Method and System for Optimizing Commodity Search Based on Historical Behavior Data

Abstract

The present invention discloses a method and system for optimizing commodity search based on historical behavior data, which relates to the technical field of optimizing commodity search, and specifically includes the following steps: According to the division results of each source path, a path signal dynamic adjustment mechanism is constructed for separately adjusting different sorting weights for the source paths of the dominant path, the dispersed path, and the mixed path; during the process of the path signal dynamic adjustment mechanism adjusting each source path, the path dynamic feedback information of each source path is obtained in real time, and after obtaining it, an analysis is carried out to evaluate whether the adjustment effect of the path signal dynamic adjustment mechanism on each source path can reach the expectation, and the path signal dynamic adjustment mechanism is optimized according to the evaluation result. The present invention solves the problem of misjudgment of multi-path repeated click signals, realizes the dynamic balance of recommendation accuracy and diversity, and improves the response ability and user satisfaction of the recommendation system.

Description

Technical Field

[0001] The present invention relates to the technical field of commodity search optimization, and in particular to a commodity search optimization method and system based on historical behavior data. Background Art

[0002] Product search is a core function used by users of e-commerce platforms when looking for specific products. By entering keywords or filtering conditions, users can find results that meet their needs from a large number of products. Common implementation methods include keyword-based text matching, product attribute-based filtering, and personalized recommendations combined with machine learning. However, traditional search methods usually only rely on keyword matching and product attribute screening, which easily ignores users' actual needs and preferences, resulting in a mismatch between search results and user expectations, thereby reducing user experience and conversion rate. Therefore, it is particularly important to optimize product search in order to improve the accuracy and relevance of search results and meet users' personalized needs. The optimization method based on historical behavior data can deeply explore users' potential intentions and build a personalized recommendation model by collecting and analyzing users' search, browsing, clicking, purchasing and other behavioral information, so that search results are not only based on the input keywords, but also can be intelligently sorted and recommended in combination with users' past behaviors, thereby improving search accuracy and user satisfaction, and ultimately improving the transaction conversion rate of the platform.

[0003] Existing product search optimization technology based on historical behavior data mainly collects multi-dimensional behavior data of users on e-commerce platforms, such as search keywords, product browsing records, click behaviors, favorite and add-to-cart behaviors, and actual purchase records, etc., uses these data to build user feature portraits, and adopts machine learning and deep learning models (such as RNN, Transformer) to predict users' potential search intentions and preferences. When a user initiates a search request, the historical behavior feature vector is extracted and the intent is analyzed in combination with the user's current search context. A candidate set of products that highly match the user's preferences is calculated, and the products are sorted through a comprehensive scoring model. The scoring model will comprehensively evaluate the relevance of products based on multiple dimensions such as user intention index and behavior feature information, so as to give priority to recommending products that better meet user needs. At the same time, this optimization process continuously adjusts model parameters through real-time monitoring of search click-through rate, conversion rate and user interaction behavior feedback to dynamically adapt to changes in user needs, continuously improve the accuracy of search results and personalized recommendation effects, thereby improving the user satisfaction and commercial conversion rate of the platform.

[0004] The prior art has the following deficiencies:

[0005] In the case where a user clicks on the same product multiple times through multiple recommended source paths (such as search results, recommendation lists, or ad slots), the system will simply superimpose these click behaviors as strong interest signals for that product. Since existing product search optimization technologies based on historical behavior data do not independently model and evaluate the click source paths when extracting click behavior features, they cannot distinguish that these click behaviors may be operations for the user to verify the consistency of different recommended sources, rather than true preference expressions. This will lead to the incorrect amplification of the interest weight for that product. Due to the lack of refined analysis of the click sources in the existing technologies, the nature of the signal with repeated paths cannot be correctly identified, resulting in the over-priority display of that product in the recommendation ranking and ignoring the potential interest needs of the user for other products. Furthermore, this problem will increase the redundancy of the recommendation results, making the user feel that the recommended content is single, reducing the satisfaction of the search experience. At the same time, the waste of resources on low-demand products by the platform will also affect the overall recommendation efficiency.

[0006] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure. Therefore, it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0007] The object of the present invention is to provide a product search optimization method and system based on historical behavior data to solve the problems in the above background art.

[0008] To achieve the above object, the present invention provides the following technical solutions: A product search optimization method based on historical behavior data, specifically including the following steps:

[0009] By real-time detecting user click behaviors, identifying whether the user clicks on the same product multiple times through multiple recommended source paths, and in the case of identifying that the user clicks on the same product multiple times through multiple recommended source paths, classifying and marking the multiple source paths generated by the user during the period of searching for the product to distinguish the click behavior characteristics of different source paths;

[0010] Real-time obtaining the click behavior information of each source path, and analyzing it after obtaining, evaluating the contribution intensity of each source path to the user's true interest signal, and classifying each source path into a dominant path, a dispersed path, and a mixed path according to the evaluation results;

[0011] According to the classification results of each source path, constructing a path signal dynamic adjustment mechanism for respectively adjusting different sorting weights for the source paths of the dominant path, the dispersed path, and the mixed path;

[0012] During the process of the path signal dynamic adjustment mechanism adjusting each source path, the path dynamic feedback information of each source path is obtained in real time, analyzed after acquisition, evaluated whether the adjustment effect of the path signal dynamic adjustment mechanism on each source path can meet the expectations, and the path signal dynamic adjustment mechanism is optimized according to the evaluation results;

[0013] Comprehensively analyze the application effect, adjustment records and optimization results of the path signal dynamic adjustment mechanism, and continuously optimize the path signal processing ability and dynamic adjustment mechanism of the recommendation ranking model in combination with the new user behavior data.

[0014] Preferably, the click behavior information of each source path is obtained in real time, analyzed after acquisition, evaluate the contribution intensity of each source path to the user's true interest signal, and divide each source path into a dominant path, a dispersed path and a mixed path according to the evaluation results. The specific steps are as follows:

[0015] Obtain the click behavior information of each source path in real time and perform preprocessing after acquisition;

[0016] Extract the path click distribution information and path click feature information from the click behavior information of each source path after preprocessing, analyze after extraction, and generate the click concentration coefficient and click sparsity index of each source path respectively;

[0017] Construct a path contribution intensity model with the generated click concentration coefficient and click sparsity index of each source path, generate the contribution evaluation coefficient of each source path, compare the generated contribution evaluation coefficient of each source path with the preset contribution evaluation coefficient threshold interval, evaluate the contribution intensity of each source path to the user's true interest signal according to the comparison result, and divide each source path into a dominant path, a dispersed path and a mixed path according to the evaluation results.

[0018] Preferably, the acquisition logic of the click concentration coefficient and click sparsity index of each source path is as follows:

[0019] Extract the path click distribution information from the click behavior information of each source path after preprocessing, specifically including the number of clicks of users on each product in each source path at different times within a period of time, the total number of clicks, and the deviation value between the number of clicks on each product and the total number of clicks in the source path, and use functions 、 and to represent them respectively according to the time series, is the time point, represents within a period of time at the moment, the user clicks on the th source path for the The number of clicks on a product represents the total number of clicks by the user within a certain period of time at the th source path represents the total number of clicks by the user within a certain period of time at the th source path for the th product, and the deviation value between the number of clicks on the product and the total number of clicks within the source path , , and are all positive integers. The defined time period is ;

[0020] Calculate the click concentration coefficient for each source path. The specific calculation formula is as follows:

[0021] ;

[0022] In the formula, is the click concentration coefficient for the th source path;

[0023] Extract the path click feature information from the click behavior information of each preprocessed source path, specifically including the total number of different products clicked by the user within each source path at different times within a certain period, the number of product categories, and the average value of the number of clicks on all products by the user within each source path, and represent them respectively by functions , and in the time series is the time point, represents the total number of different products clicked by the user within the th source path at the th time within a certain period of time represents the number of product categories of different products clicked by the user within the th source path at the th time within a certain period of time represents the average value of the number of clicks on all products by the user within the th source path at the th time within a certain period of time;

[0024] Calculate the click sparsity index for each source path. The specific calculation formula is as follows:

[0025] ;

[0026] In the formula, is the Click sparsity index of each source path.

[0027] Preferably, the click concentration coefficient of each generated source path and the click sparsity index are used to construct a path contribution intensity model, and the contribution evaluation coefficients of each source path are generated through weighted summation , and the generated contribution evaluation coefficients of each source path are compared with a pre-set contribution evaluation coefficient threshold interval , and the contribution intensity of each source path to the user's true interest signal is evaluated according to the comparison result, and each source path is divided into a dominant path, a dispersed path, and a mixed path according to the evaluation result. The specific comparison analysis and division are as follows:

[0028] If , the contribution intensity of this source path to the user's true interest signal is low intensity, then this source path is divided into a dispersed path;

[0029] If , the contribution intensity of this source path to the user's true interest signal is medium intensity, then this source path is divided into a mixed path;

[0030] If , the contribution intensity of this source path to the user's true interest signal is high intensity, then this source path is divided into a dominant path.

[0031] Preferably, according to the division results of each source path, a path signal dynamic adjustment mechanism is constructed to separately adjust the sorting weights of the dominant path, the dispersed path, and the mixed path source paths. Specifically:

[0032] According to the division results of the dominant path, the dispersed path, and the mixed path, different sorting weight adjustment rules are set respectively to form a path signal dynamic adjustment mechanism; this adjustment mechanism is based on the contribution evaluation coefficient of each source path and the optimization goal of the recommendation system, and automatically determines the sorting weight adjustment method and range of the products within each path through pre-set rules;

[0033] Different sorting strategy adjustments are made for the source paths of the dominant path, the dispersed path, and the hybrid path, specifically including: for the dominant path, the weight increase rule in the adjustment mechanism is used to increase the sorting weight of the clicked products in the path, and the products of the user's clear interests are preferentially displayed; for the dispersed path, the weight dispersion rule in the adjustment mechanism is used to reduce the sorting weight of the repeatedly clicked products, and at the same time increase the display weight of other unclicked products in the path to increase the diversity of the recommended content; for the hybrid path, the weight balance rule in the adjustment mechanism is used to maintain the sorting weight of the clicked products in the path to balance the accuracy and diversity of the recommendation.

[0034] Preferably, during the process of the path signal dynamic adjustment mechanism adjusting each source path, the path dynamic feedback information of each source path is obtained in real time, analyzed after being obtained, evaluated whether the adjustment effect of the path signal dynamic adjustment mechanism in each source path can reach the expectation, and the path signal dynamic adjustment mechanism is optimized according to the evaluation result, specifically including the following steps:

[0035] During the process of the path signal dynamic adjustment mechanism adjusting each source path, the path dynamic feedback information of each source path is obtained in real time and preprocessed after being obtained;

[0036] The sorting adjustment information and the behavior deviation information in the path dynamic feedback information of each source path after preprocessing are extracted, analyzed after being extracted, and the sorting adjustment amplitude index and the click behavior deviation index of each source path are respectively generated;

[0037] An adjustment effect evaluation model is constructed for the generated sorting adjustment amplitude index and click behavior deviation index of each source path, the adjustment evaluation coefficient of each source path is generated, the generated adjustment evaluation coefficient of each source path is compared with the preset adjustment evaluation coefficient threshold of each source path, and according to the comparison result, it is evaluated whether the adjustment effect of the path signal dynamic adjustment mechanism in each source path can reach the expectation, and the path signal dynamic adjustment mechanism is optimized according to the evaluation result.

[0038] Preferably, the acquisition logic of the sorting adjustment amplitude index and the click behavior deviation index of each source path is as follows:

[0039] The sorting adjustment information in the path dynamic feedback information of each source path after preprocessing is extracted, specifically including the sorting weight value of each product in each source path before dynamic adjustment, the sorting weight value after dynamic adjustment, and the change rate of the number of clicks of each product by the user in each source path per unit time before and after adjustment, and they are respectively calibrated as 、 and , Indicates the sorting weight value of the th product in the th source path before dynamic adjustment, Indicates the sorting weight value of the th product in the th source path after dynamic adjustment, Indicates the change rate of the number of clicks on the th product in the th source path by the user per unit time before and after adjustment, , , and are all positive integers;

[0040] Calculate the sorting adjustment amplitude index for each source path. The specific calculation formula is as follows:

[0041] ;

[0042] In the formula, is the sorting adjustment amplitude index of the th source path;

[0043] Extract the behavior offset information in the path dynamic feedback information of each preprocessed source path, specifically including the distribution ratio of the number of clicks on each product by the user in each source path before adjustment, the distribution ratio after adjustment, and the change amount of the display weight value of each product in each source path before and after adjustment, and label them as , and , Indicates the distribution ratio of the number of clicks on the th product by the user in the th source path before adjustment, Indicates the distribution ratio of the number of clicks on the th product by the user in the th source path after adjustment, Indicates the change amount of the display weight value of the th product in the th source path before and after adjustment;

[0044] Calculate the click behavior offset index for each source path. The specific calculation formula is as follows:

[0045] ;

[0046] In the formula, is the click behavior offset index of the th source path.

[0047] Preferably, the sorting adjustment amplitude index of each generated source path and the click behavior deviation index are used to construct an adjustment effect evaluation model, and the adjustment evaluation coefficients of each source path are generated through weighted summation . Then, the adjustment evaluation coefficients of each generated source path are compared with the adjustment evaluation coefficient thresholds of each preset source path . According to the comparison results, it is evaluated whether the adjustment effect of the path signal dynamic adjustment mechanism on each source path can meet the expectations, and the path signal dynamic adjustment mechanism is optimized according to the evaluation results. The specific comparison and analysis are as follows:

[0048] If , the adjustment effect of the path signal dynamic adjustment mechanism on this source path cannot meet the expectations, and the path signal dynamic adjustment mechanism needs to be optimized, specifically including: re-evaluating the sorting weight adjustment rules of the products in the path, increasing the real-time data collection and analysis of the user's latest click behavior and display preferences; dynamically adjusting the weighting parameters in the sorting adjustment amplitude index and the click behavior deviation index to optimize the balance between sorting and click distribution; re-setting the boundary conditions and priority rules for sorting weight adjustment;

[0049] If , the adjustment effect of the path signal dynamic adjustment mechanism on this source path can meet the expectations, and there is no need to optimize the path signal dynamic adjustment mechanism.

[0050] Preferably, a product search optimization system based on historical behavior data includes a user click behavior detection module, a path contribution evaluation module, a path signal adjustment mechanism construction module, a path dynamic feedback evaluation module, and a path signal comprehensive optimization module;

[0051] The user click behavior detection module, by detecting the user's click behavior in real time, identifies whether the user clicks on the same product multiple times through multiple recommended source paths. When it is identified that the user clicks on the same product multiple times through multiple recommended source paths, the multiple source paths generated by the user during the period of searching for this product are classified and marked to distinguish the click behavior characteristics of different source paths;

[0052] The path contribution evaluation module, in real time, obtains the click behavior information of each source path, and after obtaining it, analyzes and evaluates the contribution intensity of each source path to the user's real interest signal, and classifies each source path into a dominant path, a dispersed path, and a mixed path according to the evaluation results;

[0053] The path signal adjustment mechanism construction module constructs a path signal dynamic adjustment mechanism for separately adjusting the sorting weights of the source paths of the dominant path, the dispersed path, and the mixed path according to the division results of each source path;

[0054] The path dynamic feedback evaluation module, during the process of the path signal dynamic adjustment mechanism adjusting each source path, obtains the path dynamic feedback information of each source path in real time, analyzes it after obtaining, evaluates whether the adjustment effect of the path signal dynamic adjustment mechanism in each source path can meet the expectations, and optimizes the path signal dynamic adjustment mechanism according to the evaluation results;

[0055] The path signal comprehensive optimization module comprehensively analyzes the application effect, adjustment records, and optimization results of the path signal dynamic adjustment mechanism, and continuously optimizes the path signal processing ability and dynamic adjustment mechanism of the recommendation sorting model in combination with the new user behavior data.

[0056] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0057] 1. By introducing the path signal dynamic adjustment mechanism, the present invention can accurately distinguish the click behavior characteristics of users in multiple recommended source paths, solve the technical problem in the prior art that the path repeated click signal cannot be recognized. By real-time detecting the user click behavior, classifying and marking at the path level, and quantifying the contribution of each path to the user's true interest signal through the click concentration coefficient and click sparsity index. This fine-grained path modeling enables the system to filter out the interference signals with repeated sources, accurately identify the user's true interest. As a result, the recommendation sorting is more accurate, and the products that the user is interested in can be preferentially displayed, thus significantly improving the click-through rate and conversion rate of the recommendation system.

[0058] 2. The present invention combines the sorting adjustment amplitude index and the click behavior deviation index to comprehensively evaluate the effect of the path signal dynamic adjustment mechanism in sorting optimization and click behavior guidance. By constructing a dynamic adjustment mechanism, different sorting weight adjustment rules are set for different paths according to the path classification results. This mechanism preferentially recommends the products of the user's clear interest in the dominant path, increases the weight of the unclicked products in the dispersed path to improve the diversity of the recommended content, and at the same time achieves a dynamic balance between accuracy and diversity in the mixed path. In this way, the system can not only meet the user's immediate interest needs but also guide the user to explore more potential interests, improving the content coverage rate of the recommendation system and the overall satisfaction of the user.

[0059] 3. Through the real-time collection and optimization of the path dynamic feedback information, the present invention constructs the adaptive optimization ability of the recommendation system. During the application of the adjustment mechanism, the system can dynamically monitor the newly added click behaviors of users and the changes in path signals, and continuously optimize the parameters and rules of the path signal dynamic adjustment mechanism through the adjustment effect evaluation model. At the same time, continuously optimize the recommendation ranking model by combining the newly added user behavior data, so that the recommendation system can quickly respond to the changing trends of user interests and avoid the obsolescence and simplification of recommendation results. As a result, the recommendation system shows higher flexibility and continuous optimization ability in complex and changeable user demand scenarios, bringing higher user retention rate and commercial value to the platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0061] Figure 1 Schematic flowchart of the method and system for optimizing product search based on historical behavior data of the present invention.

[0062] Figure 2 Schematic diagram of the modules of the method and system for optimizing product search based on historical behavior data of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] Now, the exemplary embodiments will be described more fully with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these exemplary embodiments are provided so that the present disclosure will be more complete and comprehensive, and will fully convey the concept of the exemplary embodiments to those skilled in the art.

[0064] The present invention provides a method for optimizing product search based on historical behavior data as Figure 1 shown, which specifically includes the following steps:

[0065] By real-time detecting user click behaviors, identifying whether the user clicks on the same product multiple times through multiple recommendation source paths, and in the case of identifying that the user clicks on the same product multiple times through multiple recommendation source paths, classifying and marking the multiple source paths generated by the user during the period of searching for the product to distinguish the click behavior characteristics of different source paths;

[0066] By detecting the user's click behavior in real time, it can be identified whether the user clicks on the same product multiple times through multiple recommended source paths. This can be achieved by recording the product ID, click source path (such as search results, recommendation lists, ad positions, etc.) and click timestamp in the user's click data. The system updates the user's click sequence in real time through the click logs within the session, and when the same product ID appears in different source paths, it combines the timestamp to determine whether it conforms to the logic of multi-path clicks (such as the click time interval is within the preset range). When the product ID appears repeatedly in multiple paths and the time interval is reasonable, the system determines that the user has clicked on the same product multiple times through multiple source paths. This process can be completed by automatically matching the path attributes and time attributes of the click records without manual intervention.

[0067] In the case where it is identified that the user clicks on the same product multiple times through multiple recommended source paths, the system will classify and mark multiple source paths generated during the period when the user searches for the product to distinguish the click behavior characteristics of different source paths. This process is achieved by grouping the click logs by source path. After grouping, the system marks the click frequency, time distribution, product attribute characteristics, etc. of each path, and generates a path classification table to reflect the independent characteristics of each path. The classification and marking provide support for the calculation of subsequent path parameters, such as evaluating the contribution degree of the path to the user's interest signal, and laying a foundation for the optimization of recommendation ranking.

[0068] The purpose of doing this is to address the defect in the prior art that the click source path is not independently modeled. By classifying and modeling the path signals, the problem of click behavior signal confusion is solved. The core of the technical problem is that the repeated clicks on the path are wrongly superimposed as a strong interest signal for the product, resulting in the recommendation system amplifying the weight of this product while ignoring the potential interest in other products. This design ensures that the system can independently analyze the path contribution signals by accurately identifying and marking the click behavior characteristics of different source paths, provides data support for the subsequent dynamic adjustment mechanism, thereby improving the relevance and diversity of recommendation ranking, and effectively solving the problems of redundant recommended content and unbalanced ranking.

[0069] Obtain the click behavior information of each source path in real time, and analyze it after obtaining. Evaluate the contribution intensity of each source path to the user's real interest signal, and classify each source path into a dominant path, a dispersed path, and a mixed path according to the evaluation results;

[0070] In this embodiment, obtaining the click behavior information of each source path in real time, and analyzing it after obtaining. Evaluate the contribution intensity of each source path to the user's real interest signal, and classify each source path into a dominant path, a dispersed path, and a mixed path according to the evaluation results, specifically including the following steps:

[0071] Obtain the click behavior information of each source path in real time and perform preprocessing after obtaining it;

[0072] Obtaining the click behavior information of each source path in real time can be achieved by deploying real-time data collection modules at the front end and back end of the recommendation system. The front-end module captures click behavior information in real time through the user's interaction logs, including the clicked product ID, source path identifier (such as search results, recommendation lists, ad positions), click timestamp, and user session ID, etc. Subsequently, the data is sent to the back-end data processing pipeline through event stream technologies (such as Kafka or other message queues). In the back end, the system preliminarily classifies the received click events according to the source path identifier and stores them in a distributed database. To achieve real-time performance, the back end can use a memory computing framework (such as Spark Streaming) to process streaming data to ensure that the path click behavior is captured and recorded immediately after the user's click occurs.

[0073] The purpose of preprocessing is to provide consistency and integrity for subsequent data extraction and analysis, and to avoid interference from noise or outliers in the original data on the path contribution evaluation. Preprocessing includes the following steps: First, perform an integrity check on the obtained click data, such as detecting whether there are missing key fields (such as product ID or source path); Second, perform data deduplication to ensure that duplicate click records do not affect subsequent analysis; Third, standardize the timestamp format of click behaviors, unifying time data from different sources into a consistent time zone and format; Finally, remove abnormal data through a filtering algorithm, such as detecting whether multiple repeated clicks within a very short time are misoperations or automated behaviors. The implementation of preprocessing can be completed in the back end through software logic, such as embedding data verification and cleaning modules in the data pipeline to verify and correct each click record before entering the analysis process of the next stage.

[0074] Extract the path click distribution information and path click feature information from the click behavior information of each source path after preprocessing, and perform analysis after extraction to generate the click concentration coefficient and click sparsity index of each source path respectively;

[0075] Construct a path contribution intensity model with the generated click concentration coefficient and click sparsity index of each source path, generate the contribution evaluation coefficient of each source path, compare the generated contribution evaluation coefficient of each source path with the preset contribution evaluation coefficient threshold interval, and evaluate the contribution intensity of each source path to the user's true interest signal according to the comparison result, and divide each source path into a dominant path, a dispersed path, and a mixed path according to the evaluation result.

[0076] The pre-set contribution evaluation coefficient threshold interval can be dynamically determined in a data-driven manner by combining historical click behavior data and model optimization results. The specific method is as follows: First, collect a large amount of historical click data, covering the click behavior characteristics of users in different source paths, including the distribution of the path click concentration coefficient and the path click sparsity index. By analyzing this historical data, use a clustering algorithm (such as K-means) to group the source paths, and automatically divide the evaluation coefficient ranges corresponding to the dominant paths, dispersed paths, and mixed paths. After completing the grouping, calculate the statistical characteristics (such as mean, standard deviation, etc.) of the evaluation coefficients for each group of paths, which are used as the initial reference for the threshold interval. In addition, through an optimization algorithm (such as Bayesian optimization or grid search), combined with the actual effect of the recommendation ranking model, dynamically adjust the upper and lower limits of the threshold to improve the classification accuracy and recommendation performance. This process can be completed through an offline training pipeline in the background, and the threshold interval is updated regularly to ensure that the system can adapt to changes in user behavior patterns.

[0077] In this embodiment, the acquisition logic of the click concentration coefficient and click sparsity index for each source path is as follows:

[0078] Extract the path click distribution information in the click behavior information of each pre-processed source path, specifically including the number of clicks of users on each product in each source path at different times within a period of time, the total number of clicks, and the deviation value between the number of clicks on each product and the total number of clicks in the source path. And represent them respectively in time series using functions , and for representation. is the time point, represents within a period of time the number of clicks of the user on the th product in the th source path, represents within a period of time the total number of clicks of the user in the th source path, represents within a period of time the deviation value between the number of clicks of the user on the th product in the th source path and the total number of clicks in the source path. , , and are all positive integers, and the defined time period is ;

[0079] To extract the path click distribution information from the click behavior information of each pre - processed source path, it can be achieved through a click log analysis process based on a distributed data processing framework (such as Apache Spark or Flink). First, group the real - time collected click logs by source path, and divide each click record into the corresponding time period (such as every minute or every hour) according to the timestamp. In the group of each source path, count the products clicked by users, extract the click times of each product within a specific time period, and accumulate the total click times within this time period. In addition, by calculating the ratio of the click times of each product to the total click times within the source path, the relative distribution of product clicks can be further obtained, thus generating the deviation value between the click times and the total click times. The entire extraction process is implemented through an automated data flow processing pipeline to ensure real - time and accuracy.

[0080] The quantitative data required to obtain the path click distribution information can be collected and stored in real - time from user interaction data through a pre - deployed logging system. Specifically included are: the click times data is obtained by recording the product ID of each click behavior and counting it; the total click times data is obtained by accumulating and counting all click records within the source path over a period of time; the deviation value data is calculated based on the difference between the click times of each product and the total click times within the source path, specifically as ∣product click times−total click times / total number of products∣. To ensure the accuracy and consistency of this data, the logging system needs to record the path, timestamp, and product information of each click event in real - time and maintain it using distributed storage, thus ensuring efficient calculation and dynamic update during the data extraction process.

[0081] Calculate the click concentration coefficient of each source path. The specific calculation formula is as follows:

[0082] ;

[0083] In the formula, is the click concentration coefficient of the th source path;

[0084] This formula calculates the click concentration coefficient through integration to comprehensively and dynamically reflect the concentration degree of user click behavior of each source path over a period of time. The design of the formula combines multi - dimensional data such as click times, total click volume, and deviation value, and synthesizes the characteristics of these dimensions into a quantifiable concentration coefficient through step - by - step operations. First, represents the click times of the th product within the time , which is the basic data of click behavior within the path. By multiplying by Further reduces the impact of high-deviation-value products on centrality (the larger the deviation value, the greater the difference in the uniformity between the clicks on this product and the total distribution of path clicks). The denominator The total number of clicks within the normalized path ensures that the formula is applicable to source paths with different click magnitudes. The time-period integration operation dynamically captures the changes in centrality over the time dimension by calculating data at consecutive time points, thereby generating a click centrality coefficient that comprehensively reflects the path centrality. This design can accurately distinguish whether the user's interest expression for different paths is concentrated, especially applicable to paths with high-frequency clicks, and helps to evaluate the true level of the path contribution intensity.

[0085] The click centrality coefficient of the source path directly reflects the degree of concentration of the user's click behavior within this path, and is thus closely related to the contribution intensity of this path to the user's true interest signal. When the click centrality coefficient is large, it indicates that the user's click behavior is more concentrated on a few products, which usually means that the user has a higher interest in these products and the path's contribution to the user's true interest signal is more significant; on the contrary, when the click centrality coefficient is small, it indicates that the user's click behavior is dispersed among multiple products, which may reflect that the user's exploration behavior within the path is more than the explicit interest expression, and the path's contribution to the user's true interest signal is relatively low. Therefore, the click centrality coefficient provides a quantitative means to help the system identify whether the path reflects the user's true interest, and thus provides data support for optimizing the recommendation ranking model.

[0086] Extract the path click feature information from the click behavior information of each preprocessed source path, specifically including the total number of different products clicked by the user within each source path at different times within a period of time, the number of product categories, and the average value of the number of clicks on all products by the user within each source path, and respectively represent them with functions 、 and in time series, is the time point, represents the total number of different products clicked by the user within the th source path at the th moment within a period of time, represents the number of product categories of different products clicked by the user within the th source path at the th moment within a period of time, represents the average value of the number of clicks on all products by the user within the th source path at the th moment within a period of time;

[0087] The path click feature information in the click behavior information of each pre - processed source path can be achieved through a distributed data processing framework (such as Apache Spark or Flink) combined with grouping and aggregation operations. First, the real - time collected click log data is grouped by source path and time period (such as every hour or every minute). Within each source path group, the total number of different products clicked by users is counted by removing duplicates of product IDs, and at the same time, the total number of product categories is counted by removing duplicates of the product category field. In addition, the average value of the click times is calculated by accumulating the total number of clicks within the source path and dividing it by the number of different products. This extraction process depends on fields such as product ID, product category, and timestamp included in the click log, and the path click feature information is updated in real - time through pipelining to ensure that the system dynamically extracts users' click behavior characteristics in different time periods.

[0088] The quantitative data in the path click feature information comes from the pre - processed click log. Specifically, the total number of different products clicked is obtained by counting the number of unique product IDs within the same source path, and the de - duplication logic can be based on the uniqueness of the product ID field; the number of product categories is obtained by removing duplicates and counting the product category field in the click log. The product category field is usually the associated information of the product ID, and the system can dynamically obtain and update it according to the product's metadata table; the average value of the number of clicks on all products by users within the path is calculated by accumulating the total number of all click records within the source path and then dividing it by the total number of different products. The above data depends on the real - time records of the click log, including core fields such as product ID, product category, and click time. These fields are collected by the log module of the system front - end and stored in the background. After being statistically analyzed and aggregated by the distributed data processing framework, they form the quantitative data of the path click feature information.

[0089] Calculate the click sparsity index for each source path. The specific calculation formula is as follows:

[0090] ;

[0091] In the formula, is the click sparsity index of the th source path.

[0092] This formula calculates the click sparsity index through integration, aiming to dynamically quantify the dispersion degree of users' click behavior in a certain source path, and comprehensively consider factors such as the total number of products, product category distribution, and click uniformity to comprehensively describe the sparsity of clicks within the path. First, represents the total number of products clicked within the path, while denotes the number of product categories, and the product of the two reflects the breadth of path clicks and category diversity, which defines an important basis for sparsity; then, the denominator part introduces , which measures the uniformity of user click behavior by the absolute value of the deviation between the click count and the average value. The greater the deviation, the lower the sparsity. The summation and overall normalization adjustment in the denominator enable the calculation result to balance the click volume differences of different paths, while the integration operation dynamically captures the click distribution characteristics in the time dimension. The overall formula comprehensively reflects the sparsity characteristics of path clicks through multiple variables, making it have good generality under different time periods, product categories, and click magnitudes, and can scientifically quantify the contribution of paths to users' interest exploration behavior.

[0093] The click sparsity index of the th source path directly reflects the degree of dispersion of users' click behavior within this path, thereby affecting the evaluation of the contribution intensity of the true interest signal of this path. When the click sparsity index is relatively high, it indicates that users have clicked on a relatively large number of different products within this path, and the product category distribution is relatively wide, and at the same time, the uniformity of click behavior is relatively high. This usually means that this path is more used for exploratory behavior, and the users' interest signals are not clear enough, and the contribution to the true interest is relatively low; on the contrary, when the click sparsity index is relatively low, it indicates that users' click behavior is less dispersed and concentrated on certain specific products or categories, which may reflect clearer interest signals. Therefore, this path contributes relatively more strongly to the users' true interest signals. The click sparsity index provides a key dimension for the analysis of the interest signal of the path by quantifying the dispersion, complements the click concentration coefficient, and jointly evaluates the overall contribution intensity of the path to the interest signal.

[0094] In this embodiment, the click concentration coefficients and click sparsity indices of each generated source path are constructed to form a path contribution intensity model, and the contribution evaluation coefficients of each source path are generated through weighted summation, and the generated contribution evaluation coefficients of each source path are compared with the pre-set contribution evaluation coefficient threshold interval , and the contribution intensity of each source path to the users' true interest signal is evaluated according to the comparison result, and each source path is divided into a dominant path, a dispersed path, and a mixed path according to the evaluation result. The specific comparison analysis and division are as follows:

[0095] If , the contribution intensity of this source path to the users' true interest signal is low intensity, and this source path is divided into a dispersed path;

[0096] This situation means that the user's click behavior within this source path is relatively scattered, and it may be more manifested as exploratory operations on different products or categories, rather than a clear expression of interest. Such a path usually reflects that the user's interests have not yet converged, or the user is trying to understand more product options through extensive click behavior. Therefore, the impact of this path on the recommendation ranking is that the system needs to reduce the weight of repeatedly recommended products and increase product diversity to match the user's exploration needs, thereby enhancing the richness of the user experience.

[0097] If , and the contribution intensity of this source path to the user's true interest signal is medium intensity, then this source path is classified as a mixed path;

[0098] This situation indicates that the user's click behavior within this path contains both a certain degree of interest concentration and shows a certain degree of exploration. The user's true interest signal is between clear and vague. Such a path usually represents that the user has some preferences for specific products but is still in the process of exploring other options. For such a path, the recommendation system needs to maintain a certain weight for the products of interest while appropriately increasing diverse recommendations to balance the user's demand for specific products and potential interest in new products, optimizing the coverage of the recommendation ranking and user satisfaction.

[0099] If , and the contribution intensity of this source path to the user's true interest signal is high intensity, then this source path is classified as a dominant path.

[0100] This situation means that the user's click behavior within this source path is highly concentrated on a few products or categories, indicating that the user's interest signal is very clear, and the products within the path have strong attraction to the user. The impact of this path on the recommendation ranking is that the system should give priority to increasing the recommendation weight of the products within this path, centrally display the products that the user is obviously interested in, maximize the matching of the user's current needs, thereby improving the accuracy of the recommendation and the click-through conversion rate. Such a path is of great significance for meeting the user's immediate needs and is also more prominent for the commercial value of the platform.

[0101] The click concentration coefficient and click sparsity index generated for each source path are used to construct a path contribution intensity model, and the contribution evaluation coefficient of each source path is generated through weighted summation. The specific method is: First, determine the weight coefficients and 。The selection of weight coefficients can be optimized through a historical data-driven approach. For example, through regression analysis based on user click data or obtaining the optimal weight values through model training to ensure a reasonable balance in the contribution evaluation. Then, use the formula to perform a weighted sum of the click concentration coefficient and the click sparsity index, where reflects the influence weight of concentration on the path contribution intensity, while reflects the influence weight of sparsity on the path contribution intensity. Generally speaking, if the recommendation system pays more attention to the user's explicit interests (such as quickly meeting the user's needs), the weight of can be appropriately increased; if it pays more attention to the user's exploration behavior and diverse needs, the weight of can be appropriately increased. Through this method, the model can comprehensively reflect the concentration and sparsity characteristics of the path and effectively quantify the contribution intensity of the path to the user's true interest signal.

[0102] According to the division results of each source path, construct a path signal dynamic adjustment mechanism for separately adjusting the sorting weights of the source paths of the dominant path, the dispersed path, and the mixed path;

[0103] In this embodiment, according to the division results of each source path, construct a path signal dynamic adjustment mechanism for separately adjusting the sorting weights of the source paths of the dominant path, the dispersed path, and the mixed path. Specifically:

[0104] According to the division results of the dominant path, the dispersed path, and the mixed path, respectively set different sorting weight adjustment rules to form a path signal dynamic adjustment mechanism; this adjustment mechanism, based on the contribution evaluation coefficient of each source path and the optimization goal of the recommendation system, automatically determines the sorting weight adjustment method and amplitude of the products within each path through pre-set rules;

[0105] To achieve the goal of "setting different sorting weight adjustment rules according to the division results of the dominant path, the scattered path, and the hybrid path, and forming a dynamic adjustment mechanism for path signals", it can be realized through a dynamic rule engine combined with a real-time data processing module. The specific method is as follows: First, a set of sorting weight adjustment rules are preset in the system. These rules are defined based on the characteristics of the path types (dominant path, scattered path, hybrid path). For example, the rule for the dominant path gives priority to increasing the sorting weight of the clicked products, while the rule for the scattered path reduces the weight of repeatedly clicked products and enhances the diversity weight of unclicked products. Second, in the background of the recommendation system, the type of each source path is determined in real time through a contribution evaluation coefficient and mapped to the corresponding sorting weight adjustment rule. The adjustment rule takes the contribution evaluation coefficient as an input parameter and, combined with the optimization goals of the recommendation system (such as the balance between accuracy and diversity), automatically generates the adjustment strategy and amplitude of the sorting weight. Finally, the sorting weight of the products in the recommendation model is updated in real time through the dynamic rule engine, so as to reflect the characteristics of the path type and the user's behavior pattern in the recommendation results. The purpose of doing this is to make full use of the contribution differences of different paths to the user interest signal, give priority to recommending the products with clear interests in the dominant path, enhance exploratory recommendations in the scattered path, and achieve a dynamic balance between accuracy and diversity in the hybrid path, thereby improving the user satisfaction and overall efficiency of the recommendation system.

[0106] Different sorting strategy adjustments are made for the source paths of the dominant path, the scattered path, and the hybrid path respectively, which specifically include: for the dominant path, the weight increase rule in the adjustment mechanism is used to increase the sorting weight of the clicked products in the path and give priority to displaying the products of the user's clear interests; for the scattered path, the weight dispersion rule in the adjustment mechanism is used to reduce the sorting weight of the repeatedly clicked products and at the same time increase the display weight of other unclicked products in the path to increase the diversity of the recommended content; for the hybrid path, the weight balance rule in the adjustment mechanism is used to maintain the sorting weight of the clicked products in the path to balance the accuracy and diversity of the recommendation.

[0107] To achieve different sorting strategy adjustments for the dominant path, the scattered path, and the hybrid path, it can be realized through the dynamic sorting weight adjustment module in combination with real-time behavior data analysis. The specific method is as follows: First, design a weight adjustment function library in the recommendation system, and the function library contains three adjustment rules for path types. For the dominant path, the system identifies the products with highly concentrated user interests through the contribution evaluation coefficient and uses the "weight increase rule" to combine the click weights of these products with user behavior data (such as click times and time frequencies) to increase their sorting weights, so as to preferentially display the products with clear user interests. For the scattered path, the system adopts the "weight dispersion rule". By analyzing the repeated click times of the products in the path and the attribute data of the unclicked products (such as product categories and similarities), it reduces the sorting weights of the repeatedly clicked products and increases the display weights of other unclicked products in the path, so as to improve the diversity and coverage of the recommended content. For the hybrid path, the system uses the "weight balance rule". By performing weighted smoothing on the click weights of all products in the path (such as assigning importance weights based on product categories), and at the same time moderately balancing the display frequencies of the clicked products and the unclicked products, it takes both accuracy and diversity into account. During the implementation process, the weight adjustment module continuously updates the sorting weights of each product through real-time data streams (such as user click logs and product attributes), and dynamically applies these weights during the recommendation generation stage, so that the recommendation results always reflect the latest behavior patterns and path characteristics of users. The purpose of such a design is to maximize user satisfaction, improve the accuracy of recommendations through the dominant path, optimize the diversity of recommended content through the scattered path, and achieve the dynamic balance between diversity and accuracy through the hybrid path, so as to comprehensively improve the overall effect of the recommendation system.

[0108] During the process of the path signal dynamic adjustment mechanism adjusting each source path, the path dynamic feedback information of each source path is obtained in real time, analyzed after being obtained, evaluated whether the adjustment effect of the path signal dynamic adjustment mechanism in each source path can reach the expectation, and the path signal dynamic adjustment mechanism is optimized according to the evaluation result;

[0109] In this embodiment, during the process of the path signal dynamic adjustment mechanism adjusting each source path, the path dynamic feedback information of each source path is obtained in real time, analyzed after being obtained, evaluated whether the adjustment effect of the path signal dynamic adjustment mechanism in each source path can reach the expectation, and the path signal dynamic adjustment mechanism is optimized according to the evaluation result, which specifically includes the following steps:

[0110] During the process of the path signal dynamic adjustment mechanism adjusting each source path, the path dynamic feedback information of each source path is obtained in real time and preprocessed after being obtained;

[0111] Real-time acquisition of path dynamic feedback information from various source paths can be achieved through the front-end behavior monitoring module and the back-end real-time data processing system. The front-end monitoring module is responsible for capturing real-time user behavior data in various paths, including the timestamp of clicking on a product, the number of clicks, the product ID, the number of displays, and the sorting weight, etc. These data are recorded through data embedding technology or event logging systems. The back-end data processing system receives and integrates the real-time data from the front-end through a streaming data processing framework (such as Apache Kafka or Flink), and transmits the click behavior, display weight adjustment, and sorting change logs of the products within the path to the central data storage in a streaming pipeline manner for subsequent processing and analysis. In this way, the immediate effect of path signal adjustment can be dynamically tracked, providing high-quality raw data for subsequent evaluation.

[0112] The purpose of preprocessing is to improve the consistency, integrity, and analysis efficiency of data, ensuring that the dynamic feedback information can accurately reflect the actual effect of path adjustment in subsequent steps. Preprocessing mainly includes the following steps: First, perform data deduplication to filter out redundant data caused by network latency or repeated system sending; Second, perform integrity verification to ensure that each feedback record contains necessary fields (such as product ID, timestamp, sorting weight, etc.), and discard or complete the data with missing fields; Third, perform standardization processing to unify the timestamp into the standard time zone format and normalize the sorting weight to the same numerical range; Fourth, remove outliers by setting a reasonable range to filter out obviously abnormal data such as click frequency and number of displays (such as a large number of clicks or displays within a very short time). These preprocessing steps can be completed through the back-end automated data pipeline, combined with data validation rules and cleaning algorithms, to process each piece of dynamic feedback information in the real-time data stream one by one, ensuring that the data entering the subsequent analysis steps is accurate and reliable.

[0113] Extract the sorting adjustment information and behavior deviation information from the path dynamic feedback information of each source path after preprocessing, and conduct analysis after extraction to generate the sorting adjustment amplitude index and click behavior deviation index for each source path respectively;

[0114] Construct an adjustment effect evaluation model for the generated sorting adjustment amplitude index and click behavior deviation index of each source path, generate the adjustment evaluation coefficient for each source path, compare the generated adjustment evaluation coefficient for each source path with the pre-set adjustment evaluation coefficient threshold for each source path, and evaluate whether the adjustment effect of the path signal dynamic adjustment mechanism in each source path can meet the expectations according to the comparison result, and optimize the path signal dynamic adjustment mechanism according to the evaluation result.

[0115] The pre-set thresholds of adjustment evaluation coefficients for each source path can be dynamically determined by combining historical data analysis and model optimization. The specific method is to first collect a large amount of historical data, including the ranking adjustment amplitude index and click behavior deviation index generated by each source path under different adjustment mechanisms, as well as the adjustment evaluation coefficients and recommendation effects (such as click-through rate and conversion rate) of these paths. Then, cluster analysis algorithms (such as K-means) are used to group historical path data, and the path evaluation coefficients are divided into high-efficiency areas, low-efficiency areas, and intermediate areas according to the quality of the recommendation effect. The statistical characteristics of each group of evaluation coefficients (such as mean and standard deviation) are calculated to determine the preliminary threshold range. In addition, the threshold range is further adjusted through machine learning models (such as Bayesian optimization) to keep it consistent with the actual goals of the recommendation system (such as accuracy and diversity balance) when evaluating the effect of the adjustment mechanism. The entire process can be achieved through the background offline analysis module and real-time tuning module. Offline analysis is used to generate the basic threshold range, and real-time tuning dynamically updates the threshold in combination with the newly added data to ensure that the adjustment evaluation coefficient threshold can continuously reflect the system optimization needs.

[0116] In this embodiment, the logic for obtaining the ranking adjustment index and click behavior deviation index of each source path is as follows:

[0117] Extract the sorting adjustment information from the preprocessed dynamic feedback information of each source path, including the sorting weight value of each product in each source path before dynamic adjustment, the sorting weight value after dynamic adjustment, and the rate of change of the number of clicks of each product in each source path before and after adjustment, and mark them as , and , Indicates Source path The ranking weight value of a product before dynamic adjustment. Indicates Source path The ranking weight value of each product after dynamic adjustment. Indicates that the user is The source path is The rate of change of the number of clicks on a product per unit time before and after adjustment, , , and All are positive integers;

[0118] The sorting adjustment information in the path dynamic feedback information of each preprocessed source path can be achieved by combining front-end user behavior data tracking and back-end sorting weight logging with a real-time data processing framework. Specifically, the front-end data tracking records the user's interaction behaviors within the path (such as clicks, dwell time, etc.) and sends these behavior data to the back-end in real time. When the path dynamic adjustment mechanism takes effect, the back-end sorting model will update the sorting weight of each product in real time and record the sorting weight values before and after the adjustment in the log. At the same time, through streaming data processing tools (such as Apache Kafka or Flink), the click counts of each product in the real-time log and their corresponding timestamps are integrated, and the click change rate per unit time before and after the adjustment is calculated, and these data are marked as "sorting adjustment information". These extraction steps are completed through an automated pipeline in the back-end to ensure real-time data processing and synchronization with the dynamic adjustment mechanism.

[0119] The sorting adjustment information in the path dynamic feedback information consists of the following types of quantitative data: 1. Sorting weight values (before and after adjustment): Obtained from the weight allocation results of the sorting model before and after each path adjustment. Specifically, the sorting weight values are recorded in the system's log file. Each time the model calls the sorting algorithm, the generated weight values are synchronously written into the log to mark the priority of each product in the path; 2. Click count records: Real-time statistics through user click behavior data tracking. Each click event is recorded, including product ID, click timestamp, etc.; 3. Click change rate: Calculated in real time by the back-end. The formula is change rate = (click count after adjustment - click count before adjustment) / time interval. The time interval is calculated from the front and back timestamps, and the click count is directly extracted from the data tracking data. These data are calculated and stored in real time through a streaming processing pipeline, providing accurate input for the subsequent calculation of the sorting adjustment amplitude index.

[0120] Calculate the sorting adjustment amplitude index for each source path. The specific calculation formula is as follows:

[0121] ;

[0122] In the formula, is the sorting adjustment amplitude index of the th source path;

[0123] This formula quantifies the change degree of the sorting weight of products in the path before and after dynamic adjustment and the correlation between this change and the dynamic change of user click behavior by calculating the sorting adjustment amplitude index . Specifically, the core of the formula design is to comprehensively and dynamically capture the amplitude of the sorting weight adjustment, and at the same time combine the click change rate to reflect the effect of the adjustment. In the formula, represents the The absolute value of the change in the sorting weight of a product before and after adjustment, which is used to measure the impact of the adjustment mechanism on the product; the denominator is the normalization term of the sorting weight, which ensures that the formula is applicable to products with different weight levels and avoids the excessive influence of weights with larger absolute values on the results. The product term introduces the click change rate , which is non-linearly amplified through a logarithmic function, so that the change in the user's click behavior has a moderate amplification effect on the contribution of the exponent, while avoiding the disruption of the formula calculation by extreme click rate values. The summation of all products in the formula reflects the overall sorting adjustment amplitude of the path and dynamically captures the comprehensive optimization effect of the path after adjustment. Through this calculation method, it is possible to accurately evaluate the actual amplitude of the sorting weight adjustment and whether it is consistent with the change in user behavior, thereby providing a scientific basis for the subsequent optimization mechanism.

[0124] The sorting adjustment amplitude index of the th source path directly reflects the adjustment amplitude of the sorting weight of the products in the path by the path signal dynamic adjustment mechanism and its correlation with the change in the user's click behavior, and is thus closely related to whether the adjustment effect meets the expectations. When is large, it indicates that the adjustment mechanism has a large adjustment amplitude on the sorting weight of the products in the path, and this adjustment is closely related to the dynamic change of the user's click behavior, indicating that the adjustment mechanism is relatively successful in optimizing the path sorting and can significantly improve the accuracy or diversity of recommendations; while when is small, it indicates that the adjustment amplitude is small or the sorted order after adjustment fails to fully reflect the change in the user's click behavior, which may mean that the adjustment mechanism fails to function effectively. By analyzing the size of , it is possible to accurately evaluate whether the path signal dynamic adjustment mechanism has achieved the expected optimization effect on this path and adjust the mechanism parameters or optimization strategies accordingly.

[0125] Extract the behavior offset information in the path dynamic feedback information of each preprocessed source path, specifically including the distribution ratio of the number of clicks of each product by the user in each source path before adjustment, the distribution ratio after adjustment, and the change amount of the display weight value of each product in each source path before and after adjustment, and respectively label them as , and , represents the distribution ratio of the number of clicks of the th product by the user in the th source path before adjustment, represents the distribution ratio of the number of clicks of the th product by the user in the The click-through rate distribution ratio of each product after adjustment, represents the change in the display weight value of the th product within the th source path before and after adjustment;

[0126] Extract the behavior deviation information from the path dynamic feedback information of each preprocessed source path, which can be achieved by combining front-end event tracking and background real-time log processing with a streaming data processing framework. Front-end event tracking records each click behavior of users within the path, captures key fields such as product ID, click timestamp, and source path identifier, and sends them to the background in real time. At the same time, the background records the display weight of each product within the path through the sorting model's log and regularly synchronizes the display weight values before and after adjustment. Through a streaming data processing tool (such as Apache Kafka or Flink), the system can associate the front-end user click behavior data with the background display weight change data. Before and after the execution of the path adjustment mechanism, count the click-through rate of products within the path respectively, and calculate the distribution ratio of the click-through rate of each product (i.e., the ratio of the product's click-through rate to the total click-through rate within the path). In addition, by comparing the sorting model logs, extract the change in display weight before and after adjustment. These extraction steps are completed in the background automated data flow pipeline to ensure the real-time and accuracy of the data.

[0127] The quantitative data in the behavior deviation information includes three categories: 1. Click-through rate distribution ratio before adjustment: calculated by counting the click-through rate of each product within the path and dividing it by the total click-through rate within the path. For example, the click-through rate of each product is recorded in the user click log, and the system groups and calculates the distribution ratio according to the click timestamp and path identifier; 2. Click-through rate distribution ratio after adjustment: calculated in the same way by counting the click-through rate of each product within the path after adjustment; 3. Change in display weight value: calculated by comparing the display weights of each product within the path before and after adjustment (recorded in the sorting model log), and the formula is change = weight after adjustment - weight before adjustment. The acquisition of these quantitative data depends on the real-time recording of user behavior event tracking data and sorting model logs, and through the background data pipeline, group, count, and compare the click behavior and display weight to achieve efficient extraction and processing of behavior deviation information.

[0128] Calculate the click behavior deviation index for each source path. The specific calculation formula is as follows:

[0129] ;

[0130] In the formula, is the click behavior deviation index of the th source path.

[0131] This click behavior deviation index The calculation formula aims to quantify the degree of change in the distribution of user click behaviors within the path before and after adjustment, and comprehensively evaluate the optimization effect of the path signal dynamic adjustment mechanism on user click behaviors by combining the impact of the display weight change. The core design of the formula captures the change amplitude of the click distribution ratio of each product, and clearly measures the guiding effect of the adjustment mechanism on the click behaviors of each product within the path. The denominator part introduces the change amount of the display weight , and non-linearly amplifies the impact of the display weight change on the click behavior deviation through a logarithmic function, while suppressing the interference of extreme weight changes on the formula result. This normalization process ensures that the index still has good stability and comparability in scenarios where the path scale and display weight differ greatly. In addition, the formula reflects the distribution deviation degree of the path click behaviors as a whole through the weighted sum of the deviation values of all products, providing a scientific basis for the quantification of the dynamic adjustment effect. Through this calculation method, the effectiveness of the dynamic adjustment mechanism in optimizing the distribution of user click behaviors and improving the accuracy and diversity of recommendations can be clearly evaluated.

[0132] The click behavior deviation index of the nth source path directly reflects the guiding and optimization effect of the path signal dynamic adjustment mechanism on the distribution of user click behaviors within the path, and is thus closely related to whether the adjustment effect meets the expectations. When is relatively large, it indicates that the distribution of user click behaviors within the path has changed significantly before and after adjustment, which may suggest that the adjustment mechanism has effectively reallocated the user's attention, such as reducing the proportion of repeatedly clicked products and increasing the attention to unclicked products, thereby optimizing the diversity of recommended content; while when is relatively small, it indicates that the change in click behavior distribution before and after adjustment is small, which may mean that the adjustment mechanism has not fully guided the user's behavior and the optimization effect of the path signal is limited. By analyzing the size of , it is possible to quantitatively evaluate whether the path signal dynamic adjustment mechanism has successfully achieved the recommendation goals (such as the balance of accuracy and diversity), and provide a basis for the further optimization of the mechanism.

[0133] In this embodiment, an adjustment effect evaluation model is constructed for the sorting adjustment amplitude index and the click behavior deviation index of each generated source path, and an adjustment evaluation coefficient for each source path is generated through weighted summation, and the generated adjustment evaluation coefficients of each source path are compared with the adjustment evaluation coefficient thresholds Compare and evaluate whether the adjustment effects of the path signal dynamic adjustment mechanism on each source path can meet the expectations according to the comparison results, and optimize the path signal dynamic adjustment mechanism according to the evaluation results. The specific comparison and analysis are as follows:

[0134] If , the adjustment effect of the path signal dynamic adjustment mechanism on this source path cannot meet the expectations, and the path signal dynamic adjustment mechanism needs to be optimized, specifically including: re-evaluating the sorting weight adjustment rules for the products within the path, increasing the real-time data collection and analysis of the user's latest click behavior and display preferences; dynamically adjusting the weighting parameters in the sorting adjustment amplitude index and the click behavior deviation index to optimize the balance between sorting and click distribution; re-setting the boundary conditions and priority rules for sorting weight adjustment;

[0135] This situation means that the current path signal dynamic adjustment mechanism fails to effectively optimize the sorting weights of the products within this source path or the distribution of user click behaviors, resulting in a low matching degree between the recommended results and the user interest signals. Specifically, the sorting adjustment amplitude index and the click behavior deviation index show insufficient adjustment amplitude or insignificant distribution changes, which may cause the repeatedly clicked products within the path to be overly prioritized, without fully considering the recommendation potential of other products. In this case, the recommended results may make users feel that the content is monotonous or not relevant enough, reducing the user satisfaction and click conversion rate of the recommendation system. In addition, system resources may be wasted on recommending low-priority products, and the potential interest needs of users cannot be maximally explored. Therefore, the mechanism must be optimized by re-evaluating the rules, dynamically adjusting the parameters, etc. to improve the recommendation effect.

[0136] The re-evaluation of the sorting weight adjustment rules for products within the path, as well as real-time data collection and analysis, can be achieved through a dynamic analysis model based on user behavior and a real-time data stream processing system. Specifically, first, a real-time behavior monitoring system is needed to collect the latest click behaviors and display preference data of users, including click time, click frequency, display times, and product category preferences, etc. Through streaming processing tools (such as Kafka or Flink), this data is transmitted to the background analysis module in real time. The background analysis module constructs a behavior feature model based on the user's click history and re-evaluates the sorting weight adjustment rules for products within the path through algorithms such as clustering analysis or association rule mining. For example, the initial sorting weights of products are dynamically adjusted according to the user's click tendency for certain categories or brands. For the sorting adjustment magnitude index (PMI) and the click behavior deviation index (PCB), their weighting parameters need to be optimized through experimental data. Specifically, the weight ratio of these two indexes in sorting weight optimization and click distribution optimization can be dynamically adjusted through a multi-objective optimization algorithm to achieve a balance between accuracy and diversity. In addition, resetting the boundary conditions and priority rules for sorting weight adjustment can be completed by setting reasonable thresholds and priority algorithms. For example, the minimum and maximum magnitudes of weight adjustment are dynamically defined according to the user's historical preference data to prevent unbalanced recommendation results caused by extreme adjustments.

[0137] By re-evaluating the sorting weight adjustment rules for products within the path, it can be ensured that the sorting rules can timely reflect the latest click behaviors and display preferences of users, thus improving the accuracy of recommended content. At the same time, increasing the real-time data collection and analysis of the latest click behaviors and display preferences of users can dynamically capture the changes in user needs and avoid inaccurate recommendation results caused by data latency. Dynamically adjusting the weighting parameters in the sorting adjustment magnitude index and the click behavior deviation index can optimize the balance between accuracy and diversity of the recommendation system, thus avoiding the problem of overly single or overly scattered recommended content. Resetting the boundary conditions and priority rules for sorting weight adjustment can avoid extreme situations where the sorting adjustment amplitude is too large or too small, ensure that the recommendation results are optimized within a controllable range, and thus improve the stability and user satisfaction of the recommendation system. These optimization measures can comprehensively enhance the response ability and adaptability of the path signal dynamic adjustment mechanism, and ultimately achieve the rapid response and accurate matching of the recommendation system to user needs.

[0138] If , the adjustment effect of the path signal dynamic adjustment mechanism on this source path can reach the expected value, and there is no need to optimize the path signal dynamic adjustment mechanism.

[0139] This situation indicates that the current path signal dynamic adjustment mechanism has a significant adjustment effect on this source path, successfully achieving the optimization of recommended content. Specifically, the sorting adjustment amplitude index and the click behavior deviation index show that the adjustment amplitude of the sorting weights of the products within the path is reasonable and consistent with the changes in user behavior. The deviation of the click distribution indicates that the adjustment mechanism effectively guides users to pay attention to more products or products with higher priorities. In this case, the recommended results can accurately match the user's interest signals while providing diverse content, thereby improving user satisfaction and the click-through rate of the system. The optimized recommended path effectively allocates system resources, ensuring a positive feedback loop between user behavior and recommended content, and further enhancing the efficiency and overall value of the recommendation system.

[0140] For the sorting adjustment amplitude index of each generated source path and the click behavior deviation index Construct an adjustment effect evaluation model, and generate an adjustment evaluation coefficient for each source path through weighted summation The specific method is as follows: First, set the weight coefficients and according to the optimization objectives of the recommendation system (such as precision first, diversity first, or a balance between the two). The weight reflects the importance of the sorting adjustment amplitude index in the evaluation, and is used to measure the direct guiding effect of the sorting adjustment of the products within the path on user click behavior; the weight reflects the importance of the click behavior deviation index in the evaluation, and is used to measure whether the change in the click distribution within the path conforms to the optimization objective. Next, the and of each source path are weighted and summed according to the following formula: ; where , ensuring that the weights of the two are reasonably proportioned in the overall evaluation. The specific values of the weight coefficients can be dynamically adjusted through model training with historical data or using multi-objective optimization algorithms. For example, in the scenario of precision first, the value of can be increased, while in the scenario emphasizing diversity, the value of can be increased. Through this weighted summation method, the adjustment effect evaluation model can comprehensively consider the roles of the path signal dynamic adjustment mechanism in sorting adjustment and click behavior optimization, providing a scientific and quantitative basis for evaluating the overall optimization effect of the path.

[0141] Comprehensively analyze the application effect, adjustment records, and optimization results of the path signal dynamic adjustment mechanism, and continuously optimize the path signal processing ability and dynamic adjustment mechanism of the recommendation sorting model in combination with newly added user behavior data.

[0142] The comprehensive analysis of the application effect, adjustment records, and optimization results of the path signal dynamic adjustment mechanism can be achieved through the background data analysis module in combination with machine learning models. Specifically, first, the application effect data of the dynamic adjustment mechanism (such as the path evaluation coefficient TPG and the click-through rate and conversion rate of the recommendation system), adjustment records (including the change of path adjustment rules and parameters each time), and optimization results (such as the change of recommendation accuracy and diversity) are collected through the log recording system. These data are stored in a centralized database and integrated and analyzed through regularly triggered batch processing tasks or real-time streaming analysis tools (such as Apache Flink). Next, data mining and machine learning methods are used to model the collected data. For example, decision tree or random forest models are used to analyze the impact of different adjustment strategies on user behavior; time series analysis or regression models are used to predict the long-term optimization trend of the path signal dynamic adjustment mechanism. In addition, new user behavior data (such as click, browse, and purchase records) are input into the recommendation ranking model in real time through incremental learning methods to dynamically adjust the path signal processing rules and the parameters of the adjustment mechanism, so as to continuously improve the effect of the recommendation system.

[0143] Through the comprehensive analysis of the application effect, adjustment records, and optimization results, the effect of the path signal dynamic adjustment mechanism in actual applications can be deeply understood, providing a scientific basis for further optimizing the recommendation system. This analysis can help identify potential problems in the current adjustment strategy, such as whether there are over-adjustment or under-adjustment problems in the path signal processing rules, and whether the optimization results conform to the dynamic change trend of user needs. Combining new user behavior data can enhance the dynamic response ability of the recommendation ranking model, ensuring that the recommended content always matches the latest interest preferences of users. At the same time, continuously optimizing the path signal processing ability and dynamic adjustment mechanism helps to achieve the adaptive learning of the recommendation system, avoiding the attenuation of the recommendation effect caused by fixed rules, thereby improving the long-term performance and user satisfaction of the recommendation system. This approach not only improves the recommendation accuracy and diversity but also effectively enhances the flexibility and continuous improvement ability of the recommendation system.

[0144] such as Figure 2 The commodity search optimization system based on historical behavior data as shown includes a user click behavior detection module, a path contribution evaluation module, a path signal adjustment mechanism construction module, a path dynamic feedback evaluation module, and a path signal comprehensive optimization module;

[0145] The user click behavior detection module, by detecting the user's click behavior in real time, identifies whether the user clicks on the same commodity multiple times through multiple recommendation source paths. And in the case of identifying that the user clicks on the same commodity multiple times through multiple recommendation source paths, it classifies and marks the multiple source paths generated by the user during the period of searching for the commodity, used to distinguish the click behavior characteristics of different source paths;

[0146] A path contribution evaluation module that obtains the click behavior information of each source path in real time, analyzes it after acquisition, evaluates the contribution intensity of each source path to the user's true interest signal, and classifies each source path into a dominant path, a dispersed path, and a mixed path according to the evaluation results;

[0147] A path signal adjustment mechanism construction module that constructs a dynamic path signal adjustment mechanism for separately adjusting the sorting weights of the dominant path, the dispersed path, and the mixed path source paths according to the classification results of each source path;

[0148] A path dynamic feedback evaluation module that obtains the path dynamic feedback information of each source path in real time during the process of the path signal dynamic adjustment mechanism adjusting each source path, analyzes it after acquisition, evaluates whether the adjustment effect of the path signal dynamic adjustment mechanism on each source path can meet the expectations, and optimizes the path signal dynamic adjustment mechanism according to the evaluation results;

[0149] A path signal comprehensive optimization module that comprehensively analyzes the application effect, adjustment records, and optimization results of the path signal dynamic adjustment mechanism, and continuously optimizes the path signal processing ability and dynamic adjustment mechanism of the recommendation sorting model in combination with the newly added user behavior data.

[0150] The above formulas are all dimensionless and take their numerical calculations. The formula is a formula obtained by collecting a large amount of data for software simulation to approximate the real situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0151] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired or wireless means (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0152] It should be understood that in various embodiments of the present application, the order numbers of the above processes do not indicate the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0153] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0154] In several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the above-described embodiments are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be indirect couplings or communication connections through some interfaces, devices, or units, and can be electrical, mechanical, or other forms.

[0155] The unit described as a separation component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, that is, it may be located in one place or may be distributed across multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0156] In addition, each functional unit in various embodiments of the present application may be integrated in a processing unit, may exist separately as individual physical units, or two or more units may be integrated in one unit.

[0157] As described above, only the specific implementation manners of the present application are provided, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A commodity search optimization method based on historical behavior data, characterized in that: The specific steps include: By detecting the user's click behavior in real time, it is identified whether the user has clicked on the same product multiple times through multiple recommendation source paths. If it is identified that the user has clicked on the same product multiple times through multiple recommendation source paths, the multiple source paths generated by the user during the time period of searching for the product are classified and marked to distinguish the click behavior characteristics of different source paths. Obtain the click behavior information of each source path in real time, analyze it after acquisition, evaluate the contribution strength of each source path to the user's real interest signal, and divide each source path into dominant path, decentralized path and mixed path according to the evaluation results; The specific steps include: Obtain click behavior information of each source path in real time and perform preprocessing after acquisition; Extracting the path click distribution information and path click feature information from the preprocessed click behavior information of each source path, and analyzing them after extraction to generate the click concentration coefficient and click sparse index of each source path respectively; The logic for obtaining the click concentration coefficient and click sparse index of each source path is as follows: Extract the path click distribution information from the preprocessed click behavior information of each source path, including the number of clicks on each product in each source path at different times within a period of time, the total number of clicks, and the deviation between the number of clicks on each product and the total number of clicks in the source path, and use the function , and To express, For time point, It represents the number of clicks of the user on the mth product in the ith source path at time t within a period of time. It represents the total number of clicks by users in the i-th source path at time t within a period of time. It represents the deviation between the number of clicks on the mth product in the ith source path and the total number of clicks in the source path at time t within a period of time. , , k and g are both positive integers, and the time period is defined as ; Calculate the click concentration coefficient of each source path. The specific calculation formula is as follows: In the formula, is the click concentration coefficient of the i-th source path; Extract the path click feature information from the preprocessed click behavior information of each source path, including the total number of different products clicked by users in each source path at different times within a period of time, the number of product categories, and the average number of clicks on all products by users in each source path, and use the function to calculate the click feature information according to the time series. , and is represented by, t is the time point, It represents the total number of different products clicked by the user in the i-th source path at time t within a period of time. represents the number of product categories of different products clicked by the user in the i-th source path at time t within a period of time, It represents the average number of clicks on all products by users in the i-th source path at time t within a period of time; Calculate the click sparse index of each source path. The specific calculation formula is as follows: In the formula, is the click sparse index of the i-th source path; The click concentration coefficient of each source path generated and click sparsity index Construct a path contribution intensity model and generate the contribution evaluation coefficient of each source path through weighted summation , and the contribution evaluation coefficients of each source path generated The contribution evaluation coefficient threshold interval is set A comparison is made, and the contribution strength of each source path to the user's real interest signal is evaluated based on the comparison results. Based on the evaluation results, each source path is divided into a dominant path, a dispersed path, and a mixed path. The specific comparison analysis and division are as follows: like , the contribution intensity of the source path to the user's real interest signal is low, then the source path is divided into a dispersed path; like , the contribution intensity of the source path to the user's real interest signal is medium, then the source path is classified as a mixed path; like , the contribution intensity of the source path to the user's real interest signal is high, then the source path is classified as a dominant path; According to the division results of each source path, a path signal dynamic adjustment mechanism is constructed to adjust the dominant path, the decentralized path and the mixed path source paths with different sorting weights, specifically: According to the division results of dominant path, decentralized path and mixed path, different ranking weight adjustment rules are set respectively to form a dynamic adjustment mechanism of path signal; this adjustment mechanism is based on the contribution evaluation coefficient of each source path and the optimization goal of the recommendation system, and automatically determines the ranking weight adjustment method and amplitude of the products in each path through pre-set rules; Different sorting strategy adjustments are made to the dominant path, dispersed path and mixed path source paths respectively, including: for the dominant path, the weight enhancement rule in the adjustment mechanism is used to increase the sorting weight of the clicked products in the path, and give priority to displaying the products that the user is clearly interested in; for the dispersed path, the weight dispersion rule in the adjustment mechanism is used to reduce the sorting weight of the repeatedly clicked products, and at the same time increase the display weight of other unclicked products in the path to increase the diversity of recommended content; for the mixed path, the weight balance rule in the adjustment mechanism is used to maintain the sorting weight of the clicked products in the path to balance the accuracy and diversity of the recommendations; In the process of adjusting each source path by the path signal dynamic adjustment mechanism, the path dynamic feedback information of each source path is obtained in real time, and analyzed after acquisition to evaluate whether the adjustment effect of the path signal dynamic adjustment mechanism on each source path can meet expectations, and optimize the path signal dynamic adjustment mechanism according to the evaluation results; A comprehensive analysis is conducted on the application effect, adjustment records and optimization results of the path signal dynamic adjustment mechanism, and the path signal processing capability and dynamic adjustment mechanism of the recommendation ranking model are continuously optimized in combination with newly added user behavior data.

2. The commodity search optimization method based on historical behavior data according to claim 1, characterized in that: In the process of adjusting each source path by the path signal dynamic adjustment mechanism, the path dynamic feedback information of each source path is obtained in real time, and analyzed after acquisition to evaluate whether the adjustment effect of the path signal dynamic adjustment mechanism on each source path can meet expectations, and optimize the path signal dynamic adjustment mechanism according to the evaluation result, which specifically includes the following steps: In the process of adjusting each source path by the path signal dynamic adjustment mechanism, the path dynamic feedback information of each source path is obtained in real time, and preprocessed after being obtained; Extracting the ranking adjustment information and behavior deviation information from the pre-processed path dynamic feedback information of each source path, and analyzing them after extraction to generate a ranking adjustment amplitude index and a click behavior deviation index of each source path respectively; An adjustment effect evaluation model is constructed for the ranking adjustment amplitude index and click behavior deviation index generated for each source path, and an adjustment evaluation coefficient for each source path is generated. The generated adjustment evaluation coefficient for each source path is compared with the pre-set adjustment evaluation coefficient threshold for each source path. Based on the comparison results, it is evaluated whether the adjustment effect of the path signal dynamic adjustment mechanism on each source path can meet expectations, and the path signal dynamic adjustment mechanism is optimized based on the evaluation results.

3. The commodity search optimization method based on historical behavior data according to claim 2 is characterized in that: The logic for obtaining the ranking adjustment index and click behavior deviation index of each source path is as follows: Extract the sorting adjustment information from the preprocessed dynamic feedback information of each source path, including the sorting weight value of each product in each source path before dynamic adjustment, the sorting weight value after dynamic adjustment, and the rate of change of the number of clicks of each product in each source path before and after adjustment, and mark them as , and , Indicates the ranking weight value of the mth product in the ith source path before dynamic adjustment. It represents the ranking weight value of the mth product in the ith source path after dynamic adjustment. It represents the rate of change of the number of clicks of the user on the mth product in the i-th source path before and after the adjustment per unit time. , , k and g are both positive integers; Calculate the ranking adjustment index of each source path. The specific calculation formula is as follows: In the formula, is the ranking adjustment amplitude index of the i-th source path; The behavior deviation information in the pre-processed path dynamic feedback information of each source path is extracted, including the distribution ratio of the number of clicks of users on each product in each source path before adjustment, the distribution ratio after adjustment, and the change in the display weight value of each product in each source path before and after adjustment, and they are marked as , and , It represents the distribution ratio of the number of clicks of the user on the mth product in the i-th source path before adjustment. It represents the distribution ratio of the number of clicks of the user on the mth product in the i-th source path after adjustment. Indicates the change in display weight value of the mth product in the i-th source path before and after adjustment; Calculate the click behavior deviation index of each source path. The specific calculation formula is as follows: In the formula, is the click behavior deviation index of the i-th source path.

4. The commodity search optimization method based on historical behavior data according to claim 3 is characterized in that: The ranking adjustment index of each source path generated and click behavior deviation index Construct an adjustment effect evaluation model and generate the adjustment evaluation coefficients of each source path through weighted summation , and the adjusted evaluation coefficients of each source path generated Adjustment evaluation coefficient thresholds for each source path that are set in advance Compare and evaluate whether the adjustment effect of the path signal dynamic adjustment mechanism on each source path can meet expectations based on the comparison results, and optimize the path signal dynamic adjustment mechanism based on the evaluation results. The specific comparison and analysis are as follows: like , the adjustment effect of the path signal dynamic adjustment mechanism on the source path cannot meet the expectations, and the path signal dynamic adjustment mechanism needs to be optimized, including: re-evaluating the ranking weight adjustment rules of the products in the path, increasing the real-time data collection and analysis of the user's latest click behavior and display preference; dynamically adjusting the weighting parameters in the ranking adjustment amplitude index and the click behavior deviation index to optimize the balance between ranking and click distribution; re-setting the boundary conditions and priority rules for ranking weight adjustment; like , the adjustment effect of the path signal dynamic adjustment mechanism on the source path can achieve the expected effect, and there is no need to optimize the path signal dynamic adjustment mechanism.

5. A commodity search optimization system based on historical behavior data, used to implement the commodity search optimization method based on historical behavior data as described in any one of claims 1 to 4, characterized in that: It includes user click behavior detection module, path contribution evaluation module, path signal adjustment mechanism construction module, path dynamic feedback evaluation module and path signal comprehensive optimization module; The user click behavior detection module detects the user's click behavior in real time to identify whether the user has clicked on the same product multiple times through multiple recommendation source paths. If it is identified that the user has clicked on the same product multiple times through multiple recommendation source paths, the multiple source paths generated by the user during the time period of searching for the product are classified and marked to distinguish the click behavior characteristics of different source paths. The path contribution evaluation module obtains the click behavior information of each source path in real time, analyzes it after acquisition, evaluates the contribution strength of each source path to the user's real interest signal, and divides each source path into dominant path, decentralized path and mixed path according to the evaluation results; A path signal adjustment mechanism construction module is used to construct a path signal dynamic adjustment mechanism for adjusting different ranking weights of dominant path, decentralized path and mixed path source paths according to the division results of each source path; The path dynamic feedback evaluation module obtains the path dynamic feedback information of each source path in real time during the process of the path signal dynamic adjustment mechanism adjusting each source path, and analyzes it after acquisition to evaluate whether the adjustment effect of the path signal dynamic adjustment mechanism on each source path can meet expectations, and optimizes the path signal dynamic adjustment mechanism according to the evaluation results; The path signal comprehensive optimization module conducts a comprehensive analysis of the application effect, adjustment records and optimization results of the path signal dynamic adjustment mechanism, and continuously optimizes the path signal processing capability and dynamic adjustment mechanism of the recommendation ranking model based on the newly added user behavior data.

Images