E-commerce data update processing method and system based on real-time feedback of state differences

Abstract

This invention discloses a method and system for updating e-commerce data based on real-time feedback of state differences, belonging to the field of data processing technology. The key technical points are: acquiring real-time difference state vectors uploaded by the user; matching historical difference state vectors with a similarity greater than a set threshold from a database, and reading historical product update strategies and feedback evaluation values; generating real-time product update strategies by combining historical product update strategies, similarity, and feedback evaluation values, and distributing the real-time product update strategies to the user. This invention calculates the difference between real-time state vectors generated at adjacent times and uses a geometric progression encoding method to encode multidimensional behavioral data into lightweight state vectors. The user uploads only the encoded difference value to the e-commerce platform; this encoded difference value uniquely corresponds to a state transition path, allowing the e-commerce platform to accurately interpret behavioral intentions while preserving state changes and reducing data transmission volume.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method and system for updating e-commerce data based on real-time feedback of state differences. Background Technology

[0002] With the rapid development of e-commerce platforms, there is a large amount of product information that can be displayed on these platforms. In order to improve user experience and operational efficiency, personalized recommendation services have become a major application function of e-commerce platforms.

[0003] Existing e-commerce platforms typically involve each user client simplifying and processing the collected user behavior data locally before transmitting it to the e-commerce platform, which then updates product data based on the overall user behavior. Some e-commerce platforms also directly upload the collected user behavior data from the user clients for centralized processing. However, uploading complete raw user behavior data, such as click coordinates and dwell time, from the user clients can easily lead to high network load and response latency, making it difficult to meet real-time update requirements. Even after local simplification, retaining only core information, a large amount of data remains when many users browse simultaneously, which can easily cause system crashes during surges in user access.

[0004] Therefore, how to research and design an e-commerce data update processing method and system based on real-time feedback of state differences that can overcome the above-mentioned defects is a problem that we urgently need to solve. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for updating e-commerce data based on real-time feedback of state differences. By calculating the difference between real-time state vectors generated at adjacent times and using a geometric progression encoding method, multidimensional behavioral data is encoded into lightweight state vectors. The user uploads the encoded difference, which contains only differences, to the e-commerce platform. The encoded difference uniquely corresponds to the state transition path, allowing the e-commerce platform to accurately interpret behavioral intentions. This reduces data transmission volume while preserving state changes, effectively improving the efficiency of e-commerce data updates.

[0006] The above-mentioned technical objective of the present invention is achieved through the following technical solution:

[0007] Firstly, a method for updating e-commerce data based on real-time feedback of status differences is provided, including the following steps:

[0008] Obtain the real-time difference state vector uploaded by the user terminal. The real-time difference state vector is the difference between the real-time state vectors generated by the user terminal at adjacent times. The real-time state vector is a lightweight state vector encoded by the user behavior events captured by the user terminal in real time.

[0009] The database is matched to obtain historical difference state vectors whose similarity to the real-time difference state vector is greater than a set threshold, and the historical product update strategy and feedback evaluation value corresponding to the historical difference state vector are read.

[0010] A real-time product update strategy is generated by combining the historical product update strategy, the similarity, and the feedback evaluation value, and the real-time product update strategy is sent to the user terminal.

[0011] Furthermore, the process of generating the real-time state vector includes:

[0012] Extract state features for each event type from the user behavior events;

[0013] The feature values ​​of the state feature are converted into encoded values, and different feature values ​​of the same state feature are encoded using a geometric progression encoding.

[0014] The feature fusion engine compresses all the encoded values ​​into a fixed-dimensional vector to obtain the real-time state vector.

[0015] Furthermore, the event types include click events, page view events, and search events;

[0016] The state features of the click event include the embedded vector of the product ID, the grid partitioning code of the screen click position, and the normalized value of the click frequency within the sliding window;

[0017] The state characteristics of the page browsing event include the embedding vector of the page ID, the scroll depth percentage, and the dwell time focus index based on Z-score normalization;

[0018] The state characteristics of the search event include the semantic vector mean of the search terms and the binary encoding of the filtering conditions.

[0019] Furthermore, the step of matching historical difference state vectors from the database with a similarity greater than a set threshold to the real-time difference state vector includes:

[0020] Traverse all ID coding entries in the real-time difference state vector, classify ID coding entries with zero coding difference as first coding entries, and classify ID coding entries with non-zero coding difference as second coding entries;

[0021] Read the first encoding difference value of the first encoding item in the cache list that was last written, and write the first encoding difference value into the corresponding ID encoding item in the real-time difference state vector to obtain the updated real-time difference state vector.

[0022] Write the second encoding difference value corresponding to the second encoding item in the real-time difference state vector to the cache list to update the cache list;

[0023] Calculate the similarity between the updated real-time difference state vector and the reference difference state vector. The encoding difference corresponding to the ID encoding item in the reference difference state vector is updated through the corresponding cache list.

[0024] The reference difference state vector corresponding to the similarity greater than a set threshold is used as the historical difference state vector obtained by matching.

[0025] Furthermore, the similarity is the percentage of encoded terms with equal encoded differences between the real-time difference state vector and the reference difference state vector.

[0026] Furthermore, the similarity is the sum of the weight values ​​corresponding to the coding terms whose coding differences are equal between the real-time difference state vector and the reference difference state vector;

[0027] Wherein, if the coding difference between the coding terms of the real-time difference state vector and the reference difference state vector is equal, then the weight value of the corresponding coding term is the weight coefficient of the corresponding coding term;

[0028] If the coding difference between the real-time difference state vector and the reference difference state vector is not equal, then the weight value of the corresponding coding term is zero.

[0029] Furthermore, the process for determining the feedback evaluation value is as follows:

[0030] Obtain the next historical difference state vector T+1 from the historical difference state vector T;

[0031] The item evaluation value of the corresponding coding item is obtained by dividing the coding difference of each coding item in the historical difference state vector T+1 by the maximum coding difference of the corresponding coding item, and then multiplying it by the weight coefficient of the corresponding coding item.

[0032] Calculate the sum of all the evaluation values ​​to obtain the feedback evaluation value of the historical difference state vector T+1.

[0033] Furthermore, the step of generating a real-time product update strategy by combining the historical product update strategy, the similarity, and the feedback evaluation value includes:

[0034] If the similarity is greater than or equal to the first threshold and the feedback evaluation value is greater than or equal to the third threshold, then the historical product update strategy shall be used as the real-time product update strategy.

[0035] If the similarity is less than the first threshold and greater than or equal to the second threshold, and the feedback evaluation value is greater than or equal to the third threshold, then a local update strategy related to the coding item corresponding to the coding difference is selected from the historical product update strategies, and a real-time product update strategy is reconstructed.

[0036] Furthermore, the method also includes:

[0037] Upon receiving the real-time product update strategy, the user terminal prioritizes implementing the update strategy of keeping the display interface static.

[0038] Secondly, an e-commerce data update and processing system based on real-time feedback of state differences is provided. This system is used to implement the e-commerce data update and processing method based on real-time feedback of state differences as described in any one of the first aspects, including:

[0039] The data acquisition module is used to acquire the real-time difference state vector uploaded by the user terminal. The real-time difference state vector is the difference between the real-time state vectors generated by the user terminal at adjacent times. The real-time state vector is a lightweight state vector encoded by the user behavior events captured by the user terminal in real time.

[0040] The matching analysis module is used to match historical difference state vectors with a similarity greater than a set threshold to the real-time difference state vector from the database, and to read the historical product update strategy and feedback evaluation value corresponding to the historical difference state vector.

[0041] The strategy generation module is used to combine the historical product update strategy, the similarity and the feedback evaluation value to generate a real-time product update strategy, and to send the real-time product update strategy to the user terminal.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. The e-commerce data update processing method based on real-time feedback of state differences provided by the present invention calculates the difference between real-time state vectors generated at adjacent times and encodes multidimensional behavioral data into lightweight state vectors using a geometric progression encoding method. The user end uploads the encoded difference value, which only has differences, to the e-commerce platform. The encoded difference value uniquely corresponds to the state transition path, which can be used by the e-commerce platform to accurately interpret the behavioral intent. While retaining the state change situation, it reduces the amount of data transmission and effectively improves the efficiency of e-commerce data update.

[0044] 2. This invention addresses the matching blind spot when the encoding difference of ID encoding items is zero. By designing a dynamic update process for the cache list, the zero-difference ID encoding items are replaced with the previous cached value to restore the historical product context. Meanwhile, the non-zero-difference ID encoding items are updated in real time to ensure the continuity of subsequent matching. This can solve the problem of strategy disconnection when operating on the same product consecutively and effectively improve the accuracy of similarity matching.

[0045] 3. This invention distinguishes key behaviors by weighting coefficients, and the similarity calculation only accumulates weighted matching items to avoid interference from secondary noise. In addition, this invention quantifies the effect of the strategy by feedback evaluation values, and dynamically evaluates the effectiveness of historical strategies based on the ratio of the encoding difference of the next state vector to the maximum theoretical difference.

[0046] 4. This invention designs a dual-threshold decision mechanism. For matching cases with high similarity and high feedback value, the historical strategy is directly reused; while for matching cases with medium similarity and high feedback value, a local strategy is extracted and reconstructed, which can effectively ensure the success rate of e-commerce data updates.

[0047] 5. In this invention, the user terminal prioritizes the execution of static interface update strategies, such as background data loading and pre-rendering, to achieve seamless switching of product information and avoid interface flickering that interferes with user operation. Attached Figure Description

[0048] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0049] Figure 1 This is a flowchart from Embodiment 1 of the present invention;

[0050] Figure 2 This is a system block diagram in Embodiment 2 of the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0052] Example 1: E-commerce data update processing method based on real-time feedback of state differences, such as Figure 1 As shown, it includes the following steps:

[0053] S1: Obtain the real-time difference state vector uploaded by the user. The real-time difference state vector is the difference between the real-time state vectors generated by the user at adjacent times. The real-time state vector is a lightweight state vector encoded by the user behavior events captured by the user in real time.

[0054] S2: Match historical difference state vectors from the database whose similarity to real-time difference state vectors is greater than a set threshold, and read the historical product update strategies and feedback evaluation values ​​corresponding to the historical difference state vectors;

[0055] S3: Combine historical product update strategies, similarity, and feedback ratings to generate real-time product update strategies, and then distribute these strategies to users.

[0056] In step S1, the real-time state vector generation process includes: extracting state features of each event type from user behavior events; converting the feature values ​​of the state features into encoded values, and encoding different feature values ​​of the same state feature using a geometric progression; and compressing all encoded values ​​into a fixed-dimensional vector through a feature fusion engine to obtain the real-time state vector.

[0057] In some examples, the event types include, but are not limited to, click events, page view events, and search events; among them, the state features of click events include the embedding vector of the product ID, the grid partitioning encoding of the screen click position, and the normalized value of the click frequency within the sliding window; the state features of page view events include the embedding vector of the page ID, the scroll depth percentage, and the dwell time focus index based on Z-score normalization; the state features of search events include the semantic vector mean of the search terms and the binary encoding of the filter conditions.

[0058] In some examples, different coded terms in a real-time state vector can be encoded using different geometric sequences. For instance, a coded term can use a geometric sequence with a first term of 2 and a common ratio of 2. Different state features within a coded term can be encoded as {2, 4, 8, ..., 2}. n-1 ,2 n For example, if an encoding term uses a geometric sequence with the first term 3 and a common ratio of 3, then different state features in an encoding term can take the values ​​{3, 9, 27, ..., 3} after encoding. n-1 ,3 n}

[0059] In some examples, different coded terms in a real-time state vector can also correspond to different intervals in a geometric sequence, such as one coded term corresponding to the interval {2,4,8}, and another different coded term corresponding to the interval {16,32,64}.

[0060] In this invention, the real-time difference state vector is the difference between the real-time state vectors generated by the user terminal at adjacent time points. Specifically, it is the difference between the encoded terms in the real-time state vector at time t+1 and the encoded terms in the real-time state vector at time t. The difference in encoding for each term indicates the state feature from which it changes to another. For example, if the encoded value of state feature A is 4 and the encoded value of state feature B is 16, the encoding difference for the change from state feature A to state feature B is 12. Regardless of whether different encoded terms use different geometric sequences or different intervals of the same geometric sequence, given a defined encoding rule, the aforementioned encoding difference of 12 only indicates the change from state feature A to state feature B. Thus, after encoding a large amount of feature information into a lightweight state vector, the e-commerce platform can still accurately parse the information uploaded by the user terminal, effectively improving data transmission and processing efficiency while reducing network resource consumption.

[0061] In addition, to further reduce the amount of data in the fixed-dimensional vector, the encoded items with zero encoding difference in the fixed-dimensional vector can be compressed, such as by deletion. Further compression requires configuring a storage module in the e-commerce platform to store the non-zero encoded items in the previous real-time difference state vector. When reading complete data, the historical encoded items stored in the storage module can be filled into the real-time difference state vector to achieve complete information expansion.

[0062] In step S2, since the present invention encodes state features into lightweight state vectors, when the user performs operations on the same product at adjacent times, the encoding difference of the ID encoding item in the resulting real-time difference state vector is zero. At this time, it is impossible to obtain the product information operated by the user by parsing the real-time difference state vector. If the similarity is calculated directly on the real-time difference state vector, it is easy to match the historical difference state vector of different products, which will increase the probability of reverse product updates. Therefore, the present invention configures a cache list on the e-commerce platform to constrain the matching of historical difference state vectors.

[0063] The process of matching historical difference state vectors with a similarity greater than a set threshold to the real-time difference state vector from the database includes: traversing all ID encoding items in the real-time difference state vector, classifying ID encoding items with a coding difference of zero as first encoding items, and classifying ID encoding items with a coding difference of non-zero as second encoding items; reading the first coding difference value of the first encoding item previously written from the cache list, and writing the first coding difference value to the corresponding ID encoding item in the real-time difference state vector to obtain the updated real-time difference state vector; writing the second coding difference value corresponding to the second encoding item in the real-time difference state vector to the cache list to update the cache list; calculating the similarity between the updated real-time difference state vector and the reference difference state vector, and updating the corresponding cache list using the coding difference value of the ID encoding item in the reference difference state vector; and using the reference difference state vector with a similarity greater than the set threshold as the matched historical difference state vector.

[0064] It should be noted that different state features of the same coding term can be sorted according to the degree of feature similarity during the coding process. Then, the sorted state features are matched with a geometric sequence. For example, if the feature values ​​of the three state features are 0.7, 0.85, and 0.9, the corresponding coding values ​​can be 2, 4, and 8, respectively.

[0065] In some examples, similarity is the percentage of encoded terms with equal encoding differences between the real-time difference state vector and the reference difference state vector. For instance, if a complete real-time difference state vector has 20 encoded terms, and 16 of them have equal encoding differences with the reference difference state vector, then the calculated similarity is 0.8.

[0066] In some examples, considering that differences in ID coding terms have a greater weight in product updates than differences in coding terms in regular operations (e.g., changes in the viewed product are more important than changes in the viewed product time), weight coefficients can be configured for different coding terms according to the importance of coding term changes to product updates. The similarity is the sum of the weight values ​​corresponding to coding terms whose coding differences between the real-time difference state vector and the reference difference state vector are equal.

[0067] Specifically, if the coding difference between the real-time difference state vector and the reference difference state vector is equal, the weight value of the corresponding coding term is the weight coefficient of the corresponding coding term; if the coding difference between the real-time difference state vector and the reference difference state vector is not equal, the weight value of the corresponding coding term is zero.

[0068] This invention takes into account that the historical product update strategy may be affected by factors such as popular events, seasonal changes, and special holidays during the execution process, which may easily lead to poor execution results. Therefore, this invention analyzes the execution of the historical product update strategy through feedback evaluation values. The feedback evaluation values ​​are mainly obtained by comparing and analyzing the relative change of the encoding difference between two adjacent historical difference state vectors.

[0069] In some examples, the process of determining the feedback evaluation value is as follows: Obtain the next historical difference state vector T+1 from the historical difference state vector T; divide the coding difference of each coding item in the historical difference state vector T+1 by the maximum coding difference of the corresponding coding item, and then multiply by the weight coefficient of the corresponding coding item to obtain the item evaluation value of the corresponding coding item; calculate the sum of all item evaluation values ​​to obtain the feedback evaluation value of the historical difference state vector T+1. It should be noted that the maximum coding difference is the difference between the largest and smallest coding values ​​in a coding item.

[0070] In addition, the aforementioned threshold settings may include a first threshold and a second threshold. The second threshold is smaller than the first threshold. The first threshold is mainly to filter out product update strategies that can be directly applied, while the second threshold is mainly to select multiple similar product update strategies for combination and reconstruction when there are no directly available product update strategies.

[0071] In some examples, the first and second thresholds can be dynamically adjusted based on the amount of sample data in the database; the more sample data there is, the larger the values ​​of the first and second thresholds will be.

[0072] In some examples, since the second threshold needs to ensure that the vast majority of users successfully update their products, both the second and first thresholds can be reduced when the frequency of combined reconstruction increases to the upper limit. Conversely, when the frequency of combined reconstruction decreases to the lower limit, the first threshold can be appropriately increased, while the second threshold can remain unchanged or be increased.

[0073] In step S3, a real-time product update strategy is generated by combining historical product update strategies, similarity, and feedback evaluation values. This includes: if the similarity is greater than or equal to a first threshold and the feedback evaluation value is greater than or equal to a third threshold, then the historical product update strategy is used as the real-time product update strategy; if the similarity is less than the first threshold but greater than or equal to a second threshold and the feedback evaluation value is greater than or equal to a third threshold, then a local update strategy related to the coding item corresponding to the coding difference is selected from the historical product update strategies, and the real-time product update strategy is reconstructed and generated.

[0074] It should be noted that the product update strategy described in this invention consists of multiple independently executed update instructions, such as the exposure time of a single product, adjusting the ranking weight of search results, pushing promotional information, replacing product information, etc.

[0075] The aforementioned third threshold is generally determined by the overall feedback level in the database. For example, if 70% of the feedback evaluation values ​​are greater than 0.8, then 0.8 can be selected as the third threshold.

[0076] In some examples, to avoid visually impacting user operations due to product updates, the client prioritizes maintaining a static display interface after receiving the real-time product update strategy. This ensures that the user's visual perception of the display interface is not significantly different, enhancing the user experience.

[0077] Example 2: E-commerce data update processing system based on real-time feedback of state differences. This system is used to implement the e-commerce data update processing method based on real-time feedback of state differences as described in Example 1. Figure 2 As shown, it includes a data acquisition module, a matching analysis module, and a strategy generation module.

[0078] The system includes a data acquisition module for acquiring real-time difference state vectors uploaded by the user terminal. These real-time difference state vectors are the differences between real-time state vectors generated by the user terminal at adjacent times, and are lightweight state vectors encoded by user behavior events captured in real-time by the user terminal. A matching analysis module is used to match historical difference state vectors from a database that have a similarity greater than a set threshold with the real-time difference state vectors, and to read the historical product update strategies and feedback evaluation values ​​corresponding to the historical difference state vectors. A strategy generation module is used to combine the historical product update strategies, the similarity, and the feedback evaluation values ​​to generate a real-time product update strategy, and to send the real-time product update strategy to the user terminal.

[0079] It should be noted that the above system can be an e-commerce platform.

[0080] Working principle: This invention calculates the difference between real-time state vectors generated at adjacent times and uses a geometric progression encoding method to encode multidimensional behavioral data into lightweight state vectors. The user uploads the encoded difference, which only has differences, to the e-commerce platform. The encoded difference uniquely corresponds to the state transition path, which can be used by the e-commerce platform to accurately interpret behavioral intentions. While preserving the state change situation, it reduces the amount of data transmission and effectively improves the efficiency of e-commerce data update.

[0081] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0082] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0083] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0084] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0085] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An e-commerce data updating processing method based on real-time feedback of state difference, characterized in that, The method comprises the following steps: obtaining a real-time difference state vector uploaded by a user terminal, the real-time difference state vector being a difference between real-time state vectors generated by the user terminal at adjacent time points, the real-time state vector being a lightweight state vector encoded by a user behavior event captured by the user terminal in real time; matching a historical difference state vector from a database, the historical difference state vector having a similarity to the real-time difference state vector greater than a set threshold, and reading a historical commodity update strategy and a feedback evaluation value corresponding to the historical difference state vector; generating a real-time commodity update strategy in combination with the historical commodity update strategy, the similarity and the feedback evaluation value, and issuing the real-time commodity update strategy to the user terminal; the matching of the historical difference state vector from the database includes: traversing all ID encoding items in the real-time difference state vector, dividing ID encoding items with a zero encoding difference value into first encoding items, and dividing ID encoding items with a non-zero encoding difference value into second encoding items; reading a first encoding difference value of the first encoding item from a cache list and writing the first encoding difference value into a corresponding ID encoding item in the real-time difference state vector to obtain an updated real-time difference state vector; writing a second encoding difference value corresponding to the second encoding item in the real-time difference state vector into the cache list to update the cache list; calculating a similarity between the updated real-time difference state vector and a reference difference state vector, the encoding difference value of the ID encoding item in the reference difference state vector being updated by the corresponding cache list; taking the reference difference state vector corresponding to the similarity greater than the set threshold as the matched historical difference state vector.

2. The method of claim 1, wherein, The generation process of the real-time state vector comprises: extracting state features of each event type from the user behavior event; converting feature values of the state features into encoding values, different feature values of the same state feature being encoded by geometric progression encoding; compressing all the encoding values into a fixed dimension vector by a feature fusion engine to obtain a real-time state vector. 3.The state difference-based real-time feedback e-commerce data updating processing method of claim 2, wherein, The event types include a click event, a page browsing event and a search event; wherein the state features of the click event include an embedding vector of a commodity ID, a grid partition encoding of a screen click position, and a click frequency normalization value in a sliding window; the state features of the page browsing event include an embedding vector of a page ID, a scroll depth percentage and a stay duration focus index based on Z-score standardization; the state features of the search event include a semantic vector mean of a search word and a binary encoding of a filtering condition.

4. The method for real-time feedback based on state difference of e-commerce data update processing according to claim 1, characterized in that, The similarity is a proportion of the number of encoding items with equal encoding difference values between the real-time difference state vector and the reference difference state vector.

5. The method for real-time feedback based on state difference of e-commerce data update processing according to claim 1, characterized in that, The similarity is a sum of weight values of encoding items corresponding to equal encoding difference values between the real-time difference state vector and the reference difference state vector. If the encoding difference value of the encoding item between the real-time difference state vector and the reference difference state vector is equal, the weight value corresponding to the encoding item is the weight coefficient of the corresponding encoding item. If the encoding difference value of the encoding item between the real-time difference state vector and the reference difference state vector is not equal, the weight value corresponding to the encoding item is zero.

6. The method for real-time feedback based on state difference of e-commerce data update processing according to claim 1, characterized in that, The determination process of the feedback evaluation value is as follows: A next historical difference state vector T+1 of the historical difference state vector T is obtained. The item evaluation value of the corresponding encoding item is obtained by dividing the encoding difference value of each encoding item in the historical difference state vector T+1 by the maximum encoding difference value of the corresponding encoding item, and then multiplying the weight coefficient of the corresponding encoding item. The sum value of all the item evaluation values is calculated to obtain the feedback evaluation value of the historical difference state vector T+1.

7. The method for real-time feedback based on state difference of e-commerce data update processing according to claim 1, characterized in that, The real-time commodity update strategy is generated by combining the historical commodity update strategy, the similarity, and the feedback evaluation value, including: If the similarity is greater than or equal to the first threshold value, and the feedback evaluation value is greater than or equal to the third threshold value, the historical commodity update strategy is used as the real-time commodity update strategy. If the similarity is less than the first threshold value and greater than or equal to the second threshold value, and the feedback evaluation value is greater than or equal to the third threshold value, a local update strategy related to the encoding item corresponding to the equal encoding difference value is selected from the historical commodity update strategy to reconstruct and generate a real-time commodity update strategy. 8.The state difference-based real-time feedback e-commerce data updating processing method of claim 1, wherein, The method further includes: After receiving the real-time commodity update strategy, the user end preferentially executes an update strategy that keeps the display interface static.

9. An e-commerce data updating processing system based on real-time feedback of state difference, characterized in that, The system is used to implement the e-commerce data update processing method based on real-time feedback of state difference as claimed in any one of claims 1-8, including: A data acquisition module is configured to acquire a real-time difference state vector uploaded by a user end, wherein the real-time difference state vector is the difference between real-time state vectors generated by the user end at adjacent time points, and the real-time state vector is a lightweight state vector encoded by a user behavior event captured by the user end in real time; A matching analysis module is configured to match a historical difference state vector with a similarity greater than a set threshold value from the real-time difference state vector in a database, and read a historical commodity update strategy and a feedback evaluation value corresponding to the historical difference state vector; A strategy generation module is configured to generate a real-time commodity update strategy by combining the historical commodity update strategy, the similarity, and the feedback evaluation value, and to send the real-time commodity update strategy to the user end.

Images