Push optimization method and system based on e-commerce big data feedback

Abstract

The invention relates to the technical field of big data, in particular to a push optimization method and system based on e-commerce big data feedback, and the system comprises a data acquisition module, a model construction module, a push data generation module and a negative feedback optimization module. Obtaining multi-source heterogeneous data and a real-time e-commerce data stream, and performing compliance verification on all the obtained data; constructing a corresponding user portrait based on the verified e-commerce data, extracting a corresponding push feedback feature, and constructing a feedback driving model at the same time; inputting the user portrait and the push feedback feature into the feedback driving model, and calculating push data by using output expected feedback data; the negative feedback data of the push data is acquired in real time, and push optimization is performed based on the negative feedback data, so that feedback optimization of the e-commerce push data is realized, and the push conversion rate of the e-commerce data is improved.

Description

Technical Field

[0001] This invention relates to the field of big data technology, and in particular to a push optimization method and system based on e-commerce big data feedback. Background Technology

[0002] With the surge in business information and data, e-commerce service providers are increasingly reliant on big data technology to help managers make informed strategic decisions and enhance their competitive advantage. Big data applications in the e-commerce industry typically include precision marketing, personalized services, and personalized product recommendations. For example, personalized recommendation systems analyze user behavior, including feedback, purchase history, and social data, to identify the correlation between customers and products, uncovering users' personalized needs and interests. Based on these interests, e-commerce content can be pushed to them, effectively improving the service capabilities of the e-commerce system and retaining users.

[0003] However, in related technologies, there is a lack of corresponding feedback and optimization mechanisms after e-commerce content is pushed, resulting in very low conversion efficiency of the pushed data. Summary of the Invention

[0004] The purpose of this invention is to provide a push optimization method and system based on e-commerce big data feedback, so as to realize the feedback optimization of e-commerce push data and improve the push conversion rate of e-commerce data.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a push optimization method based on e-commerce big data feedback, comprising the following steps:

[0006] Acquire multi-source heterogeneous data and real-time e-commerce data streams, and verify the compliance of all acquired data;

[0007] Based on the verified e-commerce data, construct corresponding user profiles, extract corresponding push feedback features, and build a feedback-driven model.

[0008] The user profile and the push feedback features are input into the feedback-driven model, and the push data is calculated using the output expected feedback data.

[0009] The system acquires negative feedback data from the push notifications in real time and optimizes the push notifications based on this negative feedback data.

[0010] The multi-source heterogeneous data includes user basic data, user behavior data, product data, context data, and historical push data. The real-time e-commerce data stream is a combination of user behavior data and push feedback data acquired in real time through a message queue and a stream processing platform.

[0011] This includes compliance verification of all acquired data, including:

[0012] The acquisition channels of the multi-source heterogeneous data and the real-time e-commerce data stream are determined. If the requirements are not met, the relevant data is deleted.

[0013] Anonymize and / or pseudonymize the multi-source heterogeneous data and sensitive data in the real-time e-commerce data stream that have passed the judgment.

[0014] Prior to building the user profile, the method further includes:

[0015] After compliance verification, the data is cleaned and then merged based on the user-item-context-feedback event record format to obtain merged data.

[0016] This includes constructing corresponding user profiles based on verified e-commerce data, extracting corresponding push feedback features, and building a feedback-driven model, including:

[0017] Based on the fused data, a corresponding user profile is constructed, and the corresponding push feedback features are extracted;

[0018] A feedback-driven model is constructed based on deep learning models and traditional models, and the feedback-driven model is trained.

[0019] The process includes constructing a feedback-driven model based on deep learning and traditional models, and training the feedback-driven model, including:

[0020] By combining deep learning models and traditional models, a dual-channel feedback-driven model is constructed.

[0021] Historical e-commerce data is acquired and divided into a training set and a validation set. The training set is input into the constructed feedback-driven model for training, and the validation set is used to optimize the feedback-driven model.

[0022] The process involves inputting the user profile and the push feedback features into the feedback-driven model, and calculating the push data using the output expected feedback data, including:

[0023] The user profile and the push feedback features are input into the feedback-driven model, and the corresponding expected feedback data is generated by combining e-commerce data information.

[0024] By combining reinforcement learning strategies, the push content, push channel, and push time are extracted based on the expected feedback data.

[0025] The method further includes, after extracting the push content, push channel, and push time based on the expected feedback data:

[0026] Two types of push data are generated based on the user profile and the push content, respectively, and the corresponding push data is sent to the push channel based on the push time.

[0027] The process of acquiring negative feedback data from the push notifications in real time and optimizing the push notifications based on this negative feedback data includes:

[0028] After the push data is sent, the negative feedback data of the push data is obtained in real time, and it is determined which type of push data the current negative feedback data belongs to.

[0029] Based on the judgment result and the obtained negative feedback data type, the corresponding push data is marked, and the frequency of the corresponding push data is optimized based on the marking probability.

[0030] Secondly, the present invention provides a push optimization system based on e-commerce big data feedback, which is applied to a push optimization method based on e-commerce big data feedback as provided in the first aspect. The push optimization system based on e-commerce big data feedback includes a data acquisition module, a model building module, a push data generation module, and a negative feedback optimization module, wherein the data acquisition module, the model building module, the push data generation module, and the negative feedback optimization module are connected in sequence.

[0031] The data acquisition module is used to acquire multi-source heterogeneous data and real-time e-commerce data streams, and to perform compliance verification on all acquired data.

[0032] The model building module is used to build corresponding user profiles based on verified e-commerce data, extract corresponding push feedback features, and build a feedback-driven model.

[0033] The push data generation module is used to input the user profile and the push feedback features into the feedback-driven model, and to calculate the push data using the output expected feedback data.

[0034] The negative feedback optimization module is used to acquire negative feedback data of the push data in real time and optimize the push based on the negative feedback data.

[0035] This invention discloses a push optimization method and system based on e-commerce big data feedback. The push optimization system based on e-commerce big data feedback includes a data acquisition module, a model building module, a push data generation module, and a negative feedback optimization module. The system acquires multi-source heterogeneous data and real-time e-commerce data streams, and performs compliance verification on all acquired data. Based on the verified e-commerce data, it constructs corresponding user profiles and extracts corresponding push feedback features, while simultaneously building a feedback-driven model. The user profiles and push feedback features are input into the feedback-driven model, and push data is calculated using the output expected feedback data. Negative feedback data from the push data is acquired in real time, and push optimization is performed based on the negative feedback data to achieve feedback optimization of e-commerce push data and improve the push conversion rate of e-commerce data. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0037] Figure 1 This is a schematic diagram illustrating the steps of a push optimization method based on e-commerce big data feedback according to the first embodiment of the present invention.

[0038] Figure 2 This is a flowchart illustrating a push optimization method based on e-commerce big data feedback provided by the present invention.

[0039] Figure 3 This is a flowchart illustrating steps S103-S104 provided by the present invention.

[0040] Figure 4 This is a schematic diagram of the push optimization system based on e-commerce big data feedback according to the second embodiment of the present invention.

[0041] Figure 5 This is a schematic diagram of the electronic device of the present invention.

[0042] In the diagram: 101 - Data acquisition module, 102 - Model building module, 103 - Push data generation module, 104 - Negative feedback optimization module. Detailed Implementation

[0043] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0044] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0045] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0046] The first embodiment of this application is as follows:

[0047] See also Figures 1-3 This invention provides a push optimization method based on e-commerce big data feedback, comprising the following steps:

[0048] S101. Acquire multi-source heterogeneous data and real-time e-commerce data streams, and verify the compliance of all acquired data.

[0049] Specifically, existing web scraping technologies are used to acquire multi-source heterogeneous data and real-time e-commerce data streams. This multi-source heterogeneous data includes user basic data, user behavior data, product data, contextual data, and historical push data. User basic data includes registration information (age, gender, region – requires compliant anonymization), membership level, and account status. User behavior data includes explicit behavior, implicit behavior, and session information. Explicit behavior includes search keywords, product browsing (pages, duration, depth), product clicks, adding to cart, placing an order, successful payment, order cancellation, return and refund, favorites, following stores / brands, reviews / ratings, and sharing. Implicit behavior includes page scrolling speed, mouse movement trajectory (heatmap), page dwell time, and bounce rate. Session information includes access time, access frequency, session duration, and source channel (organic search, advertising, social media, direct access). Product data includes product attributes (category, brand, price, SKU, tags, description, images / videos) and content attributes (articles, videos, live streams, event themes, tags). Contextual data: Accessing device (PC / mobile / APP / Mini Program), operating system, network environment, geographical location (city level), current time (date, time, holidays), weather (optional). Historical push data: Pushed content (product ID, content ID), push time, push channel (APPPush, SMS, in-app message, email, ad placement), push template, user feedback on the push (exposure, click, ignore, close, conversion).

[0050] The real-time e-commerce data stream is a combination of user behavior data (especially browsing, clicking, adding to cart, and purchasing) and push feedback data acquired in real time through message queues (such as Kafka, Pulsar) and stream processing platforms (such as Flink, Spark Streaming).

[0051] To ensure that the data used complies with relevant privacy regulations (GDPR, CCPA, China's Personal Information Protection Law), the acquisition channels of the multi-source heterogeneous data and the real-time e-commerce data stream are assessed. If the requirements are not met, the relevant data is deleted; if the requirements are met, sensitive data in the multi-source heterogeneous data and the real-time e-commerce data stream that pass the assessment are anonymized and / or pseudonymous. This avoids the leakage of user privacy during subsequent e-commerce data push processes and ensures user privacy security as much as possible.

[0052] S102. Based on the verified e-commerce data, construct the corresponding user profile, extract the corresponding push feedback features, and construct a feedback-driven model.

[0053] Specifically, after compliance verification, the data undergoes data cleaning. This cleaning mainly involves removing missing values, outliers, and duplicate data, as well as unifying the ID system (user ID, product ID, session ID) of data from different sources. After data cleaning, irrelevant data is reduced to avoid complicating the data fusion process. Then, all the e-commerce data is merged based on the user-item-context-feedback event record format to obtain the merged data.

[0054] Based on the fused data, a corresponding user profile is constructed, which may include the following data: Long-term profile: historical preferences (category, brand, price range), purchasing power, activity level, and loyalty. Real-time / short-term profile (key innovation points): a dynamic interest vector is constructed based on the behavioral sequence (browsing, searching, adding to cart) of the most recent N minutes / hours and feedback on recent push notifications (click-through rate, ignore rate, conversion rate). For example: click-through rate of push notifications for the "sports shoes" category (last hour). Conversion rate of push notifications for the "discount promotion" type (today). Relevance feedback between recently viewed products and pushed products.

[0055] Furthermore, based on the fused data, item features and contextual features can be extracted. Item features include: popularity, new product status, discount level, inventory status, user reviews, and matching degree with user's historical preferences. Contextual features include: time features (weekdays / weekends, morning / evening peak hours), location features, and device features.

[0056] Based on the fused data, corresponding push feedback features are extracted. These features include individual feedback, group feedback, feedback timeliness weighting, and a feedback decay model. Individual feedback refers to historical feedback statistics (CTR, CVR, ignore rate, close rate, negative feedback markers) of users on specific types (e.g., brands, discounts, new products), specific formats (images / push / SMS), specific channels, and specific time points. Group feedback refers to aggregated feedback metrics of similar user groups on the same or similar push notifications. Feedback timeliness weighting assigns higher weight to more recent feedback. The feedback decay model models the decay of user interest and feedback effects over time. Specifically, the extraction method utilizes existing feature extraction techniques to extract keywords from the fused data. These keywords are then compared with each data type, and keywords with high similarity are grouped into their corresponding data types, thus obtaining the push feedback features corresponding to the fused data.

[0057] This paper combines traditional models (LR, GBDT) with deep learning models (Wide&Deep, DeepFM, DIN, DCN, Transformer-based sequence models) to construct a dual-channel feedback-driven model. Historical e-commerce data is acquired and divided into training and validation sets. The training set is input into the constructed feedback-driven model for training, and the validation set is used to optimize the model. The historical e-commerce data includes multi-source heterogeneous data and real-time e-commerce data streams, but the data is from the past year or so, primarily used for training and validating the feedback-driven model. The purpose of using historical e-commerce data is that complete e-commerce push results and feedback results are already available, allowing the trained feedback-driven model to more closely approximate real-world push scenarios.

[0058] S103. Input the user profile and the push feedback features into the feedback-driven model, and calculate the push data using the output expected feedback data.

[0059] Specifically, the user profile, item features, contextual features, and push feedback features are input into the feedback-driven model. The trained feedback-driven model is then invoked, and e-commerce data is used to predict the user's expected feedback data (click probability, conversion probability, negative feedback probability) for each candidate item / content under different push strategies (channel, time, template). Using a reinforcement learning strategy, the push content, push channel, and push time are extracted based on the expected feedback data. The push channel and push time with the highest feedback rate are determined based on the push content. This reduces the computational burden when generating push data later. Since the previous feedback-driven model can already output very accurate expected feedback data based on historical data, the optimal push channel and push time can be determined based on the output of the feedback-driven model, simply adapting to changes in the push content, greatly saving computation time.

[0060] Assumption:

[0061] U is the user profile feature vector.

[0062] I represents the item feature vector.

[0063] C is the context feature vector.

[0064] F is the feature vector of push feedback.

[0065] θ represents the parameters of the feedback-driven model.

[0066] The expected feedback data E can be expressed as: E = f(U, I, C, F; θ)

[0067] in:

[0068] f is a function of the feedback-driven model, which may be a combination of deep learning models (such as Wide&Deep, DeepFM, etc.) or traditional models (such as LR, GBDT, etc.).

[0069] E is a vector containing the click probability P. click Transformation probability P convert and negative feedback probability P negative .

[0070] Full form:

[0071]

[0072] Assumption:

[0073] π is a reinforcement learning strategy

[0074] S represents the state, which includes user profile, item characteristics, contextual characteristics, and expected feedback data.

[0075] A represents the action, which includes the push content, push channel, and push time.

[0076] Push strategy A can be represented as:

[0077] A = π(S)

[0078] To better align with user needs, two types of push data are generated based on the user profile and the push content, respectively. The process of generating push data based on the user profile is as follows: Based on the similarity between the long-term and short-term user profiles, the item / content with the highest similarity is selected. This candidate content is then input into the feedback-driven model, which outputs the corresponding candidate push channel and candidate push time. This results in two sets of push data, essentially two push templates. The push data generated based on the push channel and push time is designated as group A, and the other as group B for easier differentiation later. While the push content generation step also utilizes the user profile, this step primarily uses basic user profile information; it's essentially using features to assist in generating push feedback features, rather than relying on the similarity between the long-term and short-term user profiles.

[0079] The corresponding push data is sent to the push channel based on the push time.

[0080] S104. Obtain negative feedback data of the push data in real time, and optimize the push based on the negative feedback data.

[0081] Specifically, after the two types of push data are sent at their respective times, negative feedback data (ignore, turn off, mark "not interested") is acquired in real time, and it is determined which type of push data the current negative feedback data belongs to. The determination process is as follows: First, the time of the acquired negative feedback data is subtracted from the candidate push time and the push time. If the error is within 30 seconds, it is classified as the corresponding push data within the error range. For example, if the time of the acquired negative feedback data is 14:30:25, while the push time is set to 14:30 and the candidate push time is set to 14:33, then the error between the negative feedback data and the push time is within the set range, so the collected current negative feedback data is data from group A.

[0082] Based on the judgment results and the acquired negative feedback data type, the corresponding push data is marked. Before marking, the time when the negative feedback data was acquired needs to be statistically analyzed, and the deviation value is calculated, such as the aforementioned 25 seconds. The feedback probability is calculated using the deviation value and the error range, and the feedback probability is used for marking. The frequency of the corresponding push data is optimized based on the marked probability. That is, the higher the feedback probability, the lower the frequency of the corresponding push data is. The selection between A and B sets of push data is continuously performed (reducing the frequency of the corresponding push data is the selection process). Push data is generated entirely based on individual user profiles and their unique feedback history, and feedback push data is acquired. A / B testing is continuously performed to compare the effects of different models, strategies, templates, and frequency control parameters, thereby optimizing the feedback of e-commerce push data and improving the push conversion rate of e-commerce data.

[0083] Assumption:

[0084] N represents negative feedback data (such as ignored, turned off, or marked "not interested").

[0085] T push For push time

[0086] T feedback Time for negative feedback data

[0087] ΔT is the time deviation.

[0088] P negative Probability of negative feedback

[0089] F new For the new push frequency

[0090] Negative feedback data can be represented as:

[0091] ΔT=T feedback -T push|

[0092] If ΔT≤∈ (where ∈ is the set time error range, such as 30 seconds), then the negative feedback is considered to belong to the current push data.

[0093] Optimization of push frequency:

[0094] F new =F old ×(1-α×P negative )

[0095] Among them, F old α is the current push frequency; α is the adjustment coefficient, used to control the degree of impact of negative feedback on the push frequency.

[0096] The second embodiment of this application is as follows:

[0097] See also Figure 4 This invention provides a push optimization system based on e-commerce big data feedback, which is applied to a push optimization method based on e-commerce big data feedback as provided in the first embodiment. The push optimization system based on e-commerce big data feedback includes a data acquisition module 101, a model building module 102, a push data generation module 103, and a negative feedback optimization module 104. The data acquisition module 101, the model building module 102, the push data generation module 103, and the negative feedback optimization module 104 are connected in sequence.

[0098] The data acquisition module 101 is used to acquire multi-source heterogeneous data and real-time e-commerce data streams, and to perform compliance verification on all acquired data.

[0099] The model building module 102 is used to build a corresponding user profile based on the verified e-commerce data, extract the corresponding push feedback features, and build a feedback-driven model.

[0100] The push data generation module 103 is used to input the user profile and the push feedback features into the feedback-driven model, and to calculate the push data using the output expected feedback data.

[0101] The negative feedback optimization module 104 is used to acquire negative feedback data of the push data in real time and optimize the push based on the negative feedback data.

[0102] In this embodiment, firstly, the data acquisition module 101 acquires multi-source heterogeneous data and real-time e-commerce data streams, and performs compliance verification on all acquired data. The multi-source heterogeneous data includes basic user data, user behavior data, product data, context data, and historical push data. The real-time e-commerce data stream is a combination of user behavior data and push feedback data acquired in real-time through a message queue and stream processing platform. After compliance verification, the data undergoes data cleaning, and all the e-commerce data is merged based on a user-item-context-feedback event record format to obtain merged data. Then, the model building module 102 constructs a corresponding user profile based on the verified e-commerce data and extracts the corresponding push feedback features, while simultaneously building a feedback-driven model. After obtaining the feedback-driven model, it is trained and validated to ensure the model's accuracy. The user profile and the push feedback features are input into the feedback-driven model, and the push data generation module 103 calculates the push data using the output expected feedback data. The negative feedback optimization module 104 acquires the negative feedback data of the push data in real time and optimizes the push based on the negative feedback data, thus achieving both accurate push of e-commerce data and feedback optimization of e-commerce push data, thereby improving the push conversion rate of e-commerce data.

[0103] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0104] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0105] Accordingly, this application also provides an electronic device, including: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the push optimization method based on e-commerce big data feedback as described above. Figure 5 The diagram shown is a hardware structure diagram of any device with data processing capabilities, which is used in a push optimization system based on e-commerce big data feedback provided by an embodiment of the present invention. (Except for...) Figure 5In addition to the processor, memory, and network interface shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0106] Accordingly, this application also provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the push optimization method based on e-commerce big data feedback as described above. The computer-readable storage medium can be an internal storage unit of any data-processing device as described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units of any data-processing device and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the data-processing device, and can also be used to temporarily store data that has been output or will be output.

[0107] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0108] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A push optimization method based on e-commerce big data feedback, characterized in that, Includes the following steps: Acquire multi-source heterogeneous data and real-time e-commerce data streams, and verify the compliance of all acquired data; Based on the verified e-commerce data, construct corresponding user profiles, extract corresponding push feedback features, and build a feedback-driven model. The user profile and the push feedback features are input into the feedback-driven model, and the push data is calculated using the output expected feedback data. The system acquires negative feedback data from the push notifications in real time and optimizes the push notifications based on this negative feedback data.

2. The push optimization method based on e-commerce big data feedback as described in claim 1, characterized in that, The multi-source heterogeneous data includes user basic data, user behavior data, product data, context data, and historical push data. The real-time e-commerce data stream is a combination of user behavior data and push feedback data obtained in real time through message queues and stream processing platforms.

3. The push optimization method based on e-commerce big data feedback as described in claim 1, characterized in that, All acquired data undergoes compliance verification, including: The acquisition channels of the multi-source heterogeneous data and the real-time e-commerce data stream are determined. If the requirements are not met, the relevant data is deleted. Anonymize and / or pseudonymize the multi-source heterogeneous data and sensitive data in the real-time e-commerce data stream that have passed the judgment.

4. The push optimization method based on e-commerce big data feedback as described in claim 1, characterized in that, Before building user profiles, the method also includes: After compliance verification, the data is cleaned and then merged based on the user-item-context-feedback event record format to obtain merged data.

5. The push optimization method based on e-commerce big data feedback as described in claim 4, characterized in that, Based on the verified e-commerce data, corresponding user profiles are constructed, and corresponding push feedback features are extracted. Simultaneously, a feedback-driven model is built, including: Based on the fused data, a corresponding user profile is constructed, and the corresponding push feedback features are extracted; A feedback-driven model is constructed based on deep learning models and traditional models, and the feedback-driven model is trained.

6. The push optimization method based on e-commerce big data feedback as described in claim 5, characterized in that, Based on deep learning models and traditional models, a feedback-driven model is constructed and trained, including: By combining deep learning models and traditional models, a dual-channel feedback-driven model is constructed. Historical e-commerce data is acquired and divided into a training set and a validation set. The training set is input into the constructed feedback-driven model for training, and the validation set is used to optimize the feedback-driven model.

7. The push optimization method based on e-commerce big data feedback as described in claim 1, characterized in that, The user profile and the push feedback features are input into the feedback-driven model, and the push data is calculated using the output expected feedback data, including: The user profile and the push feedback features are input into the feedback-driven model, and the corresponding expected feedback data is generated by combining e-commerce data information. By combining reinforcement learning strategies, the push content, push channel, and push time are extracted based on the expected feedback data.

8. The push optimization method based on e-commerce big data feedback as described in claim 7, characterized in that, After extracting the push content, push channel, and push time based on the expected feedback data, the method further includes: Two types of push data are generated based on the user profile and the push content, respectively, and the corresponding push data is sent to the push channel based on the push time.

9. The push optimization method based on e-commerce big data feedback as described in claim 1, characterized in that, Real-time acquisition of negative feedback data from the push notifications, and optimization of push notifications based on the negative feedback data, including: After the push data is sent, the negative feedback data of the push data is obtained in real time, and it is determined which type of push data the current negative feedback data belongs to. Based on the judgment result and the obtained negative feedback data type, the corresponding push data is marked, and the frequency of the corresponding push data is optimized based on the marking probability.

10. A push optimization system based on e-commerce big data feedback, applied to the push optimization method based on e-commerce big data feedback as described in claim 1, characterized in that, The push optimization system based on e-commerce big data feedback includes a data acquisition module, a model building module, a push data generation module, and a negative feedback optimization module, which are connected in sequence. The data acquisition module is used to acquire multi-source heterogeneous data and real-time e-commerce data streams, and to perform compliance verification on all acquired data. The model building module is used to build corresponding user profiles based on verified e-commerce data, extract corresponding push feedback features, and build a feedback-driven model. The push data generation module is used to input the user profile and the push feedback features into the feedback-driven model, and to calculate the push data using the output expected feedback data. The negative feedback optimization module is used to acquire negative feedback data of the push data in real time and optimize the push based on the negative feedback data.

Images