of user A are collected between 9:00 and 10:00 on July 5, and they are arranged in sequence as 9:02 click, 9:12 access, 9:18 browsing, 9:22 transaction, 9:27 access, 9:32 click, 9:34 browsing, 9:40 transaction, 9:43 click, 9:45 transaction. After numbering this sequence, a behavior data sequence is formed. When analyzing the operation type to which each log belongs, convert the operation time into minutes for subsequent operations. The specific time of each operation is used as the index of the data sequence and arranged in ascending order of minutes to form a time-ordered data array. Using the above numbering rules, establish a behavior type array. The operation time array , According to the above data combination, each piece of data contains three items: operating platform, operating type, and operating time. For example, the first piece of data records that user A's click behavior on platform A occurred at the 2nd minute, and the second record is that the access behavior on platform B occurred at the 12th minute, and so on. If the total number of operations of the same user on different platforms in a day is N, and all operations are numbered and arranged in ascending order of time, a complete user behavior data sequence can be obtained.

[0026] S102: Based on the user behavior sequence values, analyze the distribution of click, access, browse, and transaction behaviors in each time period, calculate the occurrence times and distribution ratios of each behavior in multiple time periods, and obtain the behavior distribution information; Based on the user behavior sequence values, analyze the distribution of click, access, browse, and transaction behaviors in each time period. Assume that user A's operation time period in a day is divided into three time periods: 8:00 - 9:00, 9:00 - 10:00, and 10:00 - 11:00. Count the number of click, access, browse, and transaction operations that occur in each time period respectively. For example, within 9:00 - 10:00, there are 3 clicks, 2 accesses, 2 browses, and 3 transactions. For each behavior, calculate the number of times it appears in this time period , and compare the total number of occurrences of this behavior in all operation logs , the distribution ratio of the behavior in this time period is , for example, user A's click behavior within the time period of 9:00 - 10:00 , the total number of click behaviors , obtain the distribution ratio of the click behavior in this time period , perform the same calculation on access, browse, and transaction behaviors in turn. Assume that the distribution ratios of each behavior are respectively , , , form a distribution ratio table of different behaviors in different time periods. The table is arranged in a two-dimensional manner of behavior and time period. The time period distribution ratio matrix of each behavior , analyze the extreme values of the distribution ratio matrix, judge which behavior is the most concentrated in which time period, and sort by the ratio size. Combining the actual operation data, if the distribution ratio of the transaction behavior in 10:00 - 11:00 is higher than other time periods, it indicates that the transaction operations are intensive in this time period, and thus the behavior distribution information is obtained.

[0027] S103: According to the behavior distribution information, identify the behavior density of the user in multiple time periods, detect the distribution density of each behavior in multiple time periods, and establish behavior response section labels; Based on the behavior distribution information, identify the behavior density of the user in multiple time periods, detect the distribution density of each behavior in multiple time periods, and define the benchmark ratio for density determination as 0.7, that is, when it is determined that the behavior is of high density in this time period. Combining the foregoing example, if the click ratio from 9:00 to 10:00 is 0.5, which is lower than the density determination benchmark, and the transaction ratio from 10:00 to 11:00 is 0.8, which is higher than the determination benchmark, the transaction behavior from 10:00 to 11:00 is marked as a high-density behavior section. The distribution ratios of each behavior in all time periods are compared item by item with the benchmark ratio to form a density determination result table. Each row in the table represents a time period, the columns represent behavior types, and the cells are filled with high density or normal. Combining the determination results, select the high-density behavior section as the behavior response section label. For example, the output is transaction from 9:00 to 10:00, transaction from 10:00 to 11:00, click from 8:00 to 9:00, and then establish the behavior response section label.

[0028] The steps for obtaining the label sensitive sorting sequence are specifically as follows: S201: Call the behavior response section label. According to the click, access, browsing, and transaction behavior data of the user on various financial service products in each time period, analyze the user behavior distribution corresponding to each product attribute, identify the degree of behavior concentration and response frequency under each product attribute, and generate the product attribute behavior concentration coefficient; Call the behavior response section label. According to the user's click, access, browsing, and transaction behavior data on various financial service products in each time section, first obtain all the behavior logs related to financial product attributes of the user in different time sections, and identify the behavior type, operation time, and corresponding product attributes item by item. For example, the occurrence times of the click, access, browsing, and transaction behaviors of product A in the three sections of 8:00 - 9:00, 9:00 - 10:00, and 10:00 - 11:00 are 3, 5, 2, 4, 6, 1, 2, 4, 5, 2 respectively. Set the section number as, the product attribute number as, and the behavior number in each section as. For each product attribute, count the total number of each behavior type under this attribute in each section. For example, the occurrence times of the click behavior of product A in the 9:00 - 10:00 section. Subsequently, perform normalization processing according to the total occurrence times of each behavior type in the whole section, that is, divide the times in each section by the total times of this behavior under the same product attribute in all sections. For example, if the total times of the click behavior of product A in the three sections is, then the normalization value in the 9:00 - 10:00 section is. Calculate the normalization distribution coefficients of all behavior types and all sections in turn, construct a three-dimensional distribution matrix, and then judge whether the distribution coefficient of each behavior in each section is greater than the set behavior concentration benchmark coefficient. This benchmark coefficient is set to 0.4. If the normalization coefficient is greater than 0.4, it is considered that the behavior distribution of the product attribute in this section is concentrated. For example, the click behavior of product A in 9:00 - 10:00 is concentrated. Judge the distribution concentration of each behavior type and section under all product attributes in turn, analyze all the results, summarize the normalization distribution coefficients of each section, each behavior, and each product attribute into the behavior distribution matrix, and finally organize and output the product attribute behavior concentration coefficient.

[0029] S202: Identify the behavior pattern characteristics of the target account according to the product attribute behavior concentration coefficient, use manus to match the portrait label, construct the set of customer portrait labels of the account, and analyze the association strength between each portrait label and the corresponding product attribute to obtain the portrait attribute correlation coefficient; The specific formula for analyzing the association strength between each portrait label and the corresponding product attribute is: ; Calculate the portrait attribute correlation coefficient to obtain the portrait attribute correlation coefficient; Among them, represents the normalization value of the association strength between the j-th portrait label and the corresponding financial product attribute in the i-th account, represents the normalization value of the interaction frequency of the j-th label matching the financial product attribute in the i-th account in the time period t, represents the normalization value of the access frequency of the non-matching product attribute under the j-th label in the i-th account in the time period t, represents the interaction weight coefficient of the financial product attribute in the time period t, Represents the weight coefficient of the tag feature call for the time period t, Represents the normalized value of the number of portrait tags of the i-th account. Here, i represents the index number of the target account, j represents the index number of the customer portrait tag, T represents the total number of time periods divided within the portrait tag association period, and t represents the t-th specific time period within the portrait tag association period.

[0030] Formula: ; Detailed explanation of the formula and the derivation process of the formula calculation: The formula is used to calculate the normalized association strength between the j-th portrait tag in the i-th account and the associated financial product attributes. The obtained result is used to measure the matching ability of the current tag with the target product features and is used to drive the generation of subsequent push strategies.

[0031] Meaning of parameters and set values: Is the normalized value of the interaction frequency of the product attributes associated with the j-th tag by the i-th account in the t-th time period; Is the normalized value of the access frequency of the product attributes not associated with this tag by the i-th account in the t-th time period; Is the interaction weight coefficient of the financial product attributes in the time period t. This coefficient varies with the complexity of the product type. It is set to 0.8 for complex products, 0.6 for ordinary products, and 0.4 for simple products; Is the tag call weight coefficient in the time period t, set to 0.75; Is the normalized value of the number of portrait tags of the i-th account within the portrait modeling period, set to 0.6; T is the number of time periods in the association period, divided into 3 time periods. When t = 1, , , , When t = 2, , , , When t = 3, , , , ; Substitute the parameters into the formula for calculation: ; ; ; ; =(0.6 + 0.3)+(0.4 + 0.5)+(0.5 + 0.4)=0.9 + 0.9 + 0.9 = 2.7; ; 2.7 + 0.7746 = 3.4746; ; The result 0.0777 indicates a low association strength between the i-th account and the target financial product attribute for label j. The closer this result value is to 1, the more matching the label is to the current product attribute, and the closer it is to 0, the weaker the matching. Therefore, this label should be assigned a lower content matching priority in the push strategy. This result is used as the portrait attribute correlation coefficient to participate in label sorting and priority assignment, and ultimately used to generate a label-sensitive sorting sequence.

[0032] S203: According to the portrait attribute correlation coefficient, compare the association strength differences between each customer portrait label and the product attribute, adjust the priority order of the customer portrait labels, optimize the label call order, and establish a label-sensitive sorting sequence; According to the portrait attribute correlation coefficient, compare the association strength differences between each customer portrait label and the product attribute, adjust the priority order of the customer portrait labels, optimize the label call order, and establish a label-sensitive sorting sequence. First, based on the above correlation coefficient matrix, sort all correlation coefficients for each label and each product attribute one by one. For the same account, compare the correlation coefficients of different labels with the same product attribute in descending order of association strength. Labels with a correlation coefficient higher than 0.5 are priority labels, those between 0.2 - 0.5 are regular labels, and those below 0.2 are alternative labels. Divide the priority levels in this way, arrange the priority labels in order to form a label call queue, and preferentially call the labels with high association with the product attribute to trigger marketing content. For example, for account C, the label "High-frequency financial management" has a correlation coefficient of 0.7 with product C, the label "Medium preference" is 0.4, and the label "Risk-sensitive" is 0.15. Then the priority order is "High-frequency financial management" > "Medium preference" > "Risk-sensitive". The entire priority queue forms a label-sensitive sorting sequence, forming the final sorting structure.

[0033] The specific steps for obtaining the push content screening record are as follows: S301: Call the label-sensitive sorting sequence, analyze the stability of the customer portrait labels in the user interaction behavior according to the response frequency and the range of changes, and generate a label behavior stability coefficient; Call the tag-sensitive sorting sequence, and analyze the stability of customer portrait tags in user interaction behaviors according to the response frequency and the range of variation. First, retrieve the tag-sensitive sorting sequence obtained in the previous process. For each tag, obtain the actual number of responses of the tag in different interaction behaviors, and count the frequency of the tag being triggered within a daily, weekly, or monthly interval. For example, the "high-risk financial management preference" tag was triggered 15 times, 20 times, 18 times, 16 times, and 17 times within 30 days, forming a frequency array. Calculate the maximum and minimum values in the array to obtain the range of variation. The formula for calculating the coefficient of variation is , where is the maximum number of responses, is the minimum number of responses. In this example , , and we get . Then, combined with the average response frequency of the tag, such as the average value being 17.2, set the criterion for determining the stability of tag behavior as stable when the coefficient of variation is less than 0.3 and the average value is greater than 10. Count all tags in turn to form a tag behavior stability coefficient matrix, and the matrix elements are the behavior stability values of each tag.

[0034] S302: According to the tag behavior stability coefficient, calculate the expression accuracy of each tag corresponding to the current user's interests and preferences, evaluate the confidence of each tag, and obtain the tag confidence information; The specific formula for calculating the expression accuracy of each tag corresponding to the current user's interests and preferences is: ; Calculate the tag expression accuracy parameter; where s represents the current user, k represents the tag number, n represents the time window number, represents the behavior stability coefficient of the k-th tag for the current user s, represents the number of positive operations of the k-th tag in the n-th time window, represents the number of negative operations of the k-th tag in the n-th time window, represents the normalized value of the standard deviation of the operation frequency of the k-th tag in the n-th time window, represents the normalized value of the standard deviation of the operation duration of the k-th tag in the n-th time window, m represents the total number of time windows, represents the expression accuracy parameter of the k-th tag for the user s.

[0035] Formula: ; Detailed explanation of the formula and the derivation process of the formula calculation: The formula is used to calculate the accuracy of the tag in expressing the user's interests and preferences, and the result is used to quantify the matching degree weight of the tag to the user in the recommendation system.

[0036] Parameter meaning and setting values: : The behavior stability coefficient of label k for user s, which is inversely quantified by the standard deviation of the proportion of positive operations of the label within multiple time windows. It is set that when the fluctuation range of the positive operation ratio of a certain label within 10 consecutive monitoring cycles is less than 10%, the stability score result is 0.85; : The number of positive operations of label k in the nth time window, which comes from behaviors such as clicks, favorites, and likes. It is set that = 30, = 25, = 28; : The number of negative operations of label k in the nth time window, which comes from behaviors such as skips and dislikes. It is set that = 10, = 15, = 12; : The normalized value of the standard deviation of the operation frequency of label k in the nth time window, which is normalized by the standard deviation of the daily operation times. The normalized upper limit is 24 times. It is set that = 0.21, = 0.17, = 0.19.

[0037] : The normalized value of the standard deviation of the operation duration of label k in the nth time window, which is normalized by the standard deviation of the single operation duration. The normalized upper limit is 60 seconds. It is set that = 0.25, = 0.22, = 0.23.

[0038] m = 3: The total number of time windows; Substitute the parameters into the formula for calculation: ; ; [[ID=%56]]

[0039] ; This result indicates that label k has a high expression accuracy in the historical behavior of user s. The value 1.871 indicates that the label has a strong signal strength when expressing user interests. After exceeding the recommended system setting threshold of 1.2, it will be marked as a high-confidence label and incorporated into the main sorting weight matrix to optimize the personalized recommendation effect. The parameters will be input as weight factors into the subsequent label priority evaluation model to improve the decision-making accuracy.

[0040] S303: Based on the tag confidence information and combined with the call order of the tags, screen and adjust the push priority of each type of marketing content, and obtain the screening record of the push content; Based on the tag confidence information and combined with the call order of the tags, screen and adjust the push priority of each type of marketing content, and obtain the screening record of the push content. Sort the tag confidence information of the previous process into three levels: high, medium, and low. Set the marketing content corresponding to the high-confidence tags as the priority push category, and sequentially screen the content corresponding to the medium-confidence and low-confidence tags as supplementary and backup push content. For each round of push tasks, preferentially select high-confidence tags according to the tag call order, and arrange the push plan according to the priority, forming the corresponding relationship between the marketing content, tag confidence, and call order. Record the information such as tags, content, and time of each round of push tasks, and organize it into a data table of the screening record of the push content.

[0041] The specific steps for obtaining the push time window sequence are as follows: S401: Call the screening record of the push content, analyze the change direction and fluctuation trend of the operation interval according to the time interval between adjacent operation behaviors of the user, and generate the change trend information; Call the screening record of the push content, analyze the change direction and fluctuation trend of the operation interval according to the time interval between adjacent operation behaviors of the user. First, extract the operation time of each push and the user's actual response in the screening record of the push content, calculate the time interval between two adjacent operations, and construct a time interval sequence. For example, if the response times of the user 5 times are 10:01, 10:10, 10:18, 10:21, and 10:35 respectively, the intervals are 9 minutes, 8 minutes, 3 minutes, and 14 minutes in sequence. Analyze the change trend according to the sequence, and use the first-order difference analysis method. Subtract the previous interval from the current interval to get the difference values, which are -1, -5, and 11 respectively. A positive value indicates an increase in the interval, and a negative value indicates a shortening of the interval. Count the number of positive and negative values in the sequence to judge whether the overall interval tends to shorten or extend. If the positive and negative values are distributed alternately more, it means that the fluctuation is large, and the volatility can be further calculated. The volatility formula is , where B is the volatility, n is the number of responses, is the time interval of the i-th operation. In this example, the volatility is , and finally output the change trend information.

[0042] S402: According to the change trend information, calculate the correlation between the interaction frequency of each user behavior and the number of push responses, identify the optimal push frequency of the target user, and obtain the push frequency coefficient; According to the change trend information, calculate the correlation between the interaction frequency of each user behavior and the number of push responses, identify the optimal push frequency for the target user, and obtain the push frequency coefficient. First, extract the interaction frequency of the same behavior category from the push content screening records and the change trend information. And the corresponding number of push responses , pair all behaviors with the number of push responses. For example, in the browsing behavior, 10 pushes are made and 7 responses are received; in the click behavior, 8 pushes are made and 5 responses are received; in the transaction behavior, 4 pushes are made and 2 responses are received. Use the Pearson correlation coefficient to calculate the correlation between the interaction frequency and the number of responses. The correlation coefficient formula is , where is the correlation coefficient between the interaction frequency and the number of responses, is the interaction frequency of the i-th behavior, is the number of responses of the i-th behavior, is the mean value of the interaction frequency, is the mean value of the responses, and n is the number of behavior types. If is greater than 0.7, it is considered a strong correlation. Analyze the number of responses at each frequency, and select the frequency with the highest response rate as the optimal push frequency, and output the push frequency coefficient. For example, when the push frequency of the browsing behavior is once every 10 minutes, the response rate reaches 70%, and this frequency is the optimal frequency, and the push frequency coefficient is 0.7.

[0043] S403: Combine the push frequency coefficient, analyze the interaction frequency of the user in multiple time periods, compare the suitability of each push time period with the user's active behavior, adjust the time interval of each push task, and generate a push time window sequence; Combine the push frequency coefficient, analyze the interaction frequency of the user in multiple time periods, compare the suitability of each push time period with the user's active behavior, and adjust the time interval of each push task. First, compare the push frequency coefficient with the actual interaction frequency distribution of the user in each time period, and extract the time period close to the optimal push frequency coefficient. For example, for a certain user, the interaction frequencies from 12:00 to 13:00 and from 18:00 to 19:00 are once every 10 minutes and once every 15 minutes respectively. The time period with an interaction frequency of once every 10 minutes is given priority as the push window. Compare the interaction frequency of each push time period with the optimal push frequency coefficient in turn, sort them in descending order according to the degree of closeness, arrange the time interval of the push task, and finally output the push time window sequence.

[0044] The specific steps for obtaining the product marketing push records are as follows: S501: According to the push time window sequence, collect the number of clicks and the page stay duration of the user for different marketing content categories in each time window, analyze the user behavior data of each content category, and generate behavior feedback data; According to the push time window sequence, collect the number of clicks and page stay duration of users on different marketing content categories within each time window, and analyze the user behavior data of each content category. First, for each push time window, classify and count the marketing content categories that users browse, click, and interact with during this period. Set the content category number as j, and collect the number of clicks of each content category within the same time window. and the page stay duration , for example, in the 12:00-13:00 time window, users click on category A content 4 times, and the total stay duration is 300 seconds; click on category B content 2 times, and the stay duration is 180 seconds. Aggregate the total click volume and stay time of each content category in each window, perform multi-window statistics on the data within the entire cycle, form a two-dimensional behavior feedback table of content category - time window, and output the behavior feedback data.

[0045] S502: Based on the behavior feedback data, calculate the actual attraction degree of each type of marketing content to users. Combine the change trends of the number of clicks and stay duration, analyze the response trend of users to the content, and obtain the content response degree information; Based on the behavior feedback data, calculate the actual attraction degree of each type of marketing content to users. Combine the change trends of the number of clicks and stay duration, analyze the response trend of users to the content. First, for each content category, calculate the average number of clicks within the entire cycle and the average stay duration , for example, for category A content, the number of clicks in 4 windows is 4, 5, 6, and 3 respectively, and the total average is , and the stay durations are 300, 320, 340, and 310 respectively, and the total average seconds. Use the comprehensive score of content response degree , where is the response degree score of the j-th type of content, , are weights (which can be taken as 0.5), , are the maximum values of the average number of clicks and average stay duration of all categories respectively. Substitute the data to obtain the response degree score of category A content, sort the response degrees of all categories, analyze the change trends of clicks and stay duration, and output the content response degree information. Among them, is the average number of clicks of content category j, is the average stay duration of content category j, is the maximum average number of clicks, is the maximum average stay duration.

[0046] S503: According to the content response degree information, adjust the push parameters of the financial service product marketing content, including push frequency, push time point, and push content category, and obtain the product marketing push record; Adjust the push parameters of the financial service product marketing content according to the content response information, including the push frequency, push time point, and push content category. Obtain the product marketing push records. According to the response score of each content category, set the content with the highest response score as the priority push object. Combine the content preferences within different time windows, tilt the push frequency towards the content categories with high responses, align the push time point with the response peak window, allocate different push frequencies and push time points to different content categories, and adjust the content push order from high to low in priority. Finally, record the parameters such as content category, time window, and push frequency in each round of push tasks to form the product marketing push records.

[0047] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any other combination. When implemented using 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 processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0048] It should be understood that the term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood with reference to the context before and after.

[0049] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single item(s) or plural item(s). For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or plural.

[0050] It should be understood that in various embodiments of the present invention, the sequence numbers of the above - mentioned processes do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0051] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professionals can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present invention.

[0052] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0053] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0054] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0055] In addition, in each embodiment of the present invention, each functional unit can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0056] If the described function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. 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 discs.

[0057] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A marketing method for Manus financial service products, characterized in that The method includes: S1: Use manus to identify the user's account operation behavior data on multiple platforms, establish a user behavior data sequence according to the time sequence, identify the behavior density of the user in multiple time periods, and obtain the behavior response section label; S2: Call the behavior response section label, analyze the user's interaction behavior with various financial service products, calculate the behavior concentration degree and response frequency of the user under each product attribute, use manus to match the portrait label, compare the association strength between each portrait label and the corresponding product attribute, optimize the label call order, and generate a label sensitivity sorting sequence; S3: According to the label sensitivity sorting sequence, by analyzing the stability of the customer portrait label in the user interaction behavior, identify the label confidence level, combine the call order of the label, adjust the push priority of various marketing contents, and obtain the push content screening record; S4: Call the push content screening record, analyze the change trend of the time interval between adjacent user operation behaviors, calculate the correlation between the user behavior interaction frequency and the push response rate, identify the optimal push frequency, and combine the interaction frequency data of the user in multiple time periods to identify the matching degree of multiple push time periods, and generate a push time window sequence.

2. The marketing method of the manus financial service product according to claim 1, wherein The behavior response section label includes a behavior dense section, a behavior sparse section, and a behavior switching boundary. The label sensitivity sorting sequence is specifically the sorting label name, the sorting level identifier, and the call priority number. The push content screening record includes the candidate content identifier, the trigger policy number, and the push level assignment. The push time window sequence specifically refers to the candidate time period index, the recommended trigger time point, and the window priority number.

3. The marketing method of the manus financial service product according to claim 1, wherein The specific steps for obtaining the behavior response section label are as follows: S101: Use manus to identify the user's account operation behavior data on multiple platforms, including clicks, accesses, browsing, and transactions, and establish a user behavior data sequence according to the operation time and behavior type; S102: Based on the user behavior sequence value, analyze the distribution of click, access, browsing, and transaction behaviors in each time period, calculate the occurrence times and distribution ratios of each behavior in multiple time periods, and obtain the behavior distribution information; S103: According to the behavior distribution information, identify the behavior density of the user in multiple time periods, detect the distribution density of each behavior in multiple time periods, and establish a behavior response section label.

4. The marketing method of the manus financial service product according to claim 3, characterized in that, The specific steps for obtaining the label sensitivity sorting sequence are as follows: S201: Call the behavior response section label, and according to the user's click, access, browsing, and transaction behavior data on various financial service products in each time section, analyze the user behavior distribution corresponding to each product attribute, identify the behavior concentration degree and response frequency under each product attribute, and generate a product attribute behavior concentration coefficient; S202: According to the product attribute behavior concentration coefficient, identify the behavior pattern characteristics of the target account, use manus to match the portrait label, construct a set of customer portrait labels for the account, and analyze the association strength between each portrait label and the corresponding product attribute to obtain the portrait attribute correlation coefficient; S203: According to the portrait attribute correlation coefficient, compare the difference in the correlation strength between each customer portrait label and product attributes, adjust the priority order of customer portrait labels, optimize the call order of labels, and establish a label-sensitive sorting sequence.

5. The marketing method of the manus financial service product according to claim 4, characterized in that The specific formula for analyzing the correlation strength between each portrait label and the corresponding product attribute is: ； Calculate the portrait attribute correlation coefficient and obtain the portrait attribute correlation coefficient; Among them, represents the normalized value of the association strength between the j-th portrait label and the corresponding financial product attribute in the i-th account, represents the normalized value of the interaction frequency of the i-th account with the financial product attributes matching the j-th label during the time period t, represents the normalized value of the access frequency of the i-th account to the non-matching product attributes under the j-th label during the time period t, represents the interaction weight coefficient of the financial product attributes during the time period t, represents the weight coefficient of label feature invocation during the time period t, represents the normalized value of the number of portrait labels of the i-th account. Here, i represents the index number of the target account, j represents the index number of the customer portrait label, T represents the total number of time periods divided within the portrait label association period, and t represents the t-th specific time period in the portrait label association period.

6. The marketing method of the manus financial service product according to claim 4, wherein The specific steps for obtaining the push content screening record are as follows: S301: Call the label-sensitive sorting sequence, analyze the stability of customer portrait labels in user interaction behaviors according to the response frequency and change range, and generate a label behavior stability coefficient; S302: According to the label behavior stability coefficient, calculate the expression accuracy of each label for the current user interest and preference, evaluate the confidence of each label, and obtain label confidence information; S303: Based on the label confidence information, combined with the call order of labels, screen and adjust the push priority of each type of marketing content, and obtain a push content screening record.

7. The marketing method of manus financial service products according to claim 6, characterized in that The specific formula for calculating the expression accuracy of each label for the current user interest and preference is: ； Calculate the label expression accuracy parameter; Among them, s represents the current user, k represents the tag number, and n represents the time window number. represents the behavior stability coefficient of the k-th tag for the current user s. represents the number of positive operations of the k-th tag in the n-th time window. represents the number of negative operations of the k-th tag in the n-th time window. represents the normalized value of the standard deviation of the operation frequency of the k-th tag in the n-th time window. represents the normalized value of the standard deviation of the operation duration of the k-th tag in the n-th time window, and m represents the total number of time windows. represents the expression accuracy parameter of the k-th tag for the user s.

8. The marketing method of the manus financial service product according to claim 6, characterized in that The specific steps for obtaining the push time window sequence are as follows: S401: Call the push content screening record, analyze the change direction and fluctuation trend of the operation interval according to the time interval between adjacent user operation behaviors, and generate change trend information; S402: According to the change trend information, calculate the correlation between the interaction frequency of each user behavior and the push response quantity, identify the optimal push frequency of the target user, and obtain a push frequency coefficient; S403: Combine the push frequency coefficient, analyze the interaction frequency of the user in multiple time periods, compare the adaptation degree of each push time period and the user's active behavior, adjust the time interval of each push task, and generate a push time window sequence.

9. The marketing method of the manus financial service product according to claim 1, wherein The method further includes: S5: According to the push time window sequence, collect and analyze the click and stay behavior characteristics of users for each type of marketing content category, calculate the actual attraction degree and response trend of each content type to users, adjust the push parameters of marketing content, and obtain a product marketing push record; The product marketing push record includes a content type number, a push cycle configuration, and a response behavior summary.

10. The marketing method of the manus financial service product according to claim 9, wherein The specific steps for obtaining the product marketing push record are as follows: S501: According to the push time window sequence, collect the click quantity and page stay duration of users for different marketing content categories in each time window, analyze the user behavior data of each content category, and generate behavior feedback data; S502: Based on the behavior feedback data, calculate the actual attraction degree of each type of marketing content to users, combine the change trend of the click quantity and stay duration, analyze the response trend of users to the content, and obtain content response degree information; S503: According to the content response degree information, adjust the push parameters of financial service product marketing content, including push frequency, push time point, and push content category, and obtain a product marketing push record.

Images