Advertisement putting frequency control method and system and electronic equipment

Abstract

The invention provides an advertisement putting frequency control method and system and electronic equipment. The method comprises the following steps: acquiring user behavior data through a demand side platform and an advertisement transaction platform, and preprocessing the acquired user behavior data; according to the preprocessed user behavior data, dividing corresponding users into a plurality of levels according to advertisement acceptability by adopting a clustering algorithm; different advertisement putting frequencies are set for the users of different levels, and advertisement putting is carried out on the users of each level according to the corresponding advertisement putting frequencies; and in the process of performing advertisement putting on the users of each level, obtaining putting effect reflection data corresponding to the users of each level in real time, and optimizing the advertisement putting frequency of the users of each level by adopting a reinforcement learning algorithm. The system comprises each function module for correspondingly realizing each step. According to the electronic equipment, the method is realized when a processor executes a computer program stored in a memory. According to the invention, the problems of poor user experience and poor advertisement putting effect caused by inaccurate frequency control of an existing advertisement putting frequency control mode can be solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of advertisement delivery optimization, and more specifically, relates to an advertisement delivery frequency control method, system and electronic equipment. Background Art

[0002] With the popularity of Internet advertising, it has become mainstream for advertisers to place advertisements through real-time bidding (RTB) platforms. Advertisers can participate in advertising bidding through multiple demand-side platforms (DSPs) to achieve automated advertising display. However, in the prior art, the frequency control of advertising is usually based on simple rules and lacks in-depth analysis of user behavior and preferences, resulting in some users being over-delivered with advertisements, causing "ad fatigue", thereby reducing advertising effectiveness. In addition, different advertisers have different requirements for advertising frequency control, which further increases the complexity of advertising frequency management.

[0003] Currently, most advertising systems control advertising delivery to users based on preset frequency caps. However, this method cannot dynamically adapt to changes in user behavior, cannot accurately predict users' advertising acceptance, and does not segment and optimize user groups, resulting in reduced effectiveness of advertising delivery.

[0004] The existing advertising frequency control methods have the following major defects:

[0005] Poor user experience: Due to the lack of accurate analysis of user behavior and interests, users are often repeatedly served the same ads, which can easily cause boredom and affect user experience.

[0006] Lack of personalization: Failure to conduct personalized frequency management based on the behaviors of different user groups makes it difficult for advertisers to achieve optimal advertising effects among different user groups.

[0007] Poor advertising effect: Too high an advertising frequency may lead to advertising waste, while too low a frequency may result in missed opportunities, fail to fully stimulate user interest, and affect conversion rate. Summary of the invention

[0008] The purpose of the present invention is to solve the problem that the existing advertising delivery frequency control method leads to poor user experience and poor advertising delivery effect due to inaccurate frequency control.

[0009] In order to achieve the above-mentioned purpose, the present invention proposes a method, system and electronic device for controlling the frequency of advertisement delivery.

[0010] According to a first aspect of the present invention, there is provided a method for controlling the frequency of advertisement delivery, the method comprising the following steps:

[0011] Obtain user behavior data through demand-side platforms and advertising trading platforms, and pre-process the obtained user behavior data;

[0012] Based on the pre-processed user behavior data, a clustering algorithm is used to divide the corresponding users into several levels according to their advertising acceptance;

[0013] Set different ad delivery frequencies for users at different levels, and deliver ads to users at each level according to the corresponding ad delivery frequencies;

[0014] In the process of advertising delivery to users at each level, data reflecting the delivery effect corresponding to each level of users is obtained in real time, and a reinforcement learning algorithm is used to optimize the frequency of advertising delivery to users at each level.

[0015] Optionally, the user behavior data includes browsing data, advertising interaction data and conversion behavior data;

[0016] The browsing data includes the user's browsing history, stay time and number of times the advertisement is clicked on the network;

[0017] The advertising interaction data includes the history of user clicks on advertisements and the frequency of user interaction with the advertising content;

[0018] The conversion behavior data is a record of the user's purchase or registration behavior after an advertisement.

[0019] Optionally, the preprocessing of the acquired user behavior data is specifically as follows:

[0020] Perform feature extraction on the acquired user behavior data to obtain the user's behavior feature vector.

[0021] Optionally, the clustering algorithm is used to divide the corresponding users into several levels according to the advertisement acceptance, including:

[0022] Initialization: Select K initial cluster centers;

[0023] Distance calculation: For each user, calculate the distance between it and all cluster centers. The relevant formula is as follows:

[0024]

[0025] In the above formula, xi is the behavior feature vector of user i, c j is the cluster center, n is the dimension of the behavior feature vector;

[0026] Cluster assignment: assign each user to the cluster center closest to him;

[0027] Update cluster centers: Recalculate the center of each cluster. The relevant formula is as follows:

[0028]

[0029] In the above formula, C j For all users in the jth cluster, c j is the new cluster center;

[0030] The steps of distance calculation, cluster assignment and cluster center update are repeated in sequence until the cluster centers converge or a predetermined number of iterations is reached.

[0031] Optionally, setting different advertisement delivery frequencies for users at different levels is implemented based on the following formula:

[0032] fi(t)=fbase+α·Δri(t)

[0033] In the above formula, f i (t) is the frequency of advertising delivery to users in the i-th level at time t;

[0034] fbas e The basic frequency of advertising delivery;

[0035] α is the adjustment coefficient, which is used to control the magnitude of the change in the frequency of advertising delivery;

[0036] △r i (t) is the rate of change of advertising acceptance of users in the i-th level at time t.

[0037] Optionally, the optimization of the advertisement delivery frequency for users at each level by using a reinforcement learning algorithm is implemented based on the following formula:

[0038]

[0039] In the above formula, Q(s,a) represents the Q value of selecting action a in state s, state s is the current state determined by the user's click rate and conversion rate, and action a is to select different advertising frequencies;

[0040] α is the learning rate, γ is the discount factor, and R is the immediate reward after executing action a in state s;

[0041] is the maximum Q value of the subsequent state s'.

[0042] Optionally, the immediate reward R is expressed as:

[0043] R==β·(CTRnewCTRold)+γ·(ConversionnewConversioold)

[0044] In the above formula, CTR is click-through rate, Conversion is conversion rate, and β and γ are weight factors.

[0045] According to a second aspect of the present invention, there is provided an advertisement delivery frequency control system, the system comprising the following functional modules:

[0046] The data collection module is used to obtain user behavior data through the demand-side platform and the advertising trading platform, and pre-process the obtained user behavior data;

[0047] The clustering and stratification module is used to classify the corresponding users into several levels according to their advertising acceptance based on the pre-processed user behavior data and using a clustering algorithm;

[0048] The advertisement delivery frequency control module is used to set different advertisement delivery frequencies for users of different levels, and deliver advertisements to users of each level according to the corresponding advertisement delivery frequencies;

[0049] The advertising delivery frequency optimization module is used to obtain the delivery effect reflection data corresponding to each level of users in real time during the process of advertising delivery to users at each level, and use the reinforcement learning algorithm to optimize the advertising delivery frequency of users at each level.

[0050] According to a third aspect of the present invention, there is provided an electronic device, comprising a processor and a memory, wherein the processor implements any one of the above-mentioned methods for controlling the frequency of advertisement delivery when executing a computer program stored in the memory.

[0051] The beneficial effects of the present invention are:

[0052] The method for controlling the frequency of advertisement delivery of the present invention firstly obtains user behavior data through a demand-side platform and an advertisement trading platform, and pre-processes the obtained user behavior data; secondly, based on the pre-processed user behavior data, a clustering algorithm is used to divide the corresponding users into several levels according to their advertisement acceptance; thirdly, different advertisement delivery frequencies are set for users of different levels, and advertisement delivery is performed to users of each level according to the corresponding advertisement delivery frequencies; finally, in the process of advertisement delivery to users of each level, delivery effect reflection data corresponding to users of each level is obtained in real time, and a reinforcement learning algorithm is used to optimize the advertisement delivery frequency of users of each level.

[0053] The advertising frequency control method of the present invention stratifies user groups and uses artificial intelligence technology to dynamically optimize advertising frequency based on the behavior patterns and advertising acceptance of different user groups, so that advertisers can maximize the advertising effect within the optimal frequency range and enhance the user's advertising experience.

[0054] The advertising frequency control system and electronic device of the present invention belong to a general inventive concept as well as the above-mentioned advertising frequency control method, and have at least the same beneficial effects as the above-mentioned advertising frequency control method, and the beneficial effects will not be repeated here.

[0055] Other features and advantages of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The present invention may be better understood by referring to the following description taken in conjunction with the accompanying drawings, in which the same or similar reference numerals are used throughout the drawings to represent the same or similar components.

[0057] Figure 1 A flowchart of an implementation method for controlling the frequency of advertisement delivery according to an embodiment of the present invention is shown;

[0058] Figure 2 A structural block diagram of an advertisement delivery frequency control system according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0059] In order to enable those skilled in the art to more fully understand the technical solution of the present invention, the exemplary embodiments of the present invention will be described in more comprehensive and detailed in conjunction with the accompanying drawings below. Obviously, the one or more embodiments of the present invention described below are only one or more of the specific ways in which the technical solution of the present invention can be implemented, and are not exhaustive. It should be understood that other ways belonging to a general inventive concept can be used to implement the technical solution of the present invention, and should not be limited by the exemplary embodiments described. Based on one or more embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work should fall within the scope of protection of the present invention.

[0060] Example: Figure 1 The following is a flowchart showing the implementation of the method for controlling the frequency of advertisement delivery according to an embodiment of the present invention. Figure 1 The advertising delivery frequency control method of the embodiment of the present invention comprises the following steps:

[0061] Step S100: obtaining user behavior data through a demand-side platform (DSP) and an advertising exchange (ADX), and preprocessing the obtained user behavior data;

[0062] Step S200: Based on the pre-processed user behavior data, a clustering algorithm is used to divide the corresponding users into several levels according to their advertising acceptance;

[0063] Step S300: setting different advertisement delivery frequencies for users at different levels, and delivering advertisements to users at each level according to the corresponding advertisement delivery frequencies;

[0064] Step S400: In the process of advertising delivery to users at each level, data reflecting the delivery effect corresponding to users at each level are obtained in real time, and a reinforcement learning algorithm is used to optimize the frequency of advertising delivery to users at each level.

[0065] Further, in step S100 of the embodiment of the present invention, the user behavior data includes browsing data, advertising interaction data and conversion behavior data;

[0066] Browsing data includes the user's browsing history on the Internet, the duration of stay, and the number of times the advertisement is clicked;

[0067] Ad interaction data includes the user’s history of clicking on ads and the frequency of user interaction with ad content;

[0068] Conversion behavior data is the record of users’ purchase or registration behavior after an advertisement.

[0069] Furthermore, in step S100 of the embodiment of the present invention, the preprocessing of the acquired user behavior data is specifically as follows:

[0070] Perform feature extraction on the acquired user behavior data to obtain the user's behavior feature vector.

[0071] Furthermore, in step S200 of the embodiment of the present invention, a clustering algorithm is used to divide the corresponding users into several levels according to the advertising acceptance, including:

[0072] Initialization: Select K initial cluster centers;

[0073] Distance calculation: For each user, calculate the distance between it and all cluster centers. The relevant formula is as follows:

[0074]

[0075] In the above formula, x i is the behavior feature vector of user i, c j is the cluster center, n is the dimension of the behavior feature vector;

[0076] Cluster assignment: assign each user to the cluster center closest to him;

[0077] Update cluster centers: Recalculate the center of each cluster. The relevant formula is as follows:

[0078]

[0079] In the above formula, Cj For all users in the jth cluster, c j is the new cluster center;

[0080] The steps of distance calculation, cluster assignment and cluster center update are repeated in sequence until the cluster centers converge or a predetermined number of iterations is reached.

[0081] Specifically, in the embodiment of the present invention, after obtaining the pre-processed user behavior data, the K-means clustering algorithm is used to divide the corresponding users into several levels, each level represents the different acceptance of the users to the advertisements. For example, the users are divided into the following three levels:

[0082] Highly receptive users: frequently click on ads and have high conversion rates;

[0083] Moderate acceptance users: occasionally click on ads, with a medium conversion rate;

[0084] Low receptive users: have low interest in ads, and low click-through and conversion rates.

[0085] Traditional frequency control of advertising is mainly based on static rules, such as the total number of advertising delivery or time interval, which cannot fully consider the differences in individual user behavior. The embodiment of the present invention uses the K-means clustering algorithm to dynamically stratify users and form multiple levels of user groups based on the user's advertising acceptance, interactive behavior, click history and other characteristics. This precise hierarchical management is one of the core of the embodiment of the present invention. By distinguishing users with high, medium and low advertising acceptance, differentiated advertising can be effectively achieved.

[0086] Furthermore, in step S300 of the embodiment of the present invention, setting different advertisement delivery frequencies for users at different levels is implemented based on the following formula:

[0087] fi(t)=fbase+α·Δri(t)

[0088] In the above formula, f i (t) is the frequency of advertising delivery to users in the i-th level at time t;

[0089] fbas e The basic frequency of advertising delivery;

[0090] α is the adjustment coefficient, which is used to control the magnitude of the change in the frequency of advertising delivery;

[0091] △r i (t) is the rate of change of advertising acceptance of users at the i-th level at time t, that is, the increase or decrease in the click-through rate or conversion rate of advertisements in the recent period.

[0092] Furthermore, in step S300 of the embodiment of the present invention, the frequency of advertisement delivery to users at each level is optimized by using a reinforcement learning algorithm based on the following formula:

[0093]

[0094] In the above formula, Q(s,a) represents the Q value of selecting action a in state s, state s is the current state determined by the user's click rate and conversion rate, and action a is to select different advertising frequencies;

[0095] α is the learning rate, γ is the discount factor, and R is the immediate reward after executing action a in state s;

[0096] is the maximum Q value of the subsequent state s'.

[0097] Furthermore, in the embodiment of the present invention, the instant reward R is expressed as:

[0098] R=β·(CTRnew-CTRold)+γ·(Conversionnew-Conversionold)

[0099] In the above formula, CTR is click-through rate, Conversion is conversion rate, and β and γ are weight factors.

[0100] Specifically, in the embodiment of the present invention, in order to optimize the frequency of advertisement delivery, a reinforcement learning algorithm is used to dynamically adjust the frequency of advertisement delivery. The goal of reinforcement learning is to automatically optimize the advertisement delivery strategy based on real-time feedback from users.

[0101] In reinforcement learning, the system interacts with the environment (user behavior data) to select the optimal frequency of advertising under different conditions to maximize the advertiser's ROI. Its basic elements include:

[0102] State (S): The system determines the current state based on the user's real-time behavior and feedback (click rate, conversion rate);

[0103] Action (A): In each state, the system can select different ad delivery frequencies as actions, such as increasing, decreasing, or maintaining the current frequency;

[0104] Reward (R): After the system takes an action, it is rewarded according to the advertiser’s revenue. The reward can be expressed as:

[0105] R=β·(CTRnew-CTRola)+γ·(Conversionnew-Conversionold)

[0106] In the above formula, CTR is click-through rate, Conversion is conversion rate, and β and γ are weight factors.

[0107] Reinforcement learning can be implemented through the Q-learning algorithm. The Q-learning algorithm selects the optimal action by learning the state-action value function (Q value). Its update formula is as follows:

[0108]

[0109] In the above formula, Q(s,a) represents the Q value of selecting action a in state s, state s is the current state determined by the user's click rate and conversion rate, and action a is to select different advertising frequencies;

[0110] α is the learning rate, γ is the discount factor, and R is the immediate reward after executing action a in state s;

[0111] is the maximum Q value of the subsequent state s'.

[0112] Through the Q-learning algorithm, the system can dynamically adjust the frequency of advertising delivery under different user levels and advertising acceptance conditions to ensure that the effectiveness of each advertisement is maximized.

[0113] Specifically, the embodiment of the present invention introduces a reinforcement learning algorithm to dynamically adjust the frequency of advertising delivery according to real-time feedback (click-through rate, conversion rate). Traditional systems often find it difficult to flexibly adjust the frequency of advertising delivery during the advertising delivery process. The embodiment of the present invention uses the adaptive ability of the reinforcement learning model to enable the system to update the frequency of advertising delivery in real time according to the actual advertising delivery effect and user feedback, thereby ensuring the optimization of the advertising delivery effect.

[0114] The advertising frequency control method of the embodiment of the present invention has achieved significant improvement and optimization in the field of advertising. Compared with the existing advertising frequency control method, the embodiment of the present invention has shown outstanding effects in the accuracy of advertising, the advertiser's return on investment (ROI), user experience, resource conservation and other aspects. The following is a detailed description of the effects and advantages brought by the embodiment of the present invention:

[0115] 1. The accuracy of advertising delivery has been significantly improved:

[0116] The embodiment of the present invention uses the K-means clustering algorithm to analyze user behavior patterns and combines it with the dynamic optimization mechanism of the reinforcement learning algorithm to greatly improve the accuracy of advertising delivery. The system can divide users into multiple levels based on their advertising interaction behaviors, click data, and conversion rates, thereby setting personalized advertising delivery frequencies for users at different levels.

[0117] Traditional systems set the frequency of advertising through simple rules, while the embodiments of the present invention use intelligent clustering algorithms to carefully stratify users, ensuring that advertising is more in line with each user's advertising acceptance. Through the reinforcement learning algorithm, the system can dynamically adjust the frequency based on real-time feedback to ensure that advertisers place ads at the most appropriate frequency, thereby significantly improving the conversion rate. Experimental results show that compared with the traditional fixed frequency control method, the embodiments of the present invention improve the click-through rate (CTR) of advertisements by 15-20% and the conversion rate of advertisements by 10-15%.

[0118] 2. Optimization of user experience and effective prevention and control of ad fatigue:

[0119] The embodiment of the present invention can identify the user's advertising acceptance and prevent and control advertising fatigue by introducing an adaptive learning mechanism. When the system detects that the user's click frequency on the advertisement decreases, the frequency of advertisement delivery will be automatically reduced to avoid the user's boredom caused by excessive delivery.

[0120] By dynamically adjusting the frequency of advertising, the system can effectively reduce the interference of advertising on users, so that users no longer see the same advertisement frequently, and improve the acceptance of advertising. According to user behavior and preferences, the system provides differentiated advertising experiences for different users, further improving users' recognition of advertising content. Experimental data shows that after using the frequency optimization strategy of the embodiment of the present invention, the user's negative feedback rate on advertising (such as closing ads, blocking ads) decreased by 25%, and the user retention rate increased by 10%.

[0121] 3. Significant improvement in advertiser return on investment (ROI):

[0122] Through the intelligent optimization mechanism of the embodiment of the present invention, advertisers can display advertisements to users at the most effective time period and frequency, thereby improving the click-through rate and conversion rate of advertisements and maximizing the advertiser's return on investment (ROI). At the same time, since invalid advertisement displays are reduced, advertising resources are used more efficiently and advertising budgets are spent more reasonably. Relevant data show that after adopting the method of the embodiment of the present invention, the average ROI of advertisers has increased by 20-30% and the advertising cost has decreased by 10-15%.

[0123] 4. Efficient use and conservation of advertising resources:

[0124] The embodiment of the present invention uses intelligent frequency control to ensure that advertisements are displayed to users only at appropriate time periods and frequencies, thereby avoiding the problem of waste of advertising resources in traditional advertising frequency control. In traditional methods, due to the lack of in-depth analysis of user behavior, the advertising frequency is often set too high, resulting in a large number of invalid displays. The embodiment of the present invention dynamically adjusts the frequency through an intelligent algorithm, effectively saving advertising resources. Experimental results show that compared with traditional methods, the embodiment of the present invention can save about 15-25% of advertising resources and reduce the number of invalid displays of advertising by more than 30%.

[0125] 5. Optimization of cross-platform advertising frequency control:

[0126] The embodiments of the present invention can uniformly manage the frequency of advertisements in a multi-platform and multi-device environment, ensuring that users will not be served repeated advertisements on multiple devices or platforms. Traditional advertising systems often lack coordinated control between different platforms, resulting in users frequently receiving the same advertisements on different devices. By integrating data from multiple platforms, the embodiments of the present invention can uniformly manage the frequency of advertisement display for users and prevent the duplication of advertisement frequencies. Relevant data show that the problem of overlapping advertisement frequencies across platforms has been reduced by 40%, and the consistency of advertisement frequencies for users on different devices has been improved by 30%.

[0127] 6. Improvement of data security and privacy protection:

[0128] The embodiment of the present invention adopts advanced data encryption and privacy protection technology to ensure the security and privacy of user data throughout the entire process of advertising. User behavior data and advertising interaction data are encrypted during collection, processing and transmission to prevent data leakage and illegal use. All user behavior data is strictly encrypted during transmission and storage to ensure data security. By anonymizing user data, the system will not expose the user's personal identity information when optimizing advertising frequency, which complies with privacy protection regulations such as GDPR. The privacy protection compliance of the embodiment of the present invention reaches 100%, and the risk of data leakage is reduced by 50%, which complies with the privacy protection regulations of various regions.

[0129] Accordingly, based on the advertisement delivery frequency control method of the embodiment of the present invention, the embodiment of the present invention further proposes an advertisement delivery frequency control system.

[0130] Figure 2 FIG. 1 shows a structural block diagram of an advertisement delivery frequency control system according to an embodiment of the present invention. Figure 2 The advertising delivery frequency control system of the embodiment of the present invention includes the following functional modules:

[0131] The data collection module is used to obtain user behavior data through the demand-side platform and the advertising trading platform, and pre-process the obtained user behavior data;

[0132] The clustering and stratification module is used to classify the corresponding users into several levels according to their advertising acceptance based on the pre-processed user behavior data and using a clustering algorithm;

[0133] The advertisement delivery frequency control module is used to set different advertisement delivery frequencies for users of different levels, and deliver advertisements to users of each level according to the corresponding advertisement delivery frequencies;

[0134] The advertising delivery frequency optimization module is used to obtain the delivery effect reflection data corresponding to each level of users in real time during the process of advertising delivery to users at each level, and use the reinforcement learning algorithm to optimize the advertising delivery frequency of users at each level.

[0135] Accordingly, based on the advertising delivery frequency control method of the embodiment of the present invention, the embodiment of the present invention also proposes an electronic device, which includes a processor and a memory. When the processor executes the computer program stored in the memory, the advertising delivery frequency control method of the embodiment of the present invention is implemented.

[0136] Although one or more embodiments of the present invention are described above, it should be known to those skilled in the art that the present invention can be implemented in any other form without departing from its subject matter and scope. Therefore, the embodiments described above are illustrative rather than restrictive, and many modifications and substitutions are obvious to those skilled in the art without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A method for controlling the frequency of advertisement delivery, characterized in that: include: Obtain user behavior data through demand-side platforms and advertising trading platforms, and pre-process the obtained user behavior data; Based on the pre-processed user behavior data, a clustering algorithm is used to divide the corresponding users into several levels according to their advertising acceptance; Set different ad delivery frequencies for users at different levels, and deliver ads to users at each level according to the corresponding ad delivery frequencies; In the process of advertising delivery to users at each level, data reflecting the delivery effect corresponding to each level of users is obtained in real time, and a reinforcement learning algorithm is used to optimize the frequency of advertising delivery to users at each level.

2. The method for controlling the frequency of advertisement delivery according to claim 1, characterized in that: The user behavior data includes browsing data, advertising interaction data and conversion behavior data; The browsing data includes the user's browsing history, stay time and number of times the advertisement is clicked on the network; The advertising interaction data includes the history of user clicks on advertisements and the frequency of user interaction with the advertising content; The conversion behavior data is a record of the user's purchase or registration behavior after an advertisement.

3. The method for controlling the frequency of advertisement delivery according to claim 1, characterized in that: The preprocessing of the acquired user behavior data is specifically as follows: Perform feature extraction on the acquired user behavior data to obtain the user's behavior feature vector.

4. The method for controlling the frequency of advertisement delivery according to claim 3, characterized in that: The clustering algorithm is used to divide the corresponding users into several levels according to the advertising acceptance, including: Initialization: Select K initial cluster centers; Distance calculation: For each user, calculate the distance between it and all cluster centers. The relevant formula is as follows: In the above formula, x i is the behavior feature vector of user i, c j is the cluster center, n is the dimension of the behavior feature vector; Cluster assignment: assign each user to the cluster center closest to him; Update cluster centers: Recalculate the center of each cluster. The relevant formula is as follows: In the above formula, C j For all users in the jth cluster, c j is the new cluster center; The steps of distance calculation, cluster assignment and cluster center update are repeated in sequence until the cluster centers converge or a predetermined number of iterations is reached.

5. The method for controlling the frequency of advertisement delivery according to claim 1, characterized in that: Setting different advertising delivery frequencies for users at different levels is based on the following formula: f i (t)=f base +α·Δr i (t) In the above formula, f i (t) is the frequency of advertising delivery to users in the i-th level at time t; f base The basic frequency of advertising delivery; α is the adjustment coefficient, which is used to control the magnitude of the change in the frequency of advertising delivery; △r i (t) is the rate of change of advertising acceptance of users in the i-th level at time t.

6. The method for controlling the frequency of advertisement delivery according to claim 1, characterized in that: The optimization of the advertisement delivery frequency of each level of users by using the reinforcement learning algorithm is implemented based on the following formula: In the above formula, Q(s,a) represents the Q value of selecting action a in state s, state s is the current state determined by the user's click rate and conversion rate, and action a is to select different advertising frequencies; α is the learning rate, γ is the discount factor, and R is the immediate reward after executing action a in state s; is the maximum Q value of the subsequent state s′.

7. The method for controlling the frequency of advertisement delivery according to claim 6, characterized in that: The immediate reward R is expressed as: R=β·(CTR new -CTR old )+γ·(Conversion new -Conversion old ) In the above formula, CTR is click-through rate, Conversion is conversion rate, and β and γ are weight factors.

8. An advertisement delivery frequency control system, characterized in that: include: The data collection module is used to obtain user behavior data through the demand-side platform and the advertising trading platform, and pre-process the obtained user behavior data; The clustering and stratification module is used to classify the corresponding users into several levels according to their advertising acceptance based on the pre-processed user behavior data and using a clustering algorithm; The advertisement delivery frequency control module is used to set different advertisement delivery frequencies for users of different levels, and deliver advertisements to users of each level according to the corresponding advertisement delivery frequencies; The advertising delivery frequency optimization module is used to obtain the delivery effect reflection data corresponding to each level of users in real time during the process of advertising delivery to users at each level, and use the reinforcement learning algorithm to optimize the advertising delivery frequency of users at each level.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the processor implements the method for controlling the frequency of advertisement delivery according to any one of claims 1 to 7 when executing a computer program stored in the memory.

Images