Intelligent marketing data analysis system and method

Abstract

The invention relates to the technical field of digital marketing, in particular to an intelligent marketing data analysis system and method. According to the invention, through deep analysis of the user interaction data, the user behavior offset is accurately identified, so that the marketing strategy can be dynamically adjusted, and the putting accuracy is improved. By means of the change trend of the residence time sequence, the relative change rate is calculated in combination with the access preference, user interest hotspot distribution information is effectively extracted, and the pertinence of content recommendation is improved. Based on access path stability analysis of interest hotspots, a stable mode of user behaviors is extracted, and an advertisement putting interval is optimized, so that user loss caused by frequent putting is reduced. And user groups with fluency decline or obvious jumping behaviors are screened by utilizing change calculation of interaction fluency, so that the advertisement pushing rhythm is reasonably adjusted, and the advertisement experience is improved. The user access data is synthesized to analyze the conversion change of the user behavior after the advertisement is displayed, and the influence of different advertisement intervals on the user behavior is accurately evaluated.

Description

Technical Field

[0001] The present invention relates to the field of digital marketing technology, and in particular to an intelligent marketing data analysis system and method. Background Art

[0002] The field of digital marketing technology includes data analysis, user behavior research, advertising optimization, personalized recommendations, etc. Its core marketing content is to optimize marketing strategies in a data-driven way to improve user conversion rates and brand influence. In digital marketing, data collection, analysis and application are key links, usually involving specific technologies such as user portrait construction, market trend forecasting, and precise advertising push. At present, digital marketing widely relies on a variety of data sources, such as social media interaction data, user search records, e-commerce platform consumption data, etc., and uses statistical analysis, machine learning modeling, text mining and other methods to process data in order to explore user needs and optimize marketing decisions.

[0003] Among them, the intelligent marketing data analysis system refers to a system used to analyze and process marketing-related data to support marketing decisions and optimize delivery strategies. The system covers specific technical matters such as multi-data source collection and integration, user behavior analysis, marketing content association analysis, and advertising delivery evaluation. Its data collection part is based on crawling rules and API interface connection to achieve data aggregation from channels such as social media, e-commerce platforms, and advertising delivery platforms. User behavior analysis extracts user preference characteristics based on clickstream logs, dwell time, conversion path analysis, etc. Marketing content association analysis uses text matching, semantic analysis, keyword clustering, etc. to identify market hotspots and audience interests. Advertising delivery evaluation measures the effectiveness of marketing strategies through historical delivery data backtracking, exposure and click comparison, delivery time distribution, etc.

[0004] Traditional intelligent marketing data analysis mainly relies on static data analysis methods, which makes it difficult to capture the dynamic changes of user behavior in real time, resulting in delayed adjustments to marketing strategies. User behavior analysis is mostly limited to single indicators such as click-through rate and dwell time, and fails to integrate factors such as user access paths, bifurcation point selection, and behavior deviation, affecting the accuracy of user portraits. In the extraction of hot spots of interest, there is a lack of in-depth analysis of the rate of change of user dwell time, making it difficult to accurately grasp the real-time changes in points of interest, resulting in insufficient timeliness of content recommendations. Advertising delivery strategy evaluation often focuses on a simple comparison of exposure rate and click-through rate, and fails to fully consider the impact of delivery intervals on the smoothness of user interaction, which is prone to overexposure or underdelivery, affecting user experience. Existing marketing content evaluation methods fail to integrate delivery intervals, user behavior consistency, and conversion paths, making it difficult to accurately judge the actual effect of marketing content, resulting in a decrease in the utilization of advertising resources and limited optimization of marketing strategies. Summary of the invention

[0005] The object of the present invention is to solve the drawbacks existing in the prior art, and to propose an intelligent marketing data analysis system and method.

[0006] To achieve the above object, the present invention adopts the following technical solutions: An intelligent marketing data analysis system includes:

[0007] The user behavior deviation analysis module obtains the marketing interaction data of the user, and judges the changing trend of the marketing impact caused by the user behavior deviation according to the current exposure times of the marketing content, so as to obtain the user behavior deviation analysis result;

[0008] The interest hot spot extraction module obtains the residence time series of the user on various marketing content types based on the user behavior deviation analysis result, calculates the relative change rate in combination with the user access preference record, and obtains the user interest hot spot distribution information according to the change rate;

[0009] The path stability stratification module obtains the click path of the user corresponding to the interest hot spot and the access bifurcation point selection record based on the user interest hot spot distribution information, extracts the stable path nodes from them, and obtains the user path stability analysis result;

[0010] The advertising placement interval analysis module calculates the interaction fluency before and after the advertisement display based on the user path stability analysis result, adjusts the placement interval according to the user behaviors with decreased interaction fluency screened, and obtains the advertising placement interval optimization result;

[0011] The marketing content evaluation module obtains the user access data after the advertisement exposure according to the advertising placement interval optimization result, analyzes the conversion change of the marketing content under each advertisement interval, and generates the marketing content impact analysis result.

[0012] As a further solution of the present invention, the user behavior deviation analysis result includes the click frequency growth rate, the residence time change range, the access time distribution deviation degree, the marketing content click-through rate impact weight, and the marketing impact change trend. The user interest hot spot distribution information includes the residence time growth rate, the peak range, the change duration, the relative change rate, and the interest hot spot marketing content category. The user path stability analysis result includes stable path nodes, unstable path nodes, repeated access frequency, single access duration, and click distribution concentration. The advertising placement interval optimization result includes users with decreased interaction fluency, users who jump out at a specified time, the individual interaction rhythm comparison value, and the current behavior pattern deviation value. The marketing content impact analysis result includes the user residence time change range, the jump path after the advertisement click, the interaction heat of the advertisement display area, the influence degree of the advertising placement interval on the click-through rate, the consistency between the advertisement display time point and the user behavior, and the marketing content conversion change trend.

[0013] As a further solution of the present invention, the user behavior deviation analysis module includes:

[0014] The user behavior data acquisition sub-module acquires the marketing interaction data of the user, including the user click record, page stay time, and access timestamp, calculates the change in user click frequency, the fluctuation of page stay time, and the deviation of access time distribution within adjacent time periods, and filters the interaction data with the click frequency growth rate, the change range of stay time, and the deviation degree of access time distribution exceeding the set change threshold to obtain the user behavior deviation screening result;

[0015] The user behavior deviation weight calculation sub-module is based on the user behavior deviation screening result and uses the formula:

[0016]

[0017] Calculate the user click frequency change weight W c 、the user stay time fluctuation weight W t 、the access time distribution deviation weight W v to obtain the user behavior deviation influence weight;

[0018] Among them, C i represents the click frequency in the i-th time period, C i-1 represents the number of clicks in the previous time period i-1, n represents the total number of time periods, T j represents the stay time in the j-th time period, T j-1 represents the stay time of the previous page j-1, m represents the total number of time periods, V k represents the k-th access time point, represents the mean value of the access time, and p represents the total number of access samples;

[0019] The user behavior influence trend judgment sub-module analyzes the change trend of the user behavior deviation influence weight, including calculating the change rate of each weight within multiple time periods and filtering the time periods with the change rate exceeding the set threshold to generate the user behavior deviation analysis result.

[0020] As a further solution of the present invention, the interest hot spot extraction module includes:

[0021] The stay time series processing sub-module divides the time window based on the user behavior deviation analysis result, calls the user access record, extracts the stay time data of each marketing content category, arranges it in chronological order to form a stay time series, sets a time window with a fixed duration, and obtains the marketing content stay time series;

[0022] The stay time gradient calculation sub-module is based on the marketing content stay time series and uses the formula:

[0023]

[0024] Calculate the residence time gradient G between adjacent marketing contents x , and obtain the residence time gradient sequence;

[0025] wherein, X x is the residence time of the marketing content at time x, and X x+1 is the residence time of the marketing content at time x + 1, and ∈ is the smoothing term;

[0026] The interest hot spot distribution acquisition sub-module, based on the residence time gradient sequence, screens the marketing content categories whose gradient increase exceeds the set increase threshold, records the residence time growth rate, peak range, and change duration, calculates the relative change rate in combination with the user access preference record, and obtains the user interest hot spot distribution information according to the change rate, so as to obtain the user interest hot spot distribution information.

[0027] As a further solution of the present invention, the path stability stratification module includes:

[0028] The path record extraction sub-module, based on the user interest hot spot distribution information, obtains the user click path and the access bifurcation point selection record, and obtains the path click record and the bifurcation point characteristics according to the repeated access frequency, single access duration, and click distribution concentration;

[0029] The path stability determination sub-module, based on the path click record and the bifurcation point characteristics, screens the bifurcation points with stable click selection frequency and access duration close to the mean value, marks them as stable path nodes, and identifies the bifurcation points with access time deviating from the mean value in the click behavior, marks them as unstable path nodes, and obtains the user path stability analysis result.

[0030] As a further solution of the present invention, the advertisement delivery interval analysis module includes:

[0031] The user interaction feature extraction sub-module, based on the user path stability analysis result, obtains the user advertisement browsing path, calls the page interaction behavior data before and after the advertisement click, calculates the page residence duration, page scrolling rate, and number of interaction actions, normalizes each interaction behavior parameter, and obtains the user interaction feature;

[0032] The interaction fluency calculation sub-module, based on the user interaction feature, uses the formula:

[0033]

[0034] Calculate the change value F of the interaction fluency;

[0035] wherein, represents the page scrolling rate after the advertisement display after normalization, represents the page scrolling rate before the advertisement display after normalization, represents the page stay duration after the advertisement display after normalization, represents the page stay duration before the advertisement display after normalization, represents the number of interaction actions after the advertisement display after normalization, represents the number of interaction actions before the advertisement display after normalization;

[0036] The ad delivery interval optimization sub-module calls the change value of the interaction fluency, screens users with a decrease in interaction fluency or those who jump out within a specified time, calculates the individual interaction rhythm comparison value and the current behavior pattern deviation value, and adjusts the ad delivery interval according to the individual interaction rhythm and behavior pattern deviation to obtain the ad delivery interval optimization result.

[0037] As a further solution of the present invention, the marketing content evaluation module includes:

[0038] The ad interaction feature extraction sub-module obtains the user access data after the ad exposure according to the ad delivery interval optimization result, including the change range of the user stay time after the ad exposure, calculates the difference in the stay time before and after the ad display of the user, calls the jump path data after the ad click, extracts the page sequence accessed by the user after the click, counts the jump path length, obtains the interaction heat data of the ad display area, and calculates the number of clicks, mouse hover duration, and scrolling behavior amplitude within the ad area to obtain the ad interaction features;

[0039] The ad delivery interval influence calculation sub-module, based on the ad interaction features, uses the formula:

[0040]

[0041] calculates the ad delivery interval influence D;

[0042] where Q h is the interval time of the h-th ad delivery, is the average value of all ad delivery interval times, G h is the click-through rate of the h-th ad delivery, is the average click-through rate of all ad deliveries, and y is the number of ad delivery samples;

[0043] The marketing content conversion trend analysis sub-module analyzes the consistency between the ad display time point and the user behavior and the marketing content conversion change trend under each ad interval according to the ad delivery interval influence to obtain the marketing content influence analysis result.

[0044] An intelligent marketing data analysis method, which is executed based on the above intelligent marketing data analysis system, includes the following steps:

[0045] S1: Obtain the marketing interaction data of the user, judge the changing trend of the marketing impact caused by the user behavior deviation according to the current marketing content exposure times, and obtain the user behavior deviation analysis result;

[0046] S2: Based on the user behavior deviation analysis result, obtain the residence time series of the user on various marketing content types, calculate the relative change rate in combination with the user access preference record, and obtain the user interest hot spot distribution information according to the change rate;

[0047] S3: Based on the user interest hot spot distribution information, obtain the click path and the access bifurcation point selection record of the user corresponding to the interest hot spot, extract the stable path nodes from them, and obtain the user path stability analysis result;

[0048] S4: Based on the user path stability analysis result, calculate the interaction fluency before and after the advertisement display, adjust the delivery interval according to the user behavior with the screened decreased interaction fluency, and obtain the advertisement delivery interval optimization result;

[0049] S5: According to the advertisement delivery interval optimization result, obtain the user access data after the advertisement exposure, analyze the conversion change of the marketing content under each advertisement interval, and generate the marketing content impact analysis result.

[0050] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0051] In the present invention, through in-depth analysis of the user interaction data, the user behavior deviation is accurately identified, so that the marketing strategy can be dynamically adjusted to improve the delivery accuracy. By means of the changing trend of the residence time series and combining the access preference to calculate the relative change rate, the user interest hot spot distribution information is effectively extracted to improve the pertinence of content recommendation. Based on the access path stability analysis of the interest hot spot, the stable mode of the user behavior is extracted to optimize the advertisement delivery interval to reduce the user loss caused by frequent delivery. By using the calculation of the change of the interaction fluency, the user groups with decreased fluency or significant jump-out behavior are screened, and the advertisement push rhythm is reasonably adjusted to improve the advertisement experience. By comprehensively analyzing the user access data to analyze the conversion change of the user behavior after the advertisement display, the influence of different advertisement intervals on the user behavior is accurately evaluated, and the matching degree of the marketing content is optimized. By integrating multi-dimensional data, in the process of evaluating the marketing content, not only the click-through rate is considered, but also factors such as the user path behavior, the access time distribution, and the change of the residence time are comprehensively considered to make the evaluation more three-dimensional. Brief Description of the Drawings

[0052] Figure 1 is the system flow chart of the present invention;

[0053] Figure 2 is the flow chart of the user behavior deviation analysis module of the present invention;

[0054] Figure 3 It is a flowchart of the interest hot spot extraction module of the present invention;

[0055] Figure 4 It is a flowchart of the path stability stratification module of the present invention;

[0056] Figure 5 It is a flowchart of the advertisement placement interval analysis module of the present invention;

[0057] Figure 6 It is a flowchart of the marketing content evaluation module of the present invention. Detailed implementation manners

[0058] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0059] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0060] Please refer to Figure 1 , the present invention provides a technical solution: an intelligent marketing data analysis system includes:

[0061] The user behavior deviation analysis module obtains the marketing interaction data of the user, judges the changing trend of the marketing influence caused by the user behavior deviation according to the current exposure times of the marketing content, and obtains the user behavior deviation analysis result;

[0062] The interest hot spot extraction module obtains the residence time series of the user on various marketing content types based on the user behavior deviation analysis result, calculates the relative change rate in combination with the user access preference record, and obtains the user interest hot spot distribution information according to the change rate;

[0063] The path stability stratification module obtains the click path and the access bifurcation point selection record corresponding to the user interest hot spot based on the user interest hot spot distribution information, extracts the stable path nodes therefrom, and obtains the user path stability analysis result;

[0064] Based on the results of user path stability analysis, the advertising placement interval analysis module calculates the interaction fluency before and after the advertisement display, adjusts the placement interval according to the user behaviors with decreased interaction fluency selected, and obtains the optimized result of the advertising placement interval;

[0065] According to the optimized result of the advertising placement interval, the marketing content evaluation module obtains the user access data after the advertisement exposure, analyzes the conversion changes of the marketing content under each advertising interval, and generates the analysis result of the impact of the marketing content;

[0066] The analysis results of user behavior deviation include the growth rate of click frequency, the fluctuation range of stay time, the deviation degree of access time distribution, the influence weight of marketing content click-through rate, and the change trend of marketing influence. The information on the distribution of user interest hotspots includes the growth rate of stay time, the peak range, the duration of change, the relative change rate, and the category of marketing content of interest hotspots. The analysis results of user path stability include stable path nodes, unstable path nodes, repeated access frequency, single access duration, and click distribution concentration. The optimized result of the advertising placement interval includes users with decreased interaction fluency, users who jump out at a specified time, the comparison value of individual interaction rhythms, and the deviation value of the current behavior pattern. The analysis result of the impact of the marketing content includes the fluctuation range of user stay time, the jump path after the advertisement click, the interaction heat of the advertisement display area, the influence degree of the advertising placement interval on the click-through rate, the consistency between the advertisement display time point and user behavior, and the conversion change trend of the marketing content.

[0067] Please refer to Figure 2 , the user behavior deviation analysis module includes:

[0068] The user behavior data acquisition sub-module acquires the marketing interaction data of users, including user click records, page stay time, and access timestamps, calculates the change of user click frequency, the fluctuation of page stay time, and the deviation of access time distribution within adjacent time periods, and filters the interaction data with the growth rate of click frequency, the fluctuation range of stay time, and the deviation degree of access time distribution exceeding the set change threshold to obtain the screening result of user behavior deviation;

[0069] Obtain user click records, page dwell time, and access timestamps. For user click records, count the number of clicks by the user within a specific time period. For example, during the advertising period on a certain website, a user had 50 clicks between 9:00 - 10:00 and 30 clicks between 10:00 - 11:00. Record the click situations in these two time periods and calculate the change in click frequency. For page dwell time, record the duration the user stays on different web pages. For example, the dwell time on advertising page A is 120 seconds, and the dwell time on advertising page B is 45 seconds. Record the change in page dwell time. For access timestamps, record the specific times when the user enters and leaves the website. For example, the user enters at 9:05 and leaves at 9:25. The access timestamp records the interaction situation within this time range. Calculate the change in user click frequency, the fluctuation of page dwell time, and the deviation of access time distribution in adjacent time periods. For the change in click frequency, calculate the change in the number of clicks within a time period using the formula: where ΔC represents the click frequency change rate, C1 represents the number of clicks in the previous time period, and C2 represents the number of clicks in the current time period. If a user had 50 clicks in the previous hour and 30 clicks in the current hour, then calculate: Calculate the fluctuation of page dwell time using the formula: where ΔT represents the dwell time volatility rate, T1 represents the dwell time on the previous page, and T2 represents the dwell time on the current page. If the user stays on page A for 120 seconds and on page B for 45 seconds, then calculate: Calculate the deviation of access time distribution. For the timestamps of user entry and exit, calculate the average daily access duration of users using the formula: where V d represents the average daily access time, T exit,i′ represents the exit time of the i'-th user, T entry,i′ represents the entry time of the i'-th user, and n' represents the total number of users. Assuming the access times of 10 users are 30, 40, 50, 60, 70, 80, 90, 100, 110, and 120 seconds respectively, then: Filter out the interaction data where the growth rate of click frequency, the change range of dwell time, and the deviation degree of access time distribution exceed the set change thresholds. Assume the click frequency change rate threshold is set at ±30%, the dwell time volatility rate threshold is set at ±50%, and the access time deviation threshold is set at 20 seconds. In the above example, the click frequency change rate is -40%, exceeding the -30% threshold, and the dwell time volatility rate is -62.5%, exceeding the -50% threshold. Filter out the corresponding interaction data to obtain the user behavior deviation screening result.

[0070] The user behavior deviation weight calculation sub-module calculates based on the user behavior deviation screening result using the formula:

[0071]

[0072] Calculate the weight of the change in user click frequency W c , the weight of the fluctuation in user stay time W t , the weight of the deviation in access time distribution W v , and obtain the weight of the impact of user behavior deviation;

[0073] Among them, C i represents the click frequency in the i-th time period, C i-1 represents the number of clicks in the previous time period i - 1, and n represents the total number of time periods; T j represents the stay time in the j-th time period, T j-1 represents the stay time on the previous page j - 1, and m represents the total number of time periods; V k represents the k-th access time point, represents the mean value of the access time, and p represents the total number of access samples.

[0074] If an advertisement is displayed 1000 times in a day, then the number of exposures is 1000. The click-through rate CTR of the advertisement is the ratio of the number of times the advertisement is clicked by users to the number of exposures. Assuming that the advertisement is clicked 100 times, then calculate: The conversion rate CR is the ratio of users who complete the target behavior (such as purchase or registration) after clicking the advertisement. Assuming that 10 people complete the target behavior, then calculate: Calculate the weight of the change in user click frequency, the weight of the fluctuation in user stay time, and the weight of the deviation in access time distribution respectively, using the formula: Among them, W c represents the weight of the change in user click frequency, C i represents the click frequency in the i-th time period, and n represents the total number of time periods. Assuming that the number of clicks in 3 time periods is 30, 40, and 50, then calculate: Calculate the weight of the fluctuation in user stay time: Among them, W t represents the weight of the fluctuation in user stay time, T j represents the stay time in the j-th time period, and m represents the total number of time periods. Assuming that the stay times on 3 pages are 120, 100, and 80 seconds respectively, then calculate: Calculate the weight of the deviation in access time distribution: Among them, W v represents the weight of the deviation in access time distribution, V k represents the k-th access time point, represents the mean access time, p represents the total number of access samples. Assuming the average daily access time is 75 seconds, and the access times of a certain user are 30, 60, 90, and 120 seconds respectively, then calculate: Through operations, obtain the change weight of the user click frequency, the fluctuation weight of the user stay time, and the deviation weight of the access time distribution, and get the influence weight of the user behavior deviation.

[0075] The user behavior influence trend judgment sub-module analyzes the change trend of the influence weight of the user behavior deviation, including calculating the change rate of each weight within multiple time periods, and screening the time periods whose change rates exceed the set threshold to generate the analysis result of the user behavior deviation;

[0076] Call the influence weight of the user behavior deviation, analyze the change trends of the change weight of the user click frequency, the fluctuation weight of the user stay time, and the deviation weight of the access time distribution, calculate the change rates of each weight within different time periods, and screen the time periods whose change rates exceed the set threshold. Assume that a marketing platform records user behavior data every day, including click frequency, stay time, and access time distribution. Calculate that the change weights of the user click frequency in the past 5 days are 5, 6, 7, 8, and 9 respectively, and use the formula to calculate the click frequency change rate ΔW c : Among them, ΔW c represents the click frequency change rate, W c,当前 represents the change weight of the click frequency in the current time period, W c,初始 represents the change weight of the click frequency in the initial time period, and substitute the data for calculation: Similarly, calculate the change rate of the fluctuation weight of the user stay time. Assume that the fluctuation weights of the user stay time in the past 5 days are -10, -8, -5, -3, and 0 respectively, and use the formula: Substitute the data for calculation: Calculate the change rate of the deviation weight of the access time distribution. Assume that the deviation weights of the access time distribution of the user in the past 5 days are 30, 28, 25, 22, and 20 respectively, and use the formula: Substitute the data for calculation: Set the change rate threshold. Assume that the thresholds for the click frequency change rate, the dwell time fluctuation weight change rate, and the access time distribution offset weight change rate are set at 50%, 50%, and 30% respectively. Then, the click frequency change rate of 80% and the dwell time fluctuation weight change rate of 100% exceed the threshold, while the access time distribution offset weight change rate of -33.3% is within the threshold. Only the change trends of the click frequency and the dwell time fluctuation are screened out to judge the change trend of the marketing impact caused by the user behavior deviation. Assume that during a certain advertising campaign, the user's click frequency increases and the dwell time shortens. This may indicate that the advertisement attracts users to click but fails to retain them, and there may be problems with the advertising marketing content. On the other hand, if the click frequency decreases and the dwell time increases, it may mean that the advertisement exposure rate decreases but the marketing content is more attractive. Further analyze the direction of marketing strategy adjustment, and finally generate the analysis result of user behavior deviation.

[0077] Please refer to Figure 3 , and the interest hot spot extraction module includes:

[0078] Based on the analysis result of user behavior deviation, the dwell time series processing sub-module divides the time window, calls the user access records, extracts the dwell time data of each marketing content category, arranges them in chronological order to form a dwell time series, sets a time window with a fixed duration, and obtains the dwell time series of the marketing content;

[0079] Obtain the dwell time series of the user on multiple marketing content types. First, extract the dwell time information from the user's interaction data. Specifically, the access logs of the user on different marketing contents will record timestamps. By calculating the dwell duration of the user on a certain marketing content, an initial dwell time data set is formed. For example, if the user's access time on a certain product advertisement page is 10:02:15 and the departure time is 10:04:30, the dwell time is 135 seconds. All access records are arranged in chronological order to form a time series. Then, set a time window with a fixed duration, such as 30 minutes, and slide the window sequentially. In each time window, count the cumulative dwell time of each category of marketing content. By comparing the change trends of the dwell durations in different windows, screen out the marketing content categories with increasing or decreasing dwell times. For example, in a certain time window, the cumulative dwell time of a certain category of advertisement reaches 600 seconds, and in the next window, it increases to 900 seconds, indicating that the attractiveness of this category increases in this window. Finally, form the complete dwell time series data.

[0080] Based on the dwell time series of the marketing content, the dwell time gradient calculation sub-module uses the formula:

[0081]

[0082] Calculate the dwell time gradient G between adjacent marketing contents x, a dwell time gradient sequence is obtained;

[0083] Among them, X x is the dwell time of the marketing content at time x, and X x+1 is the dwell time of the marketing content at time x + 1, that is, the dwell time of this category in the next time window. ∈ is a smoothing term to prevent X x from being too small, which would cause the denominator to approach zero.

[0084] For example, if the dwell time of a certain marketing content category in the first window is 600 seconds and in the second window is 900 seconds, calculate its gradient change rate. If ∈ = 10 is set to smooth the gradient, then when X x = 600 seconds and X x+1 = 900 seconds, the calculated gradient is: By comparing the gradient changes between windows, the changing trend of the user's interest in a certain type of marketing content can be judged, and the dwell time gradient sequence can be obtained.

[0085] The interest hot spot distribution acquisition sub-module, based on the dwell time gradient sequence, screens out the marketing content categories whose gradient increase exceeds the set increase threshold, records the dwell time growth rate, peak range, and change duration, calculates the relative change rate in combination with the user's historical access preferences, and obtains the user interest hot spot distribution information according to the change rate to obtain the user interest hot spot distribution information;

[0086] Screen out the marketing content categories whose gradient increase exceeds the set increase threshold. Set the threshold to 10, and screen out the marketing content categories with a gradient change greater than 10. For example, if the calculated gradient value of a certain advertisement type is 12.25, it meets the screening conditions. Further record the dwell time growth rate. For example, between two adjacent time windows, its dwell time growth rate is: Among them, R represents the dwell time growth rate, and X x represents the dwell time of the marketing content at time x, and X x+1 represents the dwell time of the marketing content at time x + 1. If X x = 600 seconds and X x+1 = 900 seconds, then calculate its growth rate as: Next, record the peak range of the dwell time. For example, in a certain marketing content category, the maximum dwell time is recorded as 1200 seconds. The change duration is defined as the time window span. For example, if this marketing content category has been growing for three consecutive time windows, the change duration is 90 minutes. Combine the user's historical access preferences. For example, if a certain user's average dwell time for this category was previously 500 seconds, and the current data reaches 900 seconds, then calculate the relative change rate as: Among them, X h is the dwell time of the user's historical access preference. If Xh = 500 seconds, X x+1 = 900 seconds, then calculate its relative change rate as follows: Among them, x + 1 still represents the next window data point in the current time series. Finally, based on the change rate, judge the distribution information of user interest hotspots, record which marketing content categories show a significant increase in interest in user behavior, and obtain the distribution information of user interest hotspots.

[0087] Please refer to Figure 4 , the path stability stratification module includes:

[0088] The path record extraction sub-module, based on the distribution information of user interest hotspots, obtains the user click path and the record of access bifurcation point selection, and according to the repeated access frequency, single access duration, and click distribution concentration, obtains the path click record and the bifurcation point characteristics;

[0089] First, call the user's historical click data, sort all access records in chronological order, extract each bifurcation point, that is, different path selection points faced by the user during browsing. For example, in an e-commerce platform, the user may choose to enter a certain product category on the home page, or select a specific product detail page in the recommended marketing content. Each selection constitutes a bifurcation point. Record the selection path of the user at each bifurcation point, and accumulate the number of accesses of the user on each path. Then, calculate the repeated access frequency of each bifurcation point. For example, if the number of clicks of the user on path A is 50 times, on path B is 30 times, and on path C is 20 times, then the click frequency of path A is the highest. At the same time, calculate the stay time of the user at each bifurcation point. For example, the average access duration of the user on path A is 120 seconds, on path B is 95 seconds, and on path C is 65 seconds. Subsequently, calculate the click distribution concentration, that is, analyze whether the click behavior of the user at different bifurcation points is highly concentrated on certain specific paths. For example, if 90% of the user's clicks are distributed on path A, then the distribution concentration of path A is relatively high. Through the above steps, count the number of clicks, average access duration, and distribution concentration of all paths, and obtain the path click record and the bifurcation point characteristics.

[0090] The path stability determination sub-module, based on the path click record and the bifurcation point characteristics, screens out the bifurcation points with stable click selection frequency and access duration close to the historical average, marks them as stable path nodes, identifies the bifurcation points with access time deviating from the historical average in the click behavior, marks them as unstable path nodes, and obtains the user path stability analysis result;

[0091] Screen for bifurcation points with stable click selection frequencies and access durations close to the historical mean, and mark them as stable path nodes. Identify bifurcation points in click behavior where the access time deviates from the historical mean, and mark them as unstable path nodes. First, statistically analyze the click frequency stability of each bifurcation point, that is, calculate the fluctuation range of the click frequency of this path within multiple time windows. For example, in the past week, the click frequency changes of path A were 8%, 9%, 7%, 6%, 8%, 7%, and 9% respectively. If the fluctuation range is less than 10%, then this path is considered stable. Subsequently, calculate the fluctuation of the access duration, set the historical access mean as the benchmark. For example, the historical access mean of a certain path is 100 seconds, and the fluctuation threshold is set at ±20%. If the access duration of this path is 95 seconds on a certain day, it is considered that its change is within the stable range. Conversely, if the access duration of this path is 140 seconds, it is judged as an unstable path node. Finally, classify all bifurcation points. Bifurcation points with stable click frequencies and access durations close to the historical mean are marked as stable path nodes, while bifurcation points in click behavior where the access time deviates from the historical mean are marked as unstable path nodes. Finally, combine the stability data of all paths to obtain the user path stability analysis result.

[0092] Please refer to Figure 5 , and the advertising placement interval analysis module includes:

[0093] Based on the user path stability analysis result, the user interaction feature extraction sub-module obtains the user's advertising browsing path, calls the page interaction behavior data before and after the advertisement click, calculates the page stay duration, page scrolling rate, and the number of interaction actions, normalizes each interaction behavior parameter, and obtains the user interaction features;

[0094] First, obtain the user's advertisement browsing path, record the order of the pages visited by the user before and after the advertisement display, and calculate the residence time and interaction behaviors. For example, the user visits pages A, B, C, and D, and the residence time for each page is 30 seconds, 45 seconds, 20 seconds, and 60 seconds respectively. Then, call the interaction behavior data of the pages before and after the advertisement click, record the number of clicks, scrolling behavior, and mouse movement trajectory of the user on the pages. For example, the user clicks 3 times and scrolls 500 pixels on page A, clicks 5 times and scrolls 800 pixels on page B, clicks 2 times and scrolls 300 pixels on page C after the advertisement display, and clicks 4 times and scrolls 700 pixels on page D. Subsequently, calculate the residence duration of each page, and sum to obtain the total residence time of the user as 155 seconds. Calculate the scrolling rate of each page by dividing the scrolling distance by the corresponding residence time, and obtain the scrolling rates of pages A, B, C, and D as 16.67, 17.78, 15, and 11.67 pixels / second respectively. Then, calculate the number of interaction actions by adding up the number of clicks of the user on all pages, and obtain the total number of clicks as 14 times. After that, normalize each interaction behavior parameter. Assume that the range of the residence time of all users' pages within a period of time is from 20 seconds to 120 seconds, the range of the scrolling rate is from 10 to 20 pixels / second, and the range of the number of clicks is from 1 to 10 times. Then, the normalized residence time of the above user is (30 - 20) / (120 - 20) = 0.1, the normalized scrolling rate is (16.67 - 10) / (20 - 10) = 0.667, and the normalized number of clicks is (14 - 1) / (10 - 1) = 1.44. Finally, obtain the user interaction eigenvalue.

[0095] The interaction fluency calculation sub-module, based on the user interaction characteristics, uses the formula:

[0096]

[0097] Calculate the change value F of the interaction fluency;

[0098] Among them, represents the normalized scrolling rate of the page after the advertisement display, represents the normalized scrolling rate of the page before the advertisement display, represents the normalized residence duration of the page after the advertisement display, represents the normalized residence duration of the page before the advertisement display, represents the number of interaction actions after the advertisement display after normalization, represents the number of interaction actions before the advertisement display after normalization.

[0099] First, calculate the change rate of the page scrolling speed before and after the advertisement display. Assume that the normalized scrolling speed before the advertisement display is 0.6, and the normalized scrolling speed after the display is 0.4. Then the change rate is |0.4 - 0.6| = 0.2. Next, calculate the change rate of the stay duration. Assume that the normalized stay time before the advertisement display is 0.5, and after the display is 0.3. Then the change rate is |0.3 - 0.5| = 0.2. Subsequently, calculate the change rate of the number of interaction actions. Assume that the normalized number of clicks before the advertisement display is 0.7, and after the display is 0.5. Then the change rate is |0.5 - 0.7| = 0.2. After that, use the formula: Finally, calculate the change value of the interaction fluency.

[0100] The ad placement interval optimization sub-module calls the change value of the interaction fluency, filters out users with a decreased interaction fluency or those who jump out within a specified time, calculates the individual interaction rhythm comparison value and the current behavior pattern deviation value, and adjusts the ad placement interval according to the individual interaction rhythm and behavior pattern deviation to obtain the optimized result of the ad placement interval;

[0101] Call the change value of interaction fluency. First, screen out users with a decrease in interaction fluency or those who jump out within a specified time. Set the threshold for the change in interaction fluency to 0.4 and the threshold for the jump-out time to 5 seconds. Screen out users with a change value of interaction fluency less than 0.4 or a stay time less than 5 seconds. For example, a user has a change value of interaction fluency of 0.342, which is lower than the threshold of 0.4, and their page stay time is 4.5 seconds, which is lower than the threshold of 5 seconds, so they are screened out. Then calculate the individual interaction rhythm comparison value by comparing the user's interaction behavior parameters with the average value of all users. For example, if a user's scrolling rate is 15 pixels per second and the average value is 20 pixels per second, the comparison value is 15 / 20 = 0.75. Subsequently, calculate the deviation value of the current behavior pattern by comparing the user's current interaction behavior parameters with their historical average value. For example, if a user's current number of clicks is 5 times and the historical average value is 7 times, the deviation value is 5 / 7 ≈ 0.71. Then, combine the individual interaction rhythm comparison value and the deviation value of the current behavior pattern to determine the degree of change in the user's behavior after the advertisement is placed. Assume that the set threshold for the deviation of the behavior pattern is 0.8. Users with a deviation value lower than 0.8 are considered to be greatly affected by the advertisement and require optimization of the placement strategy. Subsequently, calculate the adjustment amount of the advertisement placement interval. Define the optimization rules: if the interaction rhythm comparison value is lower than 0.8 and the deviation value of the behavior pattern is lower than 0.8, the advertisement placement interval needs to be increased by 30%; if the interaction rhythm comparison value is lower than 0.6 and the deviation value of the behavior pattern is lower than 0.7, the placement interval is increased by 50%; if the interaction rhythm comparison value is higher than 0.9 and the deviation value of the behavior pattern is higher than 0.9, the placement interval is shortened by 20%. For example, the user's interaction rhythm comparison value is 0.75 and the deviation value of the behavior pattern is 0.71, both of which are lower than the corresponding thresholds. Therefore, the adjustment amount of their advertisement placement interval should be increased by 50%. If the current advertisement placement interval is 2 hours, the new placement interval is adjusted to 2×1.5 = 3 hours. Finally, obtain the optimization result of the advertisement placement interval. For example, for users with a decrease in interaction fluency and a large deviation in the behavior pattern, it is recommended to extend the advertisement placement interval, while for users with stable behavior, the advertisement placement interval is appropriately shortened to improve the advertisement reach efficiency.

[0102] Please refer to Figure 6 , the marketing content evaluation module includes:

[0103] The advertisement interaction feature extraction sub-module obtains the user access data after advertisement exposure according to the optimization result of the advertisement placement interval, including the change range of the user's stay time after advertisement exposure, calculates the difference in the stay time before and after the advertisement is shown to the user, calls the data of the jump path after the advertisement is clicked, extracts the sequence of pages visited by the user after clicking, counts the length of the jump path, obtains the interaction heat data of the advertisement display area, calculates the number of clicks, the mouse hover duration, and the scrolling behavior amplitude within the advertisement area, and obtains the advertisement interaction features;

[0104] Obtain the residence time of the user before and after the advertisement exposure, and calculate the change range. For example, if the residence time of the user before a certain advertisement is shown is 35 seconds and it is 50 seconds after the show, then the change range of the residence time is 50 - 35 = 15 seconds. Call the data of the jump path after the advertisement is clicked to analyze the page access sequence after the user clicks. For example, if the user visits pages A, B, and C in sequence after clicking the advertisement, then the length of the jump path is 3. Obtain the interactive heat data of the advertisement display area, and count the number of clicks, the duration of mouse hovering, and the amplitude of the scrolling behavior in this area. For example, the number of clicks of the user in a certain advertisement display area is 120 times, the duration of mouse hovering is 180 seconds, and the amplitude of the scrolling behavior is 350 pixels. Normalize the change range of the residence time, the length of the jump path, the interactive data, etc., and represent the advertisement interaction eigenvalue in the 0 - 1 interval. For example, calculate the normalized advertisement interaction eigenvalue to be 0.62. Finally, obtain the advertisement interaction eigenvalue.

[0105] The advertisement delivery interval influence calculation sub-module, based on the advertisement interaction characteristics, uses the formula:

[0106]

[0107] Calculate the advertisement delivery interval influence D;

[0108] Among them, Q h is the interval time of the h-th advertisement delivery, in minutes, which can be extracted from the advertisement delivery log data. For example, the delivery time interval of advertisement 1 is 30 minutes, and that of advertisement 2 is 60 minutes, etc. is the average value of all advertisement delivery interval times. G h is the click-through rate of the h-th advertisement delivery, which can be obtained from the website access data or the advertisement management background data. For example, the click-through rate of advertisement 1 is 0.08, and that of advertisement 2 is 0.05, etc. is the average click-through rate of all advertisement deliveries. y is the number of advertisement delivery samples, which can be directly counted from the advertisement delivery data. For example, the value in this calculation is 3.

[0109] First, analyze the relationship between the advertisement delivery interval and the click-through rate. For example, when the delivery intervals are 30 minutes, 60 minutes, and 90 minutes respectively, the corresponding advertisement click-through rates are 0.08, 0.05, and 0.03. Calculate the average click-through rate under different delivery intervals as: Calculate the average value of the delivery interval: Substitute into the formula: Calculate to obtain the advertisement delivery interval influence D = 0.666.

[0111] The marketing content conversion trend analysis sub-module analyzes the consistency between the advertisement display time points and user behaviors and the marketing content conversion change trends under each advertisement interval according to the advertisement placement interval influence degree, and obtains the marketing content influence analysis results;

[0112] Call the advertisement placement interval influence degree D = 0.666, analyze the consistency between the advertisement display time points and user behaviors, and obtain the trend data of the advertisement display time points on user behavior changes. For example, when the advertisement is displayed at 8 am, 2 pm, and 8 pm, the corresponding user click-through rates are 0.07, 0.05, and 0.02 respectively. Combine the advertisement placement interval influence degree to calculate its influence weight, set the benchmark click-through rate to 0.05, and calculate its relative deviation R h : Calculated as: The calculation results show that the relative deviations of the click-through rates at different advertisement display times are 0.266, 0, and 0.4 respectively. The click-through rate deviation at 8 pm is the largest. Combine the advertisement placement interval influence degree to obtain the marketing content influence analysis results. Among them, R h is the relative deviation of the advertisement display time point, which is obtained by calculating the relative change of the click-through rate multiplied by the placement interval influence degree. G h is the click-through rate of the h-th advertisement display time point, which can be obtained from the website access data. For example, the click-through rate in the morning is 0.07, and the click-through rate in the afternoon is 0.05, etc. 0.05 is the benchmark click-through rate, which can be calculated by the click-through rate statistics in the past period of time, such as the average click-through rate in the recent 30 days. D is the advertisement placement interval influence degree, and the calculation method is as described above, and the example value is 0.666. This result shows that the advertisement placement time point has an obvious impact on the user click-through rate, especially at 8 pm, where the relative deviation of the click-through rate is relatively large. Combining the advertisement placement interval influence degree, the advertisement display time can be optimized to improve the conversion effect of the marketing content.

[0113] An intelligent marketing data analysis method, which is executed based on the above intelligent marketing data analysis system, includes the following steps:

[0114] S1: Obtain the marketing interaction data of the user, and judge the marketing influence change trend caused by the user behavior deviation according to the current marketing content exposure times, and obtain the user behavior deviation analysis results;

[0115] S2: Based on the user behavior deviation analysis results, obtain the residence time series of the user on multiple marketing content types, calculate the relative change rate in combination with the user access preference records, and obtain the user interest hot spot distribution information according to the change rate;

[0116] S3: Based on the user interest hotspot distribution information, obtain the click paths and access bifurcation point selection records corresponding to the user's interest hotspots, extract the stable path nodes from them, and obtain the user path stability analysis result;

[0117] S4: Based on the user path stability analysis result, calculate the interaction fluency before and after the advertisement display, and adjust the delivery interval according to the user behaviors with decreased interaction fluency after screening, so as to obtain the optimized result of the advertisement delivery interval;

[0118] S5: According to the optimized result of the advertisement delivery interval, obtain the user access data after the advertisement exposure, analyze the conversion changes of the marketing content under each advertisement interval, and generate the marketing content impact analysis result.

[0119] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the technical marketing content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution marketing content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent marketing data analysis system, characterized in that: The system comprises: The user behavior deviation analysis module obtains the user's marketing interaction data, determines the marketing impact change trend caused by the user behavior deviation based on the current marketing content exposure times, and obtains the user behavior deviation analysis results; The interest hotspot extraction module obtains the user's stay time sequence on various marketing content types based on the user behavior deviation analysis results, calculates the relative change rate in combination with the user's access preference record, and obtains the user's interest hotspot distribution information according to the change rate; The path stability hierarchical module obtains the click path and access bifurcation point selection records of the user's corresponding interest hot spots based on the user's interest hot spot distribution information, extracts stable path nodes therefrom, and obtains the user path stability analysis result; The advertisement delivery interval analysis module calculates the interaction fluency before and after the advertisement display based on the user path stability analysis result, adjusts the delivery interval according to the filtered user behavior with decreased interaction fluency, and obtains the advertisement delivery interval optimization result; The marketing content evaluation module obtains user access data after advertisement exposure based on the advertisement delivery interval optimization result, analyzes the marketing content conversion changes under each advertisement interval, and generates marketing content impact analysis results.

2. The intelligent marketing data analysis system according to claim 1, characterized in that: The user behavior deviation analysis results include the click frequency growth rate, the change range of the dwell time, the degree of visit time distribution deviation, the marketing content click-through rate impact weight, and the marketing impact change trend. The user interest hotspot distribution information includes the dwell time growth rate, peak range, change duration, relative change rate, and interest hotspot marketing content category. The user path stability analysis results include stable path nodes, unstable path nodes, repeated visit frequency, single visit duration, and click distribution concentration. The advertising delivery interval optimization results include users with reduced interaction fluency, users who jump out at a specified time, individual interaction rhythm comparison values, and current behavior pattern deviation values. The marketing content impact analysis results include the user dwell time change range, the jump path after the ad click, the interaction heat of the ad display area, the degree of influence of the ad delivery interval on the click-through rate, the consistency between the ad display time point and the user behavior, and the marketing content conversion change trend.

3. The intelligent marketing data analysis system according to claim 2, characterized in that: The user behavior deviation analysis module includes: The user behavior data acquisition submodule acquires the user's marketing interaction data, including user click records, page dwell time, and access timestamps, calculates the change in user click frequency, page dwell time fluctuation, and access time distribution deviation in adjacent time periods, and screens the interaction data whose click frequency growth rate, dwell time change range, and access time distribution deviation exceed the set change threshold to obtain the user behavior deviation screening results; The user behavior deviation weight calculation submodule adopts the formula based on the user behavior deviation screening result: Calculate the user click frequency change weight W c , User stay time fluctuation weight W t , access time distribution offset weight W v , get the user behavior deviation impact weight; Among them, C i represents the click frequency in the i-th period, C i-1 represents the number of clicks in the previous time period i-1, n represents the total number of time periods, T j represents the residence time in the jth period, T j-1 represents the dwell time of the previous page j-1, m represents the total number of time periods, V k represents the kth access time point, represents the mean of access time, and p represents the total number of access samples; The user behavior impact trend judgment submodule analyzes the change trend of the user behavior deviation impact weight, including calculating the change rate of each weight in multiple time periods, and screening the time period where the change rate exceeds the set threshold, to generate a user behavior deviation analysis result.

4. The intelligent marketing data analysis system according to claim 3, characterized in that: The interest hotspot extraction module includes: The residence time sequence processing submodule divides the time window based on the user behavior deviation analysis result, calls the user access record, extracts the residence time data of each marketing content category, arranges them in chronological order to form a residence time sequence, sets a time window of fixed length, and obtains the marketing content residence time sequence; The residence time gradient calculation submodule adopts the formula based on the residence time sequence of the marketing content: Calculate the residence time gradient G between adjacent marketing content x , and obtain the residence time gradient sequence; Among them, X x is the marketing content retention time at time x, X x+1 is the residence time of marketing content at time x+1, ∈ is a smoothing term; The interest hotspot distribution acquisition submodule screens the marketing content categories whose gradient increase exceeds the set increase threshold based on the residence time gradient sequence, records the residence time growth rate, peak range, and change duration, calculates the relative change rate in combination with the user access preference record, obtains the user interest hotspot distribution information based on the change rate, and obtains the user interest hotspot distribution information.

5. The intelligent marketing data analysis system according to claim 4, characterized in that: The path stability hierarchical module includes: The path record extraction submodule obtains the user's click path and visit bifurcation point selection records based on the user's interest hotspot distribution information, and obtains the path click record and bifurcation point characteristics according to the repeated visit frequency, single visit duration, and click distribution concentration; The path stability determination submodule screens the bifurcation points whose click selection frequency is stable and whose access duration is close to the mean based on the path click records and bifurcation point characteristics, marks them as stable path nodes, identifies the bifurcation points whose access time deviates from the mean in the click behavior, marks them as unstable path nodes, and obtains the user path stability analysis results.

6. The intelligent marketing data analysis system according to claim 5, characterized in that: The advertisement delivery interval analysis module includes: The user interaction feature extraction submodule obtains the user's ad browsing path based on the user path stability analysis result, calls the page interaction behavior data before and after the ad click, calculates the page dwell time, page scrolling rate, and the number of interaction actions, normalizes each interaction behavior parameter, and obtains the user interaction feature; The interaction fluency calculation submodule adopts the formula based on the user interaction characteristics: Calculate the interaction fluency change value F; in, Represents the normalized page scrolling rate after the ad is displayed. Represents the normalized page scrolling rate before the ad is displayed. Represents the normalized page dwell time after the ad is displayed. Represents the normalized page dwell time before the ad is displayed. Represents the normalized number of interactive actions after ad display. Represents the normalized number of interactive actions before ad display; The delivery interval optimization submodule calls the interaction fluency change value, filters out users whose interaction fluency decreases or who jump out within a specified time, calculates the individual interaction rhythm comparison value and the current behavior pattern deviation value, adjusts the advertising delivery interval according to the individual interaction rhythm and behavior pattern deviation, and obtains the advertising delivery interval optimization result.

7. The intelligent marketing data analysis system according to claim 6, characterized in that: The marketing content evaluation module includes: The advertisement interaction feature extraction submodule obtains the user access data after advertisement exposure according to the advertisement delivery interval optimization result, including the change range of the user's stay time after advertisement exposure, calculates the difference of the user's stay time before and after advertisement display, calls the jump path data after advertisement click, extracts the page sequence visited by the user after the click, counts the jump path length, obtains the interaction heat data of the advertisement display area, calculates the number of clicks, the mouse hovering time and the scrolling behavior range in the advertisement area, and obtains the advertisement interaction features; The delivery interval impact calculation submodule is based on the advertisement interaction characteristics and adopts the formula: Calculate the impact D of the advertising delivery interval; Among them, Q h is the interval between the hth advertisement delivery. is the average of all advertising intervals, G h is the click-through rate of the hth ad delivery, is the average click-through rate of all ads, and y is the number of ad samples; The marketing content conversion trend analysis submodule analyzes the consistency between the advertisement display time point and the user behavior and the marketing content conversion change trend under each advertisement interval according to the influence of the advertisement delivery interval, and obtains the marketing content impact analysis result.

8. An intelligent marketing data analysis method, characterized in that: The intelligent marketing data analysis system according to any one of claims 1 to 7 is implemented, comprising the following steps: S1: Obtain the user's marketing interaction data, determine the marketing impact change trend caused by the user's behavior deviation based on the current marketing content exposure times, and obtain the user behavior deviation analysis results; S2: Based on the user behavior deviation analysis results, obtain the user's stay time series on various marketing content types, calculate the relative change rate in combination with the user's access preference records, and obtain the user's interest hotspot distribution information according to the change rate; S3: Based on the user's interest hotspot distribution information, obtain the user's click path corresponding to the interest hotspot and the visit bifurcation point selection record, extract stable path nodes from them, and obtain the user path stability analysis result; S4: Based on the user path stability analysis result, the interaction fluency before and after the advertisement is displayed is calculated, and the advertisement delivery interval is adjusted according to the filtered user behaviors with decreased interaction fluency to obtain an advertisement delivery interval optimization result; S5: According to the optimization result of the advertisement delivery interval, the user access data after advertisement exposure is obtained, the marketing content conversion changes under each advertisement interval are analyzed, and the marketing content impact analysis result is generated.

Images