Advertisement analysis system and method based on AI intelligence

Abstract

The invention discloses an AI intelligence-based advertisement analysis system and method, and particularly relates to the technical field of advertisement analysis, and the method comprises the steps: obtaining a first period of putting an advertisement on a target APP, dividing the first period into a plurality of sub-periods, and extracting the advertisement click rate of each sub-period; determining data points of each sub-cycle to obtain a data set according to the number of the newly added users and the number of the purchase users, clustering the data set to determine an advertisement putting effect index of each sub-cycle, drawing a time-advertisement putting effect broken line graph, and dividing a trend state of the advertisement putting effect of a first cycle according to the broken line graph, and when the trend state of the first period is the rising trend, the duration and the putting frequency of the advertisement in the next putting period of the target APP are increased, so that the advertisement putting frequency and the duration of the putting period can be flexibly changed according to the actual advertisement putting effect in the advertisement putting process, the budget waste is reduced, and the advertisement putting efficiency is improved. And meanwhile, the maximization of the advertising effect is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of advertising analysis, and in particular to an advertising analysis system and method based on AI intelligence. Background Art

[0002] Advertising refers to the process by which an enterprise or brand delivers advertising content to the target audience through various channels, and uses AI intelligent analysis to determine whether the actual advertising effect has successfully aroused the audience's interest and achieved the expected marketing goals;

[0003] In the advertising delivery method, generally speaking, companies or brands deliver advertisements to a certain APP, and analyze whether the marketing effect of the advertisements reaches the expected result based on the advertising effect in the APP, so as to judge whether the advertising marketing is successful. In general, companies or brands will set corresponding delivery strategies according to specific APPs, such as delivery cycle and delivery frequency, so as to maximize the effect of advertising delivery.

[0004] However, in the process of advertising, the delivery strategy on the APP is basically fixed. If the delivery frequency and duration of the delivery cycle cannot be flexibly changed according to the actual delivery effect, it may lead to budget waste and failure to maximize the advertising effect. Summary of the invention

[0005] The purpose of the present invention is to solve the problem and to propose an advertising analysis system and method based on AI intelligence.

[0006] In a first aspect of the present invention, an advertising analysis method based on AI intelligence is first proposed, the method comprising:

[0007] For the target APP, obtain the most recent cycle in which the advertisement is delivered on the target APP as the first cycle, divide the first cycle into several sub-cycles according to the preset time window, and extract the advertisement click-through rate, number of new users and number of purchasing users in each sub-cycle;

[0008] Determine the data points of each sub-period according to the advertisement click-through rate, the number of new users and the number of purchasing users to obtain a data set, and cluster the data set using a preset clustering model to determine the advertisement delivery effect index of each sub-period;

[0009] Draw a time-advertising effect line graph according to the advertising effect index of each sub-period, and divide the trend state of the advertising effect of the first period according to the line graph; the trend state includes a downward trend, a stable trend and an upward trend;

[0010] When the trend of the advertising effect in the first cycle is an upward trend, increase the duration and frequency of the next advertising cycle on the target APP.

[0011] Optionally, the data points of each sub-period are determined based on the advertisement click-through rate, the number of new users, and the number of purchasing users, including:

[0012] Get the total number of times users click on ads, the total number of times ads are displayed, the average time users stay on the page after clicking ads, and the total number of times users interact with ads in each sub-period, and get the first ad click coefficient of the corresponding sub-period. The calculation formula is:

[0013]

[0014] In the formula, Esa is the first ad click coefficient, gt, ht, sd, and ku are the total number of times users click on ads, the total number of times ads are displayed, the average time users stay on the page after clicking ads, and the total number of times users interact with ads. is the user click-through rate, ω1, ω2, and ω3 are respectively the preset proportional coefficients of the user click-through rate, the average time a user stays on the page after clicking on the ad, and the total number of times a user interacts with the ad;

[0015] The second advertisement click coefficient is obtained according to the first advertisement click coefficient of each sub-period and the weight value of the preset advertisement click coefficient of the corresponding sub-period in the advertisement delivery effect. The calculation formula is: Esx = T1 × Esa; wherein Esx is the second advertisement click coefficient, and T1 is the weight value of the preset advertisement click coefficient of the sub-period in the advertisement delivery effect;

[0016] The data point of each sub-period is determined according to the second advertisement click coefficient, the number of newly added users and the number of purchasing users in each sub-period.

[0017] Optionally, determining the data points of each sub-period according to the second advertisement click coefficient, the number of newly added users, and the number of purchasing users of each sub-period includes:

[0018] Obtain the total number of new users registered through the ads delivered, the total number of ad impressions, the user retention rate, and the average time required for new users to convert from being exposed to ads to becoming users in each sub-period, and obtain the first ad acquisition coefficient of the corresponding sub-period. The calculation formula is:

[0019]

[0020] Where Was is the coefficient of attracting new users for the first advertisement, rf, ht, re, and es are the total number of new users, the total number of advertisement displays, the user retention rate, and the average time required for new users to convert from being exposed to advertisements to being users. is the user addition rate, a1, a2, and a3 are the preset proportional coefficients of the user addition rate, user retention rate, and the average time required for new users to convert from being exposed to advertisements to being users;

[0021] The second advertising new customer acquisition coefficient is obtained according to the first advertising new customer acquisition coefficient of each sub-period and the weight value of the preset advertising new customer acquisition coefficient of the corresponding sub-period in the advertising delivery effect. The calculation formula is: Wds = T2 × Was; where Wds is the second advertising new customer acquisition coefficient, and T2 is the weight value of the preset advertising new customer acquisition coefficient of the sub-period in the advertising delivery effect;

[0022] The data point of each sub-period is determined based on the second advertisement click coefficient, the second advertisement new user acquisition coefficient and the number of purchasing users of each sub-period.

[0023] Optionally, determining the data points of each sub-period according to the second advertisement click coefficient, the second advertisement new user acquisition coefficient, and the number of purchasing users of each sub-period includes:

[0024] Get the total number of users who purchased through ads, the total number of ad impressions, the average user purchase frequency, and the average order value in each sub-period, and get the first ad conversion coefficient of the corresponding sub-period. The calculation formula is:

[0025]

[0026] In the formula, Rds is the first advertising conversion coefficient, kl, ht, nv, and py are the total number of users who purchased through advertising, the total number of advertising displays, the average user purchase frequency, and the average order value, respectively. Purchase user increase rate, b1, b2, b3 are preset proportional coefficients of user purchase user increase rate, user average purchase frequency and average order value respectively;

[0027] The second advertising conversion coefficient is obtained according to the first advertising conversion coefficient of each sub-period and the weight value of the preset advertising conversion coefficient of the corresponding sub-period in the advertising delivery effect, and the calculation formula is: Rcs = T3 × Rfs; where Rcs is the second advertising conversion coefficient, and T3 is the weight value of the preset advertising conversion coefficient of the sub-period in the advertising delivery effect;

[0028] The second advertisement click coefficient, the second advertisement new customer acquisition coefficient and the second advertisement conversion coefficient of each sub-period are used as data points of each sub-period.

[0029] Optionally, clustering the data set using a preset clustering model to determine the advertising delivery effect index of each sub-period includes:

[0030] The steps for clustering the data set according to the K-means clustering method stored in the preset clustering model are:

[0031] Step 1: Used to determine the optimal number of clusters for the data set using the elbow method;

[0032] Step 2: Randomly select a corresponding number of data points as initial clusters according to the optimal number of clusters;

[0033] Step 3: For each data point, use the Euclidean distance algorithm to calculate the distance between the data point and each initial cluster center, and assign the data point to the cluster to which the nearest cluster center belongs. After traversing all data points, the initial cluster center is obtained.

[0034] Step 4: Calculate the average value of all data points for each initial cluster and use the average value as the new cluster center;

[0035] Step 5: It is used to determine whether the cluster center has changed. If it has changed, repeat steps 3 and 4 until the cluster center no longer changes, and obtain the final cluster and the final cluster center;

[0036] The final clusters and the final cluster centers are mapped to three-dimensional coordinates to obtain the advertising effect index of each sub-period.

[0037] Optionally, the final clusters and the final cluster centers are mapped to three-dimensional coordinates to obtain the advertising effect index of each sub-period, including:

[0038] Calculate the weight coefficient of each final cluster, the calculation formula is: Where W j It is expressed as the weight coefficient of the jth final cluster, H 总 Represented as the number of all data points in the data set, H j Represents the number of data points in the jth final cluster;

[0039] Calculate the distance of each data point in each final cluster to the coordinate origin in three-dimensional coordinates to obtain the preliminary coordinate distance of the data point, and multiply the preliminary coordinate distance of the data point by the weight coefficient corresponding to the final cluster in which the data point is located to obtain the actual coordinate distance of the data point;

[0040] The actual coordinate distance of each data point is used as the advertising effect index of the corresponding sub-period.

[0041] Optionally, a time-advertising delivery effect line graph is drawn according to the advertising delivery effect index of each sub-period, and the trend state of the advertising delivery effect of the first period is divided according to the line graph, including:

[0042] Arrange the advertising effect index of each sub-period based on the time sequence to obtain an advertising effect index sequence, and draw a time-advertising effect line graph based on the advertising effect index sequence;

[0043] Calculate the slope of every two adjacent sub-periods based on the time-advertising effect line graph, and calculate the mean of all slopes to obtain the overall average slope K;

[0044] When the overall average slope K = 0, it means that the trend state of the advertising effect in the first cycle is a stable trend;

[0045] When the overall average slope K>0, it means that the trend state of the advertising effect in the first cycle is an upward trend;

[0046] When the overall average slope K<0, it means that the trend state of the advertising effect in the first cycle is a downward trend.

[0047] Optionally, when the trend state of the advertising delivery effect in the first cycle is a stable trend, the delivery cycle length and delivery frequency of the next cycle on the APP are consistent with those of the first cycle;

[0048] When the trend state of the advertising effect in the first cycle is a downward trend, the advertising on the APP is stopped.

[0049] Optionally, when the trend state of the advertising delivery effect in the first cycle is an upward trend, adjusting the duration and delivery frequency of the next delivery cycle of the advertisement on the target APP includes:

[0050] Calculate the average of the advertising effect index of each sub-period in the first period as the adjustment coefficient. Adjust the time and frequency of the advertising cycle on and off the APP according to the adjustment coefficient. The calculation formula is:

[0051] TRQ=TRE×(1+wk)

[0052] FEQ=FED×(1+wk)

[0053] Where TRQ is the time of the new delivery cycle, TRE is the time of the initial delivery cycle, FEQ is the new delivery frequency, FED is the initial delivery frequency, and wk is the adjustment coefficient.

[0054] In a second aspect of the present invention, an AI-based advertising analysis system is proposed, comprising:

[0055] Data acquisition module: for the target APP, obtain the most recent cycle in which the advertisement is placed on the target APP as the first cycle, divide the first cycle into several sub-cycles according to the preset time window, and extract the advertisement click-through rate, number of new users and number of purchasing users in each sub-cycle;

[0056] Cluster analysis module: determine the data points of each sub-period according to the advertisement click-through rate, the number of new users and the number of purchasing users, obtain a data set, and cluster the data set through a preset clustering model to determine the advertisement delivery effect index of each sub-period;

[0057] Trend judgment module: draw a time-advertising effect line graph according to the advertising effect index of each sub-period, and divide the trend state of the advertising effect of the first period according to the line graph; the trend state includes a downward trend, a stable trend and an upward trend;

[0058] Adjustment module: When the trend of the advertising effect in the first cycle is an upward trend, increase the duration and frequency of the next advertising cycle on the target APP.

[0059] Beneficial effects of the present invention:

[0060] The present invention proposes an advertising analysis system and method based on AI intelligence, which obtains the first cycle of advertising delivery on a target APP and divides it into several sub-cycles, extracts the click-through rate of advertisements in each sub-cycle, and determines the data points of each sub-cycle by the number of new users and the number of purchasing users, and obtains a data set. The data set is clustered by a preset clustering model to determine the advertising delivery effect index of each sub-cycle, and a time-advertising delivery effect line graph is drawn. The trend state of the advertising delivery effect of the first cycle is divided according to the line graph. When the trend state of the advertising delivery effect of the first cycle is an upward trend, the duration and delivery frequency of the next delivery cycle of the advertisement on the target APP are increased. In this way, in the process of advertising delivery, the frequency of advertising delivery and the duration of the delivery cycle can be flexibly changed according to the actual advertising delivery effect of the APP, thereby reducing the waste of budget and maximizing the advertising effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The present invention will be further described below in conjunction with the accompanying drawings.

[0062] Figure 1 This is a flowchart of an advertising analysis method based on AI intelligence;

[0063] Figure 2 This is a framework diagram of an AI-based advertising analysis system. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making any creative work shall fall within the scope of protection of the present invention.

[0066] The embodiment of the present invention provides an advertisement analysis method based on AI intelligence. Figure 1 , Figure 1 A flowchart of an AI-based advertising analysis method provided by an embodiment of the present invention. The method comprises the following steps:

[0067] For the target APP, obtain the most recent cycle in which the advertisement is delivered on the target APP as the first cycle, divide the first cycle into several sub-cycles according to the preset time window, and extract the advertisement click-through rate, number of new users and number of purchasing users in each sub-cycle;

[0068] Determine the data points of each sub-period according to the advertisement click-through rate, the number of new users and the number of purchasing users, obtain a data set, and cluster the data set using a preset clustering model to determine the advertisement delivery effect index of each sub-period;

[0069] Draw a time-advertising effect line graph based on the advertising effect index of each sub-period, and divide the trend status of the advertising effect of the first period according to the line graph; the trend status includes a downward trend, a stable trend and an upward trend;

[0070] When the trend of the advertising effect in the first cycle is an upward trend, increase the duration and frequency of the next advertising cycle on the target APP.

[0071] Based on an AI-based advertising analysis method provided in an embodiment of the present invention, through the above-mentioned method, during the advertising delivery process, the frequency of advertising delivery and the duration of the delivery cycle can be flexibly changed according to the actual advertising delivery effect of the APP, thereby reducing the waste of budget and maximizing the advertising effect.

[0072] It should be noted that the first cycle is divided into several sub-cycles according to the preset time window. The preset time window is set by professionals according to the actual situation. For example, assuming that the preset time window is weekly, daily, etc., the specific preset time window is determined according to the actual situation and is not limited.

[0073] In one embodiment, determining the data points of each sub-period based on the advertisement click rate, the number of new users, and the number of purchasing users includes:

[0074] Get the total number of times users click on ads, the total number of times ads are displayed, the average time users stay on the page after clicking ads, and the total number of times users interact with ads in each sub-period, and get the first ad click coefficient of the corresponding sub-period. The calculation formula is:

[0075]

[0076] In the formula, Esa is the first ad click coefficient, gt, ht, sd, and ku are the total number of times users click on ads, the total number of times ads are displayed, the average time users stay on the page after clicking ads, and the total number of times users interact with ads. is the user click-through rate, ω1, ω2, and ω3 are respectively the preset proportional coefficients of the user click-through rate, the average time a user stays on the page after clicking on the ad, and the total number of times a user interacts with the ad;

[0077] The second advertisement click coefficient is obtained according to the first advertisement click coefficient of each sub-period and the weight value of the preset advertisement click coefficient of the corresponding sub-period in the advertisement delivery effect. The calculation formula is: Esx = T1 × Esa; wherein Esx is the second advertisement click coefficient, and TI is the weight value of the preset advertisement click coefficient of the sub-period in the advertisement delivery effect;

[0078] The data point of each sub-period is determined according to the second advertisement click coefficient, the number of newly added users and the number of purchasing users in each sub-period.

[0079] It should be noted that the number of ad clicks and impressions can be recorded through the click tracking system of the target APP, and the average time users stay on the page after clicking on the ad and the number of times users interact with the ad can be obtained through tracking technology or behavioral analysis tools. These data usually come from the ad monitoring system within the APP, which automatically collects relevant indicators and conducts summary analysis by tracking users' click behaviors, browsing habits and interactions in real time; in addition, user-ad interaction refers to user-ad-related interactions, such as comments and sharing, which are all interactions, and of course can also include other forms of interaction. The specific interactive action standards are set by the advertiser itself, so I will not go into details.

[0080] It should be noted that ω1, ω2, and ω3 are set by professionals according to actual conditions. Generally, the sum of ω1, ω2, and ω3 is 1. For example, ω1, ω2, and ω3 can be 0.3, 0.3, and 0.4, respectively, or other numbers, without specific limitation.

[0081] It should be noted that T1 is the weight value of the preset ad click coefficient of the sub-period in the ad delivery effect. The specific value is set by professionals according to the time period of the actual sub-period. For example, when the sub-period is in the initial time of the first cycle of ad delivery, the weight of the user's ad click coefficient may be larger for the ad delivery effect, because in the early stage of ad delivery, the ad exposure and attractiveness are high, and the user's attention to the new ad will also be higher. At this time, the user's behavior of clicking on the ad is an important indicator for measuring the ad effect; therefore, the weight of the ad click coefficient should be larger in the early stage to better reflect the attractiveness of the ad to the user; as the delivery time progresses, the user's familiarity with the ad increases, the click rate may gradually decrease, and other indicators (such as the number of new users and the number of purchasing users) may become more important. Therefore, in the later stage, the weight of the ad click coefficient can be gradually reduced, and the weight of other effect indicators can be increased. In this way, dynamically adjusting the weight value of the ad click coefficient can improve the accuracy of the ad delivery effect evaluation; in addition, the value range of T1 is greater than 0 and less than 1. The specific value of T1 is set by professionals and is not limited or elaborated.

[0082] In one embodiment, determining the data points of each sub-period according to the second advertisement click coefficient, the number of newly added users, and the number of purchasing users of each sub-period includes:

[0083] Obtain the total number of new users registered through the ads delivered, the total number of ad impressions, the user retention rate, and the average time required for new users to convert from being exposed to ads to becoming users in each sub-period, and obtain the first ad acquisition coefficient of the corresponding sub-period. The calculation formula is:

[0084]

[0085] Where Was is the coefficient of attracting new users for the first advertisement, rf, ht, re, and es are the total number of new users, the total number of advertisement displays, the user retention rate, and the average time required for new users to convert from being exposed to advertisements to being users. is the user addition rate, a1, a2, and a3 are the preset proportional coefficients of the user addition rate, user retention rate, and the average time required for new users to convert from being exposed to advertisements to being users;

[0086] The second advertising new customer acquisition coefficient is obtained according to the first advertising new customer acquisition coefficient of each sub-period and the weight value of the preset advertising new customer acquisition coefficient of the corresponding sub-period in the advertising delivery effect. The calculation formula is: Wds = T2 × Was; where Wds is the second advertising new customer acquisition coefficient, and T2 is the weight value of the preset advertising new customer acquisition coefficient of the sub-period in the advertising delivery effect;

[0087] The data point of each sub-period is determined based on the second advertisement click coefficient, the second advertisement new user acquisition coefficient and the number of purchasing users of each sub-period.

[0088] It should be noted that the total number of new users and the total number of ad impressions can be obtained through the target APP, which can usually be achieved through background statistical functions or analysis tools; secondly, the user retention rate can be calculated by tracking the user's activity after registration, usually using data analysis software or database monitoring; finally, the average time required for new users to convert from exposure to ads can be obtained through user behavior tracking systems, or other acquisition methods, depending on the actual situation, and no limitation or elaboration is made.

[0089] It should be noted that a1, a2, and a3 are set by professionals according to actual conditions. Generally, the sum of a1, a2, and a3 is 1. For example, a1, a2, and a3 can be 0.4, 0.3, and 0.3, respectively, or other numbers, without specific limitation.

[0090] It should be noted that T2 is the weight value of the preset advertising new user acquisition coefficient of the sub-cycle in the advertising delivery effect. The specific value is set by professionals according to the time period of the actual sub-cycle. For example, in the early and middle stages of advertising delivery, since users pay more attention to new advertisements, the weight of the advertising new user acquisition coefficient at this time may be set larger to emphasize the acquisition and conversion effects of new users; in the later stages of advertising delivery, as users become more familiar with the brand, the importance of advertising new user acquisition may decline. At this time, professionals may adjust the weight accordingly and increase the proportion of other indicators (such as the number of purchasing users) to more comprehensively reflect changes in advertising effectiveness and user behavior; this flexible weight setting can more accurately reflect the actual effects of advertising delivery at different stages and provide a basis for subsequent optimization and adjustment; in addition, the value range of T2 is greater than 0 and less than 1. The specific value of T2 is set by professionals and is not limited or elaborated.

[0091] In one embodiment, determining the data points of each sub-period according to the second advertisement click coefficient, the second advertisement new user acquisition coefficient, and the number of purchasing users of each sub-period includes:

[0092] Get the total number of users who purchased through ads, the total number of ad impressions, the average user purchase frequency, and the average order value in each sub-period, and get the first ad conversion coefficient of the corresponding sub-period. The calculation formula is:

[0093]

[0094] In the formula, Rds is the first advertising conversion coefficient, kl, ht, nv, and py are the total number of users who purchased through advertising, the total number of advertising displays, the average user purchase frequency, and the average order value, respectively. Purchase user increase rate, b1, b2, b3 are preset proportional coefficients of user purchase user increase rate, user average purchase frequency and average order value respectively;

[0095] The second advertising conversion coefficient is obtained according to the first advertising conversion coefficient of each sub-period and the weight value of the preset advertising conversion coefficient of the corresponding sub-period in the advertising delivery effect, and the calculation formula is: Rcs = T3 × Rfs; where Rcs is the second advertising conversion coefficient, and T3 is the weight value of the preset advertising conversion coefficient of the sub-period in the advertising delivery effect;

[0096] The second advertisement click coefficient, the second advertisement new customer acquisition coefficient and the second advertisement conversion coefficient of each sub-period are used as data points of each sub-period.

[0097] It should be noted that the total number of ad impressions and the total number of users who purchased through the ads can be collected through the background data of the target APP. These can usually be obtained through reports or API interfaces provided by the platform. The average user purchase frequency and average order value can be obtained through analysis of user transaction records, or through other acquisition methods, which depends on the actual situation and is not limited or elaborated on.

[0098] It should be noted that b1, b2, and b3 are set by professionals according to actual conditions. Generally, the sum of b1, b2, and b3 is 1. For example, b1, b2, and b3 can be 0.25, 0.35, and 0.4, respectively, or other numbers, without specific limitation.

[0099] It should be noted that T3 is the weight value of the preset advertising conversion coefficient of the sub-cycle in the advertising delivery effect. The specific value is set by professionals based on the actual time period of the sub-cycle. For example, in the early stage of advertising delivery, the weight of the conversion coefficient may be relatively low, because users’ awareness of the advertisement is not yet in-depth at this time. As time goes by, users’ familiarity with the brand increases, and the conversion rate may increase. Therefore, the weight of the conversion coefficient should be increased accordingly in the later stage to better reflect the actual effect of the advertisement. In addition, the value range of T3 is greater than 0 and less than 1. The specific value of T3 is set by professionals and is not limited or elaborated.

[0100] It should be noted that T1, T2, and T3 are all set by professionals according to actual conditions, and T1+T2+T3=1. The specific values ​​of T1, T2, and T3 are flexibly set by professionals according to the time of each sub-cycle, and are not limited or elaborated.

[0101] In one embodiment, clustering the data set using a preset clustering model to determine the advertising delivery effect index of each sub-period includes:

[0102] The steps for clustering the data set according to the K-means clustering method stored in the preset clustering model are:

[0103] Step 1: Used to determine the optimal number of clusters for the data set using the elbow method;

[0104] Step 2: Randomly select a corresponding number of data points as initial clusters according to the optimal number of clusters;

[0105] Step 3: For each data point, use the Euclidean distance algorithm to calculate the distance between the data point and each initial cluster center, and assign the data point to the cluster to which the nearest cluster center belongs. After traversing all data points, the initial cluster center is obtained.

[0106] Step 4: Calculate the average value of all data points for each initial cluster and use the average value as the new cluster center;

[0107] Step 5: It is used to determine whether the cluster center has changed. If it has changed, repeat steps 3 and 4 until the cluster center no longer changes, and obtain the final cluster and the final cluster center;

[0108] The final clusters and the final cluster centers are mapped to three-dimensional coordinates to obtain the advertising effect index of each sub-period.

[0109] In one implementation, according to the above method, the optimal clustering results of the data set can be obtained. These clustering results can be used to identify different clusters in the data set, as well as the center and characteristics of each cluster. Such clustering results are helpful for further data analysis and interpretation.

[0110] It should be noted that the preset clustering model is pre-trained by professional personnel and adjusted based on historical data and industry experience to ensure the effectiveness and accuracy of the model in practical applications; and the K-means clustering method stored in the preset clustering model is used to cluster the data set, which is simple and efficient and can quickly process large-scale data sets; K-means clustering is a commonly used unsupervised learning algorithm for dividing a data set into pre-defined K clusters; the goal of the algorithm is to divide data points into K clusters so that each data point belongs to the cluster represented by the closest cluster center, and the data points within the cluster are as similar as possible, while the data points between different clusters may not be similar;

[0111] Cluster analysis can help evaluate the effectiveness of advertising in each sub-cycle on the target APP; if the second ad click coefficient, the second ad new customer acquisition coefficient and the second ad conversion coefficient of the sub-cycle are larger, then the sub-cycle advertising is more effective on the target APP; conversely, if the second ad click coefficient, the second ad new customer acquisition coefficient and the second ad conversion coefficient of the sub-cycle are smaller, then the sub-cycle advertising is less effective on the target APP; through cluster analysis, we can understand the effectiveness of advertising in each sub-cycle on the target APP, thereby determining the effectiveness of advertising in the first cycle on the target APP, and adjust the duration and frequency of the next advertising cycle on the target APP based on the effectiveness of advertising in the first cycle on the target APP.

[0112] Therefore, by taking the second advertisement click coefficient, the second advertisement attracting new customers coefficient and the second advertisement conversion coefficient of each sub-period as the data points of each sub-period, a data set is obtained, and the data set is clustered to determine the advertising delivery effect index of each sub-period; and the second advertisement click coefficient, the second advertisement attracting new customers coefficient and the second advertisement conversion coefficient of the sub-period present a common linear relationship, that is, the larger the second advertisement click coefficient, the larger the second advertisement attracting new customers coefficient and the larger the second advertisement conversion coefficient of the sub-period, the larger the advertising delivery effect index of the corresponding sub-period.

[0113] It should be noted that step 1: using the elbow method to determine the optimal number of clusters for the data set; the elbow method is one of the common methods used to determine the optimal number of clusters K in K-means clustering. The basic idea is to find an "elbow point" by observing the relationship between the number of clusters K and the total internal deviation square sum (WCSS) of the clustering results. The K value corresponding to the "elbow point" is the position where the WCSS decreases significantly after the number of clusters increases to a certain extent. WCSS refers to the sum of the squares of the distances from each data point to the center of the cluster to which it belongs, which indicates the compactness of the clustering results.

[0114] Step 2: According to the optimal number of clusters, randomly select the corresponding number of data points from the data points as the initial cluster centers of clustering. The initial cluster centers refer to the K cluster centers randomly selected before starting the clustering process. These initial cluster centers are the starting points of the algorithm.

[0115] Step 3: For each data point in the data set, calculate the distance between it and each cluster center using the Euclidean distance. For example, calculate the distance between data point s i and the initial cluster center e j The distance between i , e j ), and its calculation formula is And assign the data point to the cluster to which the nearest cluster center belongs. Euclidean distance refers to the distance between two points in Euclidean space, and is also one of the most commonly used distance metrics. In K-means clustering, Euclidean distance is usually used to calculate the distance between data points and cluster centers.

[0116] Step 4: For each cluster, calculate the mean of all its data points and use the mean as the new cluster center. The calculation formula is: where k j Represented as the set of data points in the jth cluster, s′ i Represented as the new cluster center of the jth cluster;

[0117] Step 5: Repeat steps 3 and 4 until the cluster center no longer changes, and obtain the final cluster and the final cluster center.

[0118] It should be specifically explained that the step of assigning a data point to the cluster to which the nearest cluster center belongs is to sort the distances between the data point and each cluster center from large to small, and select the smallest distance to assign the data point to the cluster to which it belongs.

[0119] In one embodiment, in one embodiment, the final clusters and the final cluster centers are mapped to three-dimensional coordinates to obtain the advertising effect index of each sub-period, including:

[0120] Calculate the weight coefficient of each final cluster, the calculation formula is: Where W j It is expressed as the weight coefficient of the jth final cluster, H 总 Represented as the number of all data points in the data set, H j Represents the number of data points in the jth final cluster;

[0121] Calculate the distance of each data point in each final cluster to the coordinate origin in three-dimensional coordinates to obtain the preliminary coordinate distance of the data point, and multiply the preliminary coordinate distance of the data point by the weight coefficient corresponding to the final cluster in which the data point is located to obtain the actual coordinate distance of the data point;

[0122] The actual coordinate distance of each data point is used as the advertising effect index of the corresponding sub-period.

[0123] In one implementation, according to the above method, the advertising effect index of each sub-period obtained is more accurate and can better reflect the advertising effect of each sub-period on the target APP.

[0124] It should be noted that because the second ad click coefficient, the second ad new customer acquisition coefficient and the second ad conversion coefficient are selected as data points for clustering, these three data points can now be used as coordinates and mapped to the three-dimensional coordinate axis, and each sub-period will be mapped to a point on the three-dimensional coordinate axis; after clustering is completed, all data points will be distributed in different final clustering clusters on the three-dimensional coordinate axis. The closer the final clustering cluster is to the original center and the sub-period corresponding to the data points included therein, the larger the advertising delivery effect index on the target APP, indicating that the advertising delivery effect is better, and vice versa; at the same time, the weight coefficient of each final clustering cluster is calculated to correct the distance from the data point to the original center. In this way, the different importance of each final clustering cluster can be ensured, and at the same time, the influence of noise and outliers on the clustering results can be reduced to a certain extent, so that the advertising delivery effect index of each cycle calculated in the end is more accurate.

[0125] In one implementation, the advertising effect index of each sub-period on the target APP calculated by the above method has the advantages of multi-dimensional analysis, identification of group characteristics, improved accuracy, visual display and dynamic adjustment; this method can comprehensively evaluate the advertising effect, reduce the impact of noise and outliers, so that the advertising effect index finally calculated can more accurately reflect the advertising effect in the first period.

[0126] In one embodiment, a time-advertising delivery effect line graph is drawn according to the advertising delivery effect index of each sub-period, and the trend state of the advertising delivery effect of the first period is divided according to the line graph, including:

[0127] Arrange the advertising effect index of each sub-period based on the time sequence to obtain an advertising effect index sequence, and draw a time-advertising effect line graph based on the advertising effect index sequence;

[0128] Calculate the slope of every two adjacent sub-periods based on the time-advertising effect line graph, and calculate the mean of all slopes to obtain the overall average slope K;

[0129] When the overall average slope K = 0, it means that the trend state of the advertising effect in the first cycle is a stable trend;

[0130] When the overall average slope K>0, it means that the trend state of the advertising effect in the first cycle is an upward trend;

[0131] When the overall average slope K<0, it means that the trend state of the advertising effect in the first cycle is a downward trend.

[0132] It should be noted that a variety of data visualization tools and libraries can be used to draw a time-advertising effect line graph, such as Matplotlib or Seaborn in Python, ggPlot2 in R language, and commercial software such as Excel and Tableau. These tools provide a wealth of drawing functions, allowing users to visualize the advertising effect index and time series data, and generate intuitive line graphs through simple code or interface operations, thereby helping to analyze the changing trend of advertising effects. Using these tools, you can not only quickly draw the required charts, but also facilitate further analysis and customized settings of the data; of course, you can also choose other drawing methods, depending on the actual situation, without limitation or elaboration; in addition, in the time-advertising effect line graph, the horizontal axis is the time series, which is the time point of each cycle; the vertical axis represents the advertising effect index.

[0133] It should be noted that when the overall average slope is close to zero, it means that the advertising effect remains relatively stable in the first cycle, and it shows a steady trend at this time; when the overall average slope is greater than zero, it means that the advertising effect is constantly improving and the advertising strategy is relatively successful, and it shows an upward trend at this time; when the overall average slope is less than zero, it means that the advertising effect has declined, and it shows a downward trend at this time.

[0134] In one embodiment, when the trend state of the advertising delivery effect in the first cycle is a stable trend, the delivery cycle length and delivery frequency of the next cycle on the APP are consistent with those of the first cycle;

[0135] When the trend state of the advertising effect in the first cycle is a downward trend, the advertising on the APP is stopped.

[0136] It should be noted that when the trend of the advertising effect in the first cycle is a stable trend, it means that the effect of advertising on the APP remains stable without significant growth or decline. In this case, maintaining the same duration and frequency of the advertising cycle can ensure continuous exposure and brand awareness, while observing whether the advertising effect will be further improved in a stable state; when the trend of the advertising effect in the first cycle is a downward trend, it means that the effectiveness of the advertisement is weakening, which may lead to waste of resources and loss of budget; in this case, stopping advertising can avoid further budget losses and free up resources for the company or brand to make more effective market strategy adjustments or to place ads on more potential platforms; such decisions can ensure the efficiency and effectiveness of advertising, help companies better manage marketing budgets, and achieve higher returns on investment.

[0137] In one embodiment, when the trend state of the advertising delivery effect in the first cycle is an upward trend, increasing the duration and delivery frequency of the advertisement in the next delivery cycle on the target APP includes:

[0138] Calculate the average of the advertising effect index of each sub-period in the first period as the adjustment coefficient. Adjust the duration and frequency of the advertising cycle on and off the APP according to the adjustment coefficient. The calculation formula is:

[0139] TRQ=TRE×(1+wk)

[0140] FEQ=FED×(1+wk)

[0141] Where TRQ is the duration of the new delivery cycle, TRE is the duration of the initial delivery cycle, FEQ is the new delivery frequency, FED is the initial delivery frequency, and wk is the adjustment coefficient.

[0142] It should be noted that the duration of the initial delivery cycle and the initial delivery frequency are the duration and delivery frequency of the first delivery cycle, which can be directly obtained through the background data of the target APP for advertising delivery; when the trend state of the advertising delivery effect of the first cycle is an upward trend, it means that the advertising delivery effect of the first cycle is very good. In order to maximize the advertising effect, the duration and delivery frequency of the next delivery cycle of the advertisement on the target APP should be appropriately increased to ensure that the advertisement can maintain a high exposure rate among users and further enhance brand awareness and user engagement.

[0143] Based on the same inventive concept, the embodiment of the present invention also provides an advertising analysis system based on AI intelligence. Figure 2 , Figure 2 A framework diagram of an AI-based advertising analysis system provided in an embodiment of the present invention, the system comprising:

[0144] Data acquisition module: for the target APP, obtain the most recent cycle in which the advertisement is placed on the target APP as the first cycle, divide the first cycle into several sub-cycles according to the preset time window, and extract the advertisement click-through rate, number of new users and number of purchasing users in each sub-cycle;

[0145] Cluster analysis module: Determine the data points of each sub-period according to the advertisement click-through rate, the number of new users and the number of purchasing users, obtain the data set, cluster the data set through the preset clustering model to determine the advertising effect index of each sub-period;

[0146] Trend judgment module: Draw a time-advertising effect line graph based on the advertising effect index of each sub-period, and divide the trend status of the advertising effect of the first period according to the line graph; the trend status includes downward trend, stable trend and upward trend;

[0147] Adjustment module: When the trend of the advertising effect in the first cycle is an upward trend, increase the duration and frequency of the next advertising cycle on the target APP.

[0148] An AI-based advertising analysis system provided in an embodiment of the present invention can flexibly change the frequency and duration of the advertising delivery cycle according to the actual advertising delivery effect of the APP during the advertising delivery process through the above-mentioned method, thereby reducing the waste of budget and maximizing the advertising effect.

[0149] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. An advertising analysis method based on AI intelligence, characterized in that: The following steps are involved: For the target APP, obtain the most recent cycle in which the advertisement is delivered on the target APP as the first cycle, divide the first cycle into several sub-cycles according to the preset time window, and extract the advertisement click-through rate, number of new users and number of purchasing users in each sub-cycle; Determine the data points of each sub-period according to the advertisement click-through rate, the number of new users and the number of purchasing users to obtain a data set, and cluster the data set using a preset clustering model to determine the advertisement delivery effect index of each sub-period; Draw a time-advertising effect line graph according to the advertising effect index of each sub-period, and divide the trend state of the advertising effect of the first period according to the line graph; the trend state includes a downward trend, a stable trend and an upward trend; When the trend of the advertising effect in the first cycle is an upward trend, increase the duration and frequency of the next advertising cycle on the target APP.

2. The advertising analysis method based on AI intelligence according to claim 1 is characterized in that: The data points for each sub-period are determined based on the ad click-through rate, the number of new users, and the number of purchasing users. They include: Get the total number of times users click on ads, the total number of times ads are displayed, the average time users stay on the page after clicking ads, and the total number of times users interact with ads in each sub-period, and get the first ad click coefficient of the corresponding sub-period. The calculation formula is: In the formula, Esa is the first ad click coefficient, gt, ht, sd, and ku are the total number of times users click on ads, the total number of times ads are displayed, the average time users stay on the page after clicking ads, and the total number of times users interact with ads. is the user click-through rate, ω1, ω2, and ω3 are respectively the preset proportional coefficients of the user click-through rate, the average time a user stays on the page after clicking on the ad, and the total number of times a user interacts with the ad; The second advertisement click coefficient is obtained according to the first advertisement click coefficient of each sub-period and the weight value of the preset advertisement click coefficient of the corresponding sub-period in the advertisement delivery effect. The calculation formula is: Esx = T1 × Esa; wherein Esx is the second advertisement click coefficient, and T1 is the weight value of the preset advertisement click coefficient of the sub-period in the advertisement delivery effect; The data point of each sub-period is determined according to the second advertisement click coefficient, the number of newly added users and the number of purchasing users in each sub-period.

3. The advertising analysis method based on AI intelligence according to claim 2 is characterized in that: The data points for each sub-period are determined based on the second advertisement click coefficient, the number of new users, and the number of purchasing users in each sub-period, including: Obtain the total number of new users registered through the ads delivered, the total number of ad impressions, the user retention rate, and the average time required for new users to convert from being exposed to ads to becoming users in each sub-period, and obtain the first ad acquisition coefficient of the corresponding sub-period. The calculation formula is: Where Was is the coefficient of attracting new users for the first advertisement, rf, ht, re, and es are the total number of new users, the total number of advertisement displays, the user retention rate, and the average time required for new users to convert from being exposed to advertisements to being users. is the user addition rate, a1, a2, and a3 are the preset proportional coefficients of the user addition rate, user retention rate, and the average time required for new users to convert from being exposed to advertisements to being users; The second advertising new customer acquisition coefficient is obtained according to the first advertising new customer acquisition coefficient of each sub-period and the weight value of the preset advertising new customer acquisition coefficient of the corresponding sub-period in the advertising delivery effect. The calculation formula is: Wds = T2 × Was; where Wds is the second advertising new customer acquisition coefficient, and T2 is the weight value of the preset advertising new customer acquisition coefficient of the sub-period in the advertising delivery effect; The data point of each sub-period is determined based on the second advertisement click coefficient, the second advertisement new user acquisition coefficient and the number of purchasing users of each sub-period.

4. The AI-based advertising analysis method according to claim 3, characterized in that: The data points of each sub-period are determined based on the second advertisement click coefficient, the second advertisement new user acquisition coefficient and the number of purchasing users in each sub-period, including: Get the total number of users who purchased through ads, the total number of ad impressions, the average user purchase frequency, and the average order value in each sub-period, and get the first ad conversion coefficient of the corresponding sub-period. The calculation formula is: In the formula, Rds is the first advertisement conversion coefficient, kl, ht, nv, and py are the total number of users who purchased through advertisements, the total number of advertisement displays, the average purchase frequency of users, and the average order value, respectively. Purchase user increase rate, b1, b2, b3 are preset proportional coefficients of user purchase user increase rate, user average purchase frequency and average order value respectively; The second advertising conversion coefficient is obtained according to the first advertising conversion coefficient of each sub-period and the weight value of the preset advertising conversion coefficient of the corresponding sub-period in the advertising delivery effect, and the calculation formula is: Rcs = T3 × Rfs; where Rcs is the second advertising conversion coefficient, and T3 is the weight value of the preset advertising conversion coefficient of the sub-period in the advertising delivery effect; The second advertisement click coefficient, the second advertisement new customer acquisition coefficient and the second advertisement conversion coefficient of each sub-period are used as data points of each sub-period.

5. The advertising analysis method based on AI intelligence according to claim 1 is characterized in that: Clustering the data set using a preset clustering model to determine the advertising delivery effect index of each sub-period includes: The steps for clustering the data set according to the K-means clustering method stored in the preset clustering model are: Step 1: Used to determine the optimal number of clusters for the data set using the elbow method; Step 2: Randomly select a corresponding number of data points as initial clusters according to the optimal number of clusters; Step 3: For each data point, use the Euclidean distance algorithm to calculate the distance between the data point and each initial cluster center, and assign the data point to the cluster to which the nearest cluster center belongs. After traversing all data points, the initial cluster center is obtained. Step 4: Calculate the average value of all data points for each initial cluster and use the average value as the new cluster center; Step 5: It is used to determine whether the cluster center has changed. If it has changed, repeat steps 3 and 4 until the cluster center no longer changes, and obtain the final cluster and the final cluster center; The final clusters and the final cluster centers are mapped to three-dimensional coordinates to obtain the advertising effect index of each sub-period.

6. The advertising analysis method based on AI intelligence according to claim 5 is characterized in that: The final clusters and the final cluster centers are mapped to three-dimensional coordinates, and the advertising effect indexes of each sub-period are obtained, including: Calculate the weight coefficient of each final cluster, the calculation formula is: Where W j It is expressed as the weight coefficient of the jth final cluster, H 总 Represented as the number of all data points in the data set, H j Represents the number of data points in the jth final cluster; Calculate the distance of each data point in each final cluster to the coordinate origin in three-dimensional coordinates to obtain the preliminary coordinate distance of the data point, and multiply the preliminary coordinate distance of the data point by the weight coefficient corresponding to the final cluster in which the data point is located to obtain the actual coordinate distance of the data point; The actual coordinate distance of each data point is used as the advertising effect index of the corresponding sub-period.

7. The advertising analysis method based on AI intelligence according to claim 1 is characterized in that: Draw a time-advertising effect line graph based on the advertising effect index of each sub-period, and divide the trend status of the advertising effect of the first period into the following according to the line graph: Arrange the advertising effect index of each sub-period based on the time sequence to obtain an advertising effect index sequence, and draw a time-advertising effect line graph based on the advertising effect index sequence; Calculate the slope of every two adjacent sub-periods based on the time-advertising effect line graph, and calculate the mean of all slopes to obtain the overall average slope K; When the overall average slope K = 0, it means that the trend state of the advertising effect in the first cycle is a stable trend; When the overall average slope K>0, it means that the trend state of the advertising effect in the first cycle is an upward trend; When the overall average slope K<0, it means that the trend state of the advertising effect in the first cycle is a downward trend.

8. The AI-based advertising analysis method according to claim 7, characterized in that: When the trend of the advertising effect in the first cycle is a stable trend, the duration and frequency of the next cycle on the APP will be the same as the first cycle; When the trend state of the advertising effect in the first cycle is a downward trend, the advertising on the APP is stopped.

9. The AI-based advertising analysis method according to claim 1, characterized in that: When the trend of the advertising effect in the first cycle is an upward trend, the duration and frequency of the next advertising cycle on the target APP are adjusted as follows: Calculate the average of the advertising effect index of each sub-period in the first period as the adjustment coefficient. Adjust the time and frequency of the advertising cycle on and off the APP according to the adjustment coefficient. The calculation formula is: TRQ=TRE×(1+wk) FEQ=FED×(1+wk) Where TRQ is the time of the new delivery cycle, TRE is the time of the initial delivery cycle, FEQ is the new delivery frequency, FED is the initial delivery frequency, and wk is the adjustment coefficient.

10. An AI-based advertising analysis system, characterized in that: The system comprises: Data acquisition module: for the target APP, obtain the most recent cycle in which the advertisement is placed on the target APP as the first cycle, divide the first cycle into several sub-cycles according to the preset time window, and extract the advertisement click-through rate, number of new users and number of purchasing users in each sub-cycle; Cluster analysis module: determine the data points of each sub-period according to the advertisement click-through rate, the number of new users and the number of purchasing users, obtain a data set, and cluster the data set through a preset clustering model to determine the advertisement delivery effect index of each sub-period; Trend judgment module: draw a time-advertising effect line graph according to the advertising effect index of each sub-period, and divide the trend state of the advertising effect of the first period according to the line graph; the trend state includes a downward trend, a stable trend and an upward trend; Adjustment module: When the trend of the advertising effect in the first cycle is an upward trend, increase the duration and frequency of the next advertising cycle on the target APP.

Images