E-commerce precise advertisement putting method based on big data mining

Abstract

An e-commerce precise advertisement putting method based on big data mining comprises the following steps: collecting user data: collecting multi-dimensional data of a user through an e-commerce platform, including but not limited to browsing history, search history, purchase record, social media interaction, geographical location information and the like; all data should be encrypted and desensitized to ensure data privacy security, modeling is performed on features such as interest, purchasing power and purchasing cycle of a user based on user data by applying clustering analysis, association rule mining and other big data analysis technologies, a dynamically updated user portrait is formed, and a personalized user tag is established on the basis, so that the user experience is improved. , such as price sensitivity, product preference and the like. A differential privacy technology is adopted to protect user data, and the personal privacy of a user is ensured not to be leaked through a noise injection and encryption algorithm. Even in the big data analysis process, the sensitive information of the user data cannot be restored, so that the trust of the user to the platform is greatly enhanced, and the risk of data leakage or abuse is avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet e-commerce, specifically an e-commerce precise advertisement placement method based on big data mining. BACKGROUND

[0002] The e-commerce precise advertisement placement method based on big data mining mainly analyzes user historical behavior data, consumption preferences, interest tags, etc., and combines real-time market dynamics and social data to construct a precise user portrait. Through data mining techniques such as clustering analysis, association rule analysis, and machine learning, the system can accurately identify potential user groups and their needs, and push the most relevant advertisement content to each user. This method can significantly improve the conversion rate of advertisement placement, reduce advertisement waste, and improve the marketing effectiveness of e-commerce platforms. At the same time, combined with A / B testing and effect evaluation models, advertisers can optimize placement strategies in real time to achieve more personalized and intelligent advertisement promotion.

[0003] While big data mining-based e-commerce precise advertising placement methods have significant advantages in improving advertising effectiveness and user experience, they also have some drawbacks and challenges: privacy and data security issues: big data mining requires the collection and processing of large amounts of user data, including personal information, browsing history, purchase behavior, etc., which can easily lead to user privacy leakage and data security risks. If data processing is not done properly, it may lead to data leakage, misuse or hacking, affecting user trust. Data quality and accuracy: the quality of data directly affects the effectiveness of advertising placement. If the data source is unreliable or contains errors, the user portrait mined may not be accurate, resulting in a significant reduction in the effectiveness of advertising placement. For example, some user data may be outdated or incomplete, affecting the accuracy of the prediction. Ad fatigue caused by over-personalization: while precise advertising placement can improve relevance, if it relies too much on user portraits, users may frequently receive similar or repetitive ads, leading to ad fatigue and even negative emotions. This not only reduces the effectiveness of advertising, but also may lead to user loss. Algorithmic bias and unfairness: data mining algorithms may produce bias in certain specific groups when analyzing user data. For example, the algorithm may prefer to recommend certain specific products or brands, ignoring the needs of niche users, resulting in unfair advertising or discrimination against certain user groups. High technology and resource costs: big data-based advertising placement methods require strong data storage, computing and analysis capabilities. For small and medium-sized e-commerce platforms, it may be difficult to afford the high cost of infrastructure. In addition, the development and maintenance of related algorithm models also require a large amount of technical investment. Real-time and dynamic adjustment difficulties: while big data analysis can provide accurate predictions, advertising placement strategies need to be quickly adjusted as market and user needs change. How to maintain efficient real-time response and optimization in a complex and changing environment is still a technical challenge. Legal and regulatory risks: with the increasing strictness of data privacy regulations (such as GDPR), e-commerce platforms must comply with relevant laws when conducting precise advertising placement to avoid illegal collection and use of user data. Otherwise, they may face legal lawsuits and heavy fines, affecting brand image and operation.

[0004] Therefore, we propose a big data mining-based e-commerce precise advertising placement method. SUMMARY

[0005] To achieve the above purpose, the present application provides the following technical scheme: a big data mining-based e-commerce precise advertising placement method, comprising the following steps:

[0006] 1.1 Collect user data: collect multi-dimensional data of users through e-commerce platforms, including but not limited to browsing history, search history, purchase records, social media interactions, geographic location information, etc. All data should be encrypted and desensitized to ensure data privacy and security.

[0007] 1.2 User Profiling: Based on user data, apply clustering analysis, association rule mining, and other big data analysis techniques to model user interests, purchasing power, and purchasing cycles, etc. to form a dynamic and updated user profile. On this basis, establish personalized user tags such as price sensitivity and product preferences.

[0008] 1.3 Differential Privacy Protection Technology: During the collection and analysis of user data, introduce differential privacy protection mechanisms to perturb sensitive data, ensuring that even in the process of data analysis, the privacy of individual users is not compromised.

[0009] 1.4 Advertisement Recommendation Algorithm: Based on the constructed user profile, use intelligent recommendation algorithms such as deep learning, collaborative filtering, and neural networks to select the most suitable advertisement content from the advertisement library. This process should consider the user's immediate needs and potential interests to ensure the real-time and relevance of the advertisement content.

[0010] 1.5 Advertisement Fatigue Monitoring: Monitor user interactions with advertisements (such as clicks, views, dwell time, and jumps) to calculate advertisement fatigue and dynamically adjust the frequency and type of advertisement push to avoid repeated push of the same advertisement. If a certain advertisement has weak response, immediately adjust the push strategy.

[0011] 1.6 Fairness Evaluation Model: Monitor the effects of advertisement placement in real time and establish a fairness evaluation model to analyze whether advertisements are excessively concentrated in certain user groups or brands. If there is advertisement bias, correct it in time to ensure fair exposure opportunities for different brands and groups.

[0012] 1.7 Advertisement Placement Optimization Feedback Mechanism: Collect user interaction data (such as click-through rate, conversion rate, and social sharing) in real time, use A / B testing and other methods to evaluate advertisement effectiveness, and adjust advertisement push strategies based on the results to form a continuous optimization loop.

[0013] Preferably, the user profile is updated in real time based on the user's dynamic behavior changes (such as seasonal purchasing preferences and activity participation) during the construction process, maintaining the high relevance of advertisement recommendations.

[0014] Preferably, the differential privacy protection mechanism uses the Laplace noise injection algorithm to ensure the accuracy of data statistical analysis while avoiding the disclosure of individual user information.

[0015] Preferably, the advertisement fatigue monitoring includes calculating the number of advertisements received by the user within a certain time period and the change in click-through rate. If the click-through rate of a certain advertisement is lower than the preset threshold within a certain time period, the push frequency of that advertisement is automatically reduced to avoid excessive exposure.

[0016] Preferably, the fairness evaluation model includes multi-dimensional evaluation based on advertisement exposure distribution, advertisement conversion rate, user group characteristics, etc., and dynamically adjusts the advertisement delivery strategy through an optimization algorithm to ensure fair exposure of different brands, products, and groups.

[0017] Preferably, the advertisement delivery optimization feedback mechanism further includes comparative analysis based on historical data and real-time data, prediction of the potential conversion effect of the advertisement using a machine learning algorithm, and automatic adjustment of the content and display mode of the advertisement.

[0018] Compared with the prior art, the present application provides an e-commerce precise advertisement delivery method based on big data mining, which has the following beneficial effects:

[0019] 1. The e-commerce precise advertisement delivery method based on big data mining protects user data using differential privacy technology, ensures user privacy through noise injection and encryption algorithms, and greatly enhances user trust in the platform, avoiding the risk of data leakage or misuse.

[0020] 2. The e-commerce precise advertisement delivery method based on big data mining can push advertisements according to the user's immediate needs and potential interests by constructing a dynamically updated user portrait and combining deep learning and intelligent recommendation algorithms. The system not only identifies long-term user preferences, but also responds to changes in user behavior in real time, providing highly personalized advertising content and significantly improving the relevance and click-through rate of advertisements.

[0021] 3. The e-commerce precise advertisement delivery method based on big data mining can precisely monitor the interaction between advertisements and users by introducing an advertisement fatigue monitoring mechanism, dynamically adjusting the frequency and content of advertisements during the advertisement pushing process, avoiding repeated display of the same advertisement, and reducing user's aversion to overexposure, thereby maintaining the effectiveness of the advertisement.

[0022] 4. The e-commerce precise advertisement delivery method based on big data mining establishes a fairness evaluation model to evaluate whether there is bias or excessive concentration in advertisement delivery in real time. By continuously optimizing the exposure distribution and conversion rate of advertisements, it ensures fair competition among different advertisers, brands, and user groups, and avoids the unfairness or bias of the algorithm.

[0023] 5. The e-commerce precise advertisement delivery method based on big data mining can continuously optimize the advertisement delivery strategy by collecting user feedback data in real time and combining A / B testing mechanism. The system continuously compares and analyzes historical data and real-time data to predict the potential effect of the advertisement, thereby realizing automatic adjustment of the advertisement pushing strategy and further improving the effect of the advertisement delivery and the marketing return of the platform. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work are within the protection scope of the present application.

[0025] Embodiments

[0026] Embodiments of an e-commerce precise advertisement placement method based on big data mining

[0027] The e-commerce precise advertisement placement method based on big data mining comprises the following steps:

[0028] 1.1 Collecting user data: Collect multi-dimensional data of users through e-commerce platforms, including but not limited to browsing history, search history, purchase records, social media interactions, geographic location information, etc. All data should be processed through encryption and desensitization to ensure data privacy and security.

[0029] 1.2 Building user portraits: Based on user data, apply clustering analysis, association rule mining and other big data analysis techniques to model user's interests, purchasing power, purchase cycle and other characteristics, form dynamic updated user portraits, and establish personalized user tags such as price sensitivity and product preference on this basis.

[0030] 1.3 Differential privacy protection technology: In the process of collecting and analyzing user data, introduce differential privacy protection mechanism to perturb sensitive data, ensuring that even in the process of data analysis, the private information of a single user will not be leaked.

[0031] 1.4 Advertisement recommendation algorithm: Based on the constructed user portraits, select the most suitable advertisement content for users from the advertisement library through intelligent recommendation algorithms such as deep learning, collaborative filtering and neural networks. This process should consider the user's immediate needs and potential interests to ensure the real-time and relevance of the advertisement content.

[0032] 1.5 Advertisement fatigue monitoring: By monitoring user interactions with advertisements (such as clicks, views, dwell time, jumps, etc.), calculate the advertisement fatigue, dynamically adjust the frequency and types of advertisement push, and avoid repeated push of the same advertisement. If a certain advertisement has weak response, adjust the push strategy immediately.

[0033] 1.6 Fairness evaluation model: Through real-time monitoring of the effect of advertisement placement, establish a fairness evaluation model to analyze whether the advertisement placement is excessively concentrated in certain user groups or brands. If there is advertisement bias, correct it in time to ensure fair exposure opportunities for different brands and different groups.

[0034] 1.7 Advertisement delivery optimization feedback mechanism: By collecting user's advertisement interaction data (such as click-through rate, conversion rate, social sharing, etc.) in real time, using A / B testing and other methods to evaluate the effect of the advertisement, and adjusting the advertisement pushing strategy according to the effect, a continuous optimization loop is formed.

[0035] Specifically, during the construction of the user portrait, the user's dynamic behavior changes (such as seasonal purchase preferences, activity participation, etc.) are updated in real time to maintain the high relevance of the advertisement recommendation.

[0036] Specifically, the differential privacy protection mechanism uses Laplace noise injection algorithm, which can ensure the accuracy of data statistical analysis while avoiding the leakage of individual user information.

[0037] Specifically, the advertisement fatigue monitoring includes calculating the number of advertisements received by the user within a certain period of time and the change of click-through rate, if the click-through rate of a certain advertisement is lower than the preset threshold within a certain period of time, the pushing frequency of the advertisement will be automatically reduced to avoid overexposure.

[0038] Specifically, the fairness evaluation model includes multi-dimensional evaluation based on advertisement exposure distribution, advertisement conversion rate, user group characteristics, etc., and dynamically adjusts the advertisement delivery strategy through optimization algorithm to ensure fair exposure of different brands, products and groups.

[0039] Specifically, the advertisement delivery optimization feedback mechanism further includes comparative analysis based on historical data and real-time data, using machine learning algorithms to predict the potential conversion effect of the advertisement, and automatically adjusting the content and display mode of the advertisement.

[0040] Through the technical scheme, in the application, the differential privacy technology is used to protect user data, and noise injection and encryption algorithm are used to ensure that the personal privacy of users is not leaked. Even in the process of big data analysis, the sensitive information of user data cannot be restored, which greatly enhances the trust of users to the platform and avoids the risk of data leakage or abuse. By constructing a dynamically updated user portrait and combining deep learning and intelligent recommendation algorithm, the application can push advertisements according to the immediate needs and potential interests of users. The system can not only identify the long-term preferences of users, but also respond to changes in user behavior in real time, provide highly personalized advertising content, and significantly improve the relevance and click rate of advertisements. By introducing an advertisement fatigue monitoring mechanism, the application can accurately monitor the interaction between advertisements and users, dynamically adjust the frequency and content during the advertisement pushing process, avoid repeated display of the same advertisement, and reduce the user's aversion caused by excessive exposure, thereby maintaining the effectiveness of the advertisement. By establishing a fairness evaluation model, it can evaluate whether there is bias or excessive concentration in the advertisement placement in real time. By continuously optimizing the exposure distribution and conversion rate of advertisements, it ensures fair competition among different advertisers, brands and user groups, and avoids the unfairness or bias of the algorithm. By collecting user feedback data in real time and combining A / B testing mechanism, it can continuously optimize the advertisement placement strategy. The system continuously compares and analyzes historical data and real-time data to predict the potential effect of the advertisement, so as to realize the automatic adjustment of the advertisement pushing strategy and further improve the effect of the advertisement placement and the marketing return of the platform.

[0041] User portrait construction and advertisement recommendation

[0042] User data collection and privacy protection: The e-commerce platform collects user browsing history, search history, purchase records and other data through user behavior logs, and encrypts and desensitizes all sensitive data. Use differential privacy technology to perturb user data to ensure that even in the data analysis process, the information of a single user cannot be leaked.

[0043] User portrait construction: The system identifies the user's interest points, purchase frequency, price sensitivity and other characteristics through clustering analysis algorithm, and combines deep learning model to predict the user's potential interest. For example, for users who frequently purchase fashion products, the system will identify them as "fashion experts" and push related advertisements. The user portrait will be dynamically updated according to the user's behavior, such as a user showing interest in purchasing a certain electronic product in a certain time period, the system will adjust the user's interest label in time.

[0044] Advertisement recommendation: Based on user profiling, the system uses collaborative filtering algorithms and deep neural network recommendation algorithms to select the most matching advertisements from the advertisement library according to user interests. For example, the system recommends high-cost-performance product advertisements to a price-sensitive user and the latest clothing advertisements to some fashion users. The display of advertisement content will be dynamically adjusted according to the changes in user interests to ensure the relevance of the advertisement content.

[0045] Advertisement frequency and fatigue monitoring: The system continuously tracks user interactions with advertisements. If the click-through rate of a certain advertisement decreases significantly, the system will reduce its push frequency or replace the advertisement content. For example, if a user's click-through rate on fashion shoe advertisements decreases, the system will adjust to recommend other related product advertisements to prevent fatigue caused by repeated exposure to the same advertisement.

[0046] Fairness evaluation and advertisement optimization

[0047] Fairness evaluation model: The advertisement delivery system regularly evaluates the fairness of advertisement exposure through algorithms, including the time distribution of exposure, exposure opportunities for each brand or advertiser, and the fairness of advertisement conversion rates. If a brand advertisement receives significantly more exposure than other brands, the system will adjust strategies to balance the exposure of different advertisements.

[0048] Advertisement effect feedback and optimization: The system collects effect data of different advertisement contents and push strategies based on A / B testing. By analyzing indicators such as click-through rate and conversion rate, the system automatically optimizes the display order, frequency, and content of advertisements. For example, if an advertiser's advertisement significantly improves conversion rate through A / B testing, the system will automatically push this advertisement to more users to improve the overall effectiveness of the advertisement.

[0049] Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An e-commerce precision advertising method based on big data mining, characterized by comprising the following steps: 1.1 Collection of User Data: Multi-dimensional user data is collected through e-commerce platforms, including but not limited to browsing history, search history, purchase history, social media interactions, geographic location information, etc. All data must be encrypted and desensitized to ensure data privacy and security. 1.2 Build user profiles: Based on user data, apply big data analysis techniques such as cluster analysis and association rule mining to model user characteristics such as interests, purchasing power, and purchase cycle to form a dynamically updated user profile. On this basis, establish personalized user tags, such as price sensitivity and product preferences. 1.3 Differential Privacy Protection Technology: During the collection and analysis of user data, a differential privacy protection mechanism is introduced to perturb sensitive data, ensuring that even during the data analysis process, the private information of a single user will not be leaked. 1.4 Ad Recommendation Algorithm: Based on the constructed user profile, intelligent recommendation algorithms such as deep learning, collaborative filtering, and neural networks are used to select the advertising content that best meets user needs from the ad library. This process should consider the user's immediate needs and potential interests, ensuring the real-time and relevance of advertising content. 1.5 Ad Fatigue Monitoring: By monitoring user interactions with ads (such as clicks, views, dwell time, redirects, etc.), we calculate ad fatigue and dynamically adjust the frequency and types of ad pushes to avoid repeated pushes of the same ad. If an ad receives a weak response, we immediately adjust the push strategy. 1.6 Fairness Assessment Model: By monitoring the effectiveness of advertising in real time, we establish a fairness assessment model to analyze whether advertising is overly concentrated on certain user groups or brands. If advertising bias occurs, we promptly correct it to ensure fair exposure for different brands and groups. 1.7 Advertising delivery optimization feedback mechanism: By collecting users' advertising interaction data (such as click-through rate, conversion rate, social sharing, etc.) in real time, using A / B testing and other methods to evaluate advertising effectiveness, and adjusting advertising push strategies based on the results, a continuous optimization closed loop is formed.

2. The method for precise e-commerce advertising based on big data mining according to claim 1, characterized in that: During the construction process, the user portrait is updated in real time based on the user's dynamic behavioral changes (such as seasonal purchasing preferences, activity participation, etc.) to maintain a high degree of relevance for advertising recommendations.

3. The e-commerce precision advertising method based on big data mining according to claim 1 is characterized by: The differential privacy protection mechanism adopts the Laplace noise injection algorithm, which can avoid leaking users' personal information while ensuring the accuracy of data statistical analysis.

4. The method for precise e-commerce advertising based on big data mining according to claim 1, characterized in that: The advertising fatigue monitoring includes calculating the number of advertisements received by users in a specific time period and the changes in their click-through rate. If the click-through rate of an advertisement is lower than a preset threshold within a certain time period, the push frequency of the advertisement is automatically reduced to avoid overexposure.

5. The e-commerce precise advertising delivery method based on big data mining according to claim 1 is characterized by: The fairness assessment model includes assessments based on multiple dimensions such as advertising exposure distribution, advertising conversion rate, and user group characteristics, and dynamically adjusts advertising delivery strategies through optimization algorithms to ensure fair exposure for different brands, products, and groups.

6. The method for precise e-commerce advertising based on big data mining according to claim 1, characterized in that: The advertising delivery optimization feedback mechanism further includes comparative analysis based on historical data and real-time data, using machine learning algorithms to predict the potential conversion effect of advertisements, and automatically adjusting the content and display of advertisements.