Shopping guide strategy optimization method and device

Abstract

The invention discloses a shopping guide strategy optimization method and device, and relates to the technical field of big data. A specific embodiment of the method comprises the steps of determining a user behavior path based on acquired user search behavior data; wherein the user behavior path comprises a decision action and a decision state; the decision action comprises an operation behavior executed by the user; the decision state comprises a page state of the user; performing Markov modeling according to the decision action to obtain a value function, and determining the value degree of each decision state based on the value function; and optimizing the search shopping guide strategy according to the value degree of each decision state. According to the embodiment, the search shopping guide strategy can be optimized by evaluating the value of the decision state between different decision actions in combination with the user behavior path, so that the search shopping guide order forming rate and the user experience are improved.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a shopping guide strategy optimization method and device. Background Art

[0002] In e-commerce search scenarios, the behavior paths reflected by users’ search behavior data are very scattered and complex. Currently, we mainly identify abnormal situations during the user search process by describing the user conversion and churn rates at key links, and then optimize shopping guide strategies.

[0003] In the process of implementing the present invention, the inventors found that there are at least the following problems in the related art:

[0004] Analyzing user behavior only through conversions at a single stage results in a low conversion rate for search guide strategies and a poor user experience. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a user search behavior data analysis method, which can combine the user behavior path and perform value assessment on the decision status between different decision actions to optimize the search shopping guide strategy, thereby improving the search shopping guide success rate and user experience.

[0006] To achieve the above object, according to one aspect of an embodiment of the present invention, a method for analyzing user search behavior data is provided, comprising:

[0007] Determine the user behavior path based on the acquired user search behavior data; wherein the user behavior path includes a decision action and a decision state; the decision action includes the operation behavior performed by the user; the decision state includes the page state where the user is located;

[0008] Markov modeling is performed based on decision actions to obtain a value function, and the value of each decision state is determined based on the value function;

[0009] Optimize the search guide strategy according to the value of each decision state.

[0010] Optionally, Markov modeling is performed based on the decision actions to obtain a value function, including:

[0011] Define the key elements of the model; the key elements of the model include state space, action space, state transition probability and reward function;

[0012] Construct a state transition matrix based on the key elements of the model;

[0013] Determine the value function based on the decision action and state transition matrix.

[0014] Optionally, the method further comprises:

[0015] The decision states in the state transfer matrix are defined as nodes, and the state transfer probabilities exceeding a preset probability threshold in the state transfer matrix are defined as edges;

[0016] Construct a transfer path graph based on nodes and edges;

[0017] The key loss paths are extracted from the transfer path diagram to optimize the search guide strategy based on the key loss paths.

[0018] Optionally, the method further comprises:

[0019] The decision states in the state transfer matrix are removed one by one, and the state transfer matrix is ​​updated. The decrease in the user conversion probability is determined based on the updated state transfer matrix, so as to optimize the search guide strategy according to the decrease in the user conversion probability.

[0020] Optionally, before determining the user behavior path based on the acquired user search behavior data, the method further includes:

[0021] Clean the acquired user search behavior data, and build a user behavior operation log based on the cleaned user search behavior data;

[0022] Based on time similarity or category similarity, a complete session set is screened out from the user behavior operation log to determine the user behavior path according to the user search behavior data corresponding to the complete session set.

[0023] Optionally, the search shopping guide strategy includes a strategy for optimizing user search results by adjusting the display and sorting of shopping guide items; optimizing the search shopping guide strategy according to the value of each decision state includes:

[0024] The shopping guide items corresponding to the decision states whose values ​​exceed the preset threshold are placed in front, and the shopping guide items corresponding to the decision states whose values ​​do not exceed the preset threshold are removed.

[0025] According to a second aspect of an embodiment of the present invention, a shopping guide strategy optimization device is provided, characterized in that it includes:

[0026] A path determination module is used to determine the user behavior path based on the acquired user search behavior data; wherein the user behavior path includes a decision state and a decision action; the decision state includes the page state of the user, and the decision action includes the operation behavior performed by the user;

[0027] The value determination module is used to perform Markov modeling on the decision action to obtain the value function, and determine the value of each decision state based on the value function;

[0028] The strategy optimization module is used to optimize the search guide strategy according to the value of each decision state.

[0029] According to a third aspect of an embodiment of the present invention, there is provided an electronic device, including:

[0030] one or more processors;

[0031] a memory for storing one or more programs,

[0032] When one or more programs are executed by one or more processors, the one or more processors are enabled to implement the method of any of the above embodiments.

[0033] According to a fourth aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method of any of the above embodiments is implemented.

[0034] According to a fifth aspect of an embodiment of the present invention, there is provided a computer program product, including a computer program, which implements the method of any of the above embodiments when the computer program is executed by a processor.

[0035] One embodiment of the above invention has the following advantages or beneficial effects: determining the user behavior path based on the acquired user search behavior data; wherein the user behavior path includes decision actions and decision states; decision actions include operation behaviors performed by the user; decision states include the page state in which the user is located; Markov modeling is performed based on the decision actions to obtain a value function, and the value of each decision state is determined based on the value function; the search shopping guide strategy is optimized based on the value of each decision state; thereby, the user behavior path can be combined to perform value evaluation on the decision states between different decision actions to optimize the search shopping guide strategy, thereby improving the search shopping guide success rate and user experience.

[0036] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0038] Figure 1 is a schematic diagram of the main process of the shopping guide strategy optimization method according to an embodiment of the present invention;

[0039] Figure 2 is a flow chart of a shopping guide strategy optimization method according to a preferred embodiment of the present invention;

[0040] Figure 3 is a flow chart of a data preparation phase according to a preferred embodiment of the present invention;

[0041] Figure 4is a schematic diagram of main modules of a shopping guide strategy optimization device according to an embodiment of the present invention;

[0042] Figure 5 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0043] Figure 6 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0045] It should be noted that the acquisition, storage and application of personal information etc. involved in the embodiments of the present invention are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0046] In e-commerce search scenarios, the behavior paths reflected by user search behavior data are very scattered and complex. Currently, the main focus is on determining abnormal situations by describing the user conversion and churn rates of key links during the user search process, thereby optimizing the entire search behavior process. User behavior analysis based only on the conversion of a single link results in a low search guide order conversion rate and a poor user experience.

[0047] In view of this, according to one aspect of an embodiment of the present invention, a method for analyzing user search behavior data is provided.

[0048] Figure 1 FIG. 1 is a schematic diagram of the main process of the method for analyzing user search behavior data according to an embodiment of the present invention. Figure 1 As shown, the user search behavior data analysis method according to an embodiment of the present invention includes the following steps S101 to S103.

[0049] Step S101, determining a user behavior path based on the acquired user search behavior data; wherein the user behavior path includes a decision action and a decision state; the decision action includes the operation behavior performed by the user; and the decision state includes the page state where the user is located.

[0050] User search behavior data refers to the relevant operation records generated when users search on e-commerce platforms or search engines. Specifically, these data include the search keywords entered by users, the products or pages clicked, the filtering operations performed, the sorting behaviors, the operations of adding to the shopping cart, and the final purchase behaviors, etc. The user behavior path refers to the complete behavior link of the user from the initial search operation to the final decision-making behavior (such as purchase or abandonment) during the search and browsing process. This path can be described by a series of decision actions and decision states. Among them, the decision action includes the operation behavior performed by the user, which can be the choice or operation made by the user at each node or in the state, such as clicking on a certain product or applying a certain filtering condition. The decision state includes the page state of the user, which can be the state of the user after each operation, such as "browsing products", "filtered products" or "added to the shopping cart", etc. The user behavior path can present the complete decision-making process of the user in the entire shopping or search process, and show the path of the user from search to the final action.

[0051] Based on the acquired user search behavior data, the user behavior path can be determined. Each user operation can be arranged in chronological order, the relationship between these operations can be analyzed, the user's behavior pattern can be identified, and a coherent behavior sequence can be formed. For example, starting from the user's search keyword, it can be tracked whether the user narrows the search range through the filtering function, whether the user clicks on a certain product, whether the user views other recommended products, and finally whether the user purchases or abandons the operation. The sequence and correlation of these steps form the user's behavior path.

[0052] Step S102, perform Markov modeling according to the decision action to obtain a value function, and determine the value of each decision state based on the value function.

[0053] The value of a decision state is a measure of the contribution of each user to the overall business goal (such as platform transaction volume or user conversion rate) under different decision states. In an e-commerce platform or recommendation system, each decision state has its influence on the final conversion (purchase behavior), and the value reflects the importance of the decision state in the user behavior path. A high-value decision state means that the state is more critical in user conversion or business indicator improvement.

[0054] Specifically, the modeling process can be achieved by constructing a state transition matrix and calculating the value function. First, the state space (such as search result page, filter page, product details page, etc.) and action space (such as click, filter, sort, etc.) are defined based on user behavior data, and then the probability of users transferring to other decision states after performing certain decision actions in different decision states is counted to form a state transition matrix. At the same time, the reward value is assigned to the state-action pair according to the immediate reward function (such as the increased conversion rate after clicking on the product or the user stay time brought by the filter conditions). The value function of each decision state is iteratively calculated using the Bellman equation to reflect the cumulative long-term benefits under the decision state, thereby determining the value of each decision state. The value function can also be dynamically optimized based on the reinforcement learning model, and the Markov decision process can be modeled based on the decision action using the reinforcement learning algorithm (such as deep Q learning). The value estimation of the decision-action pair of the decision state is updated by collecting the user behavior path in real time. According to the current user behavior and reward, the optimal value function is gradually approached, and it can dynamically adapt to changes in user behavior. Compared with static modeling, the reinforcement learning method can more accurately determine the value of each decision state in scenarios where user behavior data is insufficient or the behavior pattern changes rapidly.

[0055] Through these methods, it is possible to systematically analyze user behavior in different decision-making states and provide valuable insights for the platform's business decisions, such as optimizing high-value states or adjusting low-value states to improve overall user conversion rates.

[0056] Step S103, optimizing the search and shopping guide strategy according to the value of each decision state.

[0057] Search guide strategies include providing personalized product recommendations and search result display strategies for users. By analyzing the user's search behavior and preferences, relevant products, services or product inventory are presented to users to improve the user's shopping experience and conversion rate. For example, by optimizing the shopping guide items in the user's search results, the user's interest in the product is increased and the purchase behavior is promoted; the shopping guide items include category tiling, filtering options, search drop-down, etc. Among them, category tiling is a way to intuitively display the classification of products. By tiling the products in a grid or list according to different categories, users can quickly browse and select the product category of interest. The filtering option is a tool that provides users with fine-grained search. It can filter out irrelevant products by setting conditions, so that users can more accurately locate products that meet their needs. Specifically, the filtering options include price range, brand, color, specification, rating and other dimensions. Users can dynamically adjust the conditions and update the search results in real time; for example, on the e-commerce platform, users search for "laptops" and can choose to filter "price below 5,000 yuan", "screen size: 14 inches", etc. The search drop-down is a way to recommend related words, products or phrases based on historical data or algorithms when users enter keywords, helping users complete input and find target content faster. It can not only improve input efficiency, but also stimulate users' potential interests by providing diverse recommended information, thereby guiding users to explore more products; for example, when users enter "mobile phone" in the search box, the drop-down will display "mobile phone case", "Double Eleven discount for mobile phones", etc. The reasonable layout of these shopping guide items can help users find products that meet their needs faster, thereby improving shopping efficiency and satisfaction.

[0058] In addition to adjusting shopping guide items, search shopping guide strategies can also optimize user experience and improve conversion rates by adjusting product inventory. Based on the value of each decision state, inventory can be managed more accurately, and sufficient inventory can be maintained for products related to high-value decision states, so that user needs can be met in a timely manner during peak user demand periods to avoid loss caused by out-of-stock. At the same time, for products with lower value or less searched by users, inventory investment can be appropriately reduced to reduce inventory backlog risks and optimize supply chain management. Specifically, by analyzing user operations in different decision states, such as product clicks, viewing detail pages, and adding to shopping carts, the platform can evaluate which products are more concerned by users. The cumulative value of these behaviors can be used to predict future product demand and prioritize resource allocation in inventory strategies. For example, when a product has a high click-through rate and add-to-cart rate and is related to a high-value decision state, the platform can predict that the product will have a higher sales conversion in the future and replenish inventory in advance to ensure that the product is available during the user decision process. Search guide strategy optimization based on decision status value can adjust the shopping guide items, and can also improve user experience and platform conversion rate through precise inventory management and other methods, more flexibly adjust search guide strategies, improve user service experience, and effectively increase the success rate of product sales.

[0059] The embodiment of the present invention determines the user behavior path based on the acquired user search behavior data; wherein the user behavior path includes decision actions and decision states; performs correlation analysis on the decision actions, and determines the value of each decision state based on the correlation analysis results; optimizes the search shopping guide strategy according to the value of each decision state; thereby, combining the user behavior path, performing value assessment on the decision states between different decision actions to optimize the search shopping guide strategy, thereby improving the search shopping guide success rate and user experience.

[0060] Optionally, Markov modeling is performed according to the decision action to obtain the value function, including: defining the key elements of the model; wherein the key elements of the model include state space, action space, state transition probability and reward function; constructing a state transition matrix based on the key elements of the model; and determining the value function according to the decision action and the state transition matrix.

[0061] Key elements include state space, action space, state transition probability, and reward function. The state space represents the different decision states of users on the platform, and each decision state corresponds to the specific situation of the user at a certain moment. The action space describes the actions or behavioral decisions that users may take in different decision states, such as searching for new keywords, clicking on a certain product, sorting a list of products, or placing an order directly. The state transition probability is used to describe the possibility of a user moving from one decision state to another. The reward function defines the benefits of user behavior, such as the amount of an order placed by a user after successfully completing a purchase. The platform can optimize the user's decision path based on the reward function to maximize the overall benefits of the platform.

[0062] After defining the above key elements, a state transition matrix can be constructed based on it. The state transition matrix is ​​a probability table that describes the transition of user behavior from one decision state to another. The state transition matrix can predict the next decision state that the user may enter under different decision actions. The construction of the state transition matrix depends on the user's behavior data, which includes a series of operation steps from search to click to purchase. By analyzing these behavior trajectories, a corresponding probability value can be assigned to each decision state transition, so that the matrix can accurately reflect the dynamic changes of user behavior. After the state transition matrix is ​​constructed, the value function can be determined according to the decision action and the state transition matrix. The value function can be used to evaluate the long-term value of each decision state, reflecting the maximum cumulative reward that can be obtained by taking the best decision in this decision state. The value function can be calculated by the Bellman equation, which calculates the value of each decision state by recursion. Based on the value function, the value of each decision state is further determined. The high-value decision state represents the key node in the user behavior path, which can affect user conversion and search efficiency. According to these values, the search guide strategy can be optimized, such as enhancing user interaction on high-value decision states, highlighting related filter items or recommended tags, to improve user experience and conversion rate.

[0063] The embodiments of the present invention can better understand user behavior patterns, optimize search and recommendation strategies, and thus improve conversion rates and user satisfaction.

[0064] Optionally, the method also includes: defining the decision states in the state transfer matrix as nodes, and defining the state transfer probabilities in the state transfer matrix that exceed a preset probability threshold as edges; constructing a transfer path graph based on the nodes and edges; extracting key loss paths from the transfer path graph to optimize the search guide strategy based on the key loss paths.

[0065] In the process of constructing the transition path graph, each decision state in the state transition matrix can be defined as a node, and each node represents a specific stage of the user's operation process. Between these nodes, according to the state transition probability in the state transition matrix, the edges connecting the nodes are constructed. These edges reflect the probability of the user transitioning from one decision state to another. In order to optimize the analysis, the state transition exceeding the preset probability threshold can be defined as an edge, focusing on the most likely user behavior path, thereby reducing data complexity. Based on these nodes and edges, a user behavior transition path graph can be constructed. The transition path graph is a visual expression of the entire behavior path of the user from the initial search behavior to the final purchase decision, showing the user's transition process between different nodes and its corresponding probability. By analyzing this transition path graph, the most common behavior flow experienced by users can be intuitively seen, and the possible transfer paths of users at different stages can be identified. The key churn path is extracted from the transition path graph. The key churn path is the key decision node and path where users may leave the platform and terminate the operation during the shopping process, such as users clicking on multiple products but not adding them to the shopping cart, or users adding products to the shopping cart but not completing the order. By identifying these key churn paths, the platform can analyze the specific reasons why users give up purchasing. You can optimize the search guide strategy based on these key churn paths, such as adjusting the display order of shopping guide items, the display method of recommended content, or optimizing the user's interactive experience at a certain node to reduce the probability of churn. For example, if you find that users often churn on a certain product details page, you can optimize the content layout of the page, add promotional information, or recommend alternative products. This optimization strategy based on key churn paths can effectively improve the user's purchase conversion rate and increase the overall revenue of the platform.

[0066] Optionally, the method also includes: removing decision states in the state transfer matrix one by one, and updating the state transfer matrix, determining the decrease in the user conversion probability based on the updated state transfer matrix, so as to optimize the search guide strategy according to the decrease in the user conversion probability.

[0067] In the process of optimizing the search guide strategy, the decision states in the state transfer matrix can be removed one by one to evaluate the impact of each decision state on the user conversion rate. First, for the existing state transfer matrix, the decision states are removed one by one step by step. After each decision state is removed, the state transfer matrix is ​​updated accordingly. Based on the updated state transfer matrix after removal, the user conversion probability in the entire user behavior path can be recalculated. This conversion probability represents the overall possibility of the user completing a series of behaviors from search to purchase. By comparing the changes in the user conversion probability before and after removal, the contribution of each decision state to the conversion rate can be evaluated. If the user conversion probability decreases significantly after a decision state is removed, it means that the decision state is crucial to the user conversion. On the contrary, if the decrease is small, the state has a weak impact on the user behavior path. By analyzing these declines, we can determine which decision states play a key role in the final user conversion. Based on these analysis results, we can optimize the search guide strategy in a targeted manner. For example, we can highlight or pre-process the decision states that contribute more to the user conversion rate, so that users can more easily access the shopping guide items in these states. At the same time, for decision states with less impact, we can consider reducing their recommendation frequency or adjusting them to improve the overall efficiency of the shopping guide strategy. By gradually removing states and observing changes in user conversion rates, we can accurately optimize the shopping guide process, thereby improving user conversion rates and overall revenue.

[0068] Optionally, before determining the user behavior path based on the acquired user search behavior data, it also includes: performing data cleaning on the acquired user search behavior data, and constructing a user behavior operation log based on the cleaned user search behavior data; screening out a complete session set from the user behavior operation log based on time similarity or category similarity, so as to determine the user behavior path based on the user search behavior data corresponding to the complete session set.

[0069] The data cleaning process includes removing invalid, duplicate or abnormal data, especially data affected by robot operation, crawler interference or order-brushing behavior, so that the retained data is more authentic and effective. Based on the cleaned user search behavior data, a user behavior operation log is constructed. The user behavior operation log is a data record organized by individual users, which contains each user's browsing, clicking, purchasing and other operation behaviors. Each operation is accompanied by a timestamp and a behavior category tag, such as search keywords, filter clicks, product clicks, and order placement. Based on time similarity or category similarity, a complete session set can be filtered out. Time similarity means that continuous operations are close in time, indicating that these operations belong to the same session; while category similarity indicates the correlation between product categories between operations. If users concentrate on operating products of the same or similar categories in a short period of time, these operations are classified as the same session. Through these two similarity standards, the user's continuous operation sequence on the platform can be accurately identified to form a complete session set. Based on the complete session set obtained through screening, the user behavior path is determined using the user search behavior data corresponding to these sessions. The complete session set represents all the user's operation behaviors in a certain demand scenario and reflects the user's decision-making mode in this scenario. Determining the user behavior path through these data can improve the accuracy of user behavior analysis.

[0070] Optionally, the search shopping guide strategy includes a strategy for optimizing user search results by adjusting the display and sorting of shopping guide items; optimizing the search shopping guide strategy according to the value of each decision state, including: putting in front shopping guide items corresponding to decision states whose values ​​exceed a preset threshold, and removing shopping guide items corresponding to decision states whose values ​​do not exceed the preset threshold.

[0071] Pre-process the shopping guide items corresponding to high-value decision states, that is, when users search, give priority to displaying content related to high-value decision states. For example, if the filtering function contributes greatly to the user's conversion rate, the filtering options can be designed to be more prominent and convenient, or even pre-select some filtering conditions by default according to user habits. Correspondingly, for decision states with lower value, you can choose to simplify or remove the corresponding shopping guide items, thereby reducing user interference and unnecessary choices, and helping users find the products they need faster. In addition, you can also use machine learning methods such as reinforcement learning to dynamically adjust shopping guide strategies. By continuously learning users' real-time operational behavior data, the value of the decision state is updated in real time, and the sorting and presentation of shopping guide items are automatically adjusted according to changes, effectively improving user experience and the platform's revenue conversion rate.

[0072] Figure 2 is a flow chart of a method for analyzing user search behavior data according to a preferred embodiment of the present invention; Figure 2As shown in the figure, the search guide strategy is optimized based on user behavior data. First, the data set is prepared by recording the user session granularity, collecting the specific search behavior and interactive operations of users on the e-commerce platform to form a complete set of behavior data sets. The model is constructed using the Markov decision process. First, the core elements in the model are defined, including decision state, decision action, discount factor (weighing the value of current and future decisions) and reward function (evaluating the return of users in each operation). These elements together constitute a mathematical model that can simulate the user decision process. After the model is built, the state transfer function is generated. The state transfer function describes the probability of a user transferring from one decision state to another. In the initial model generation stage, this state transfer function is still in its initial form and needs to be further optimized and improved. In the strategy evaluation stage, the state transfer function generated previously is used to calculate the value function of each decision state and decision action, so as to evaluate the effectiveness of different strategies, that is, to find the strategy that can bring the greatest benefit and ensure that the shopping guide tool can maximize the user's conversion rate. Identify the key churn path based on the state transfer matrix. In the process of users transferring from one decision state to another, if there is a large amount of user loss on certain paths, special attention should be paid to these key loss paths. By analyzing these key loss paths, key nodes can be found and targeted optimization measures can be taken, such as increasing the supply of certain products or optimizing related search terms to reduce user loss. At the same time, the state transition matrix can be used to query those state transitions that can bring high returns, and the N queries with the highest returns can be selected. These queries are regarded as the next steps that are most helpful to users. By recommending these queries to users, users can be effectively guided to complete purchase conversions. The top N queries are used as the recommended guide items for the shopping guide tool of the current node, so as to help users find the products of interest more quickly, and finally optimize and improve the platform shopping guide strategy.

[0073] Figure 3 is a flow chart of the data preparation stage of the preferred embodiment of the present invention; Figure 3 As shown in Figure 1, a data set is obtained from an e-commerce platform. The data set mainly consists of users' browsing and clicking logs and purchase records. Meaningless browsing and clicking data, such as cheating behaviors of robots and crawlers, are excluded, and fake order data is excluded. The records are combined into user behavior operation logs o=<uuid, query,event_id,time> , where uuid represents the user, query represents the search keyword, event_id represents the user's various click and purchase behavior events (such as filter clicks, sort clicks, product clicks, purchase orders, etc.), and time represents the time when the user behavior event is triggered; then identify the user session and organize the log data, let session S = {o1, o2, ..., o i}, all operations oi The execution user is the same, and then the order of the logs is determined based on time. If the highly relevant categories of the query are consistent, then o is determined. i +1 belongs to session S, otherwise o i +1 starts a new session. The business meaning of each session is the operation record of the same user for the same type of demand, organized in time series.

[0074] Specifically, we collect and organize users’ search logs from e-commerce platforms. These logs contain users’ browsing, clicking, and purchasing records. During the collection process, we filter the log data to exclude meaningless operations, such as the behavior of robots and crawlers, and abnormal order data, such as order-brushing behavior. We group the operation logs according to the user’s unique identifier (UUID) to ensure that each group of log data belongs to the same user. After grouping, we sort the operation logs of each user in chronological order so that we can better analyze the time series of user behavior. After sorting, we start to analyze the user’s operation behaviors one by one. First, we determine whether the current operation is the first operation of the user. If so, we create a new session. If it is not the first operation, we need to check whether the operation is highly correlated with the search keywords of the previous operation. If so, we consider that the two belong to the same session. Otherwise, we create a new session. In the same session, we further determine whether the event type of the user operation is a purchase event. If it is a purchase event, we consider that the session is over, and add the complete session to the final dataset. Otherwise, we continue to record the behavior in the session until the session end condition is met or a new purchase occurs. In this way, the system can effectively organize the user's operation records for the same type of requirements into a session, providing clear structured data for subsequent analysis.

[0075] In a preferred embodiment of the present invention, in the process of building a model using the Markov decision process, the search session process between the user and the search engine is as follows: the user enters "men's shoes" as a query in the search box and clicks the "search" button; the search engine executes a search request and displays the product page with the keyword "men's shoes"; the user browses the product display page and can click on some of the products to enter the detail page; if the user finds that there are no satisfactory products, he can filter and rearrange the page through product modules such as screening and sorting, or refine the needs through product modules such as category tiling and middle interlacing, or re-enter new search keywords by actively rewriting; the search engine receives the request, executes a second search request, and displays the product page corresponding to the second search; after several such "search-sort / re-search-browse" rounds, the search session will eventually end when the user purchases a product or gives up browsing.

[0076] First, we model the context information and user behavior in the search session, and define concepts such as search result page, user behavior event, and transaction conversion rate, which are the basis for defining states and state transition relationships. Define query access history, let q be the query of a search session, for the initial time step t=0, the corresponding initial query access path is h0=q, for any other time step t>=1, the corresponding access path is ht=(ht-1, qt), where ht-1 is the access path of the t-1th step, and qt is the access query of the tth step; different users will choose to re-search, purchase, or leave at different time steps; and purchase and leaving are two terminal states, so the transaction conversion probability and abandonment probability can be defined based on the terminal state. Define the transaction conversion probability, for any path node in a search session, ht (t>0), let B(ht) represent the random event of the user's purchase behavior after observing ht, that is, the transaction conversion probability of ht, denoted as b(ht), which is the probability of event B(ht) occurring under ht. Define the abandonment probability. For any path node ht (t>0) in a search session, let L(ht) represent the random event that the user leaves the search session after observing ht. Then the abandonment probability of ht, denoted by l(ht), is the probability of event L(ht) occurring under ht. Define the re-search probability. For any path node ht (t>0) in a search session, let C(ht) represent the random event that the user initiates a secondary search after observing ht. Then the re-search probability of ht, denoted by c(ht), is the probability of event C(ht) occurring under ht. Define the decision process of the search session. The search session decision process can be formally defined as M=<T,H,S,A,R,P> , in which T is the maximum number of decision steps for the search; is the query access history of all covered queries in the search session, where Ht is the set of all possible queries at time t ( ); is the state space, is a set of non-terminal states containing all continuing sessions, and They are the terminal state sets that include all transaction conversion events and leaving events; is the action interval, including all product modules that trigger re-search (in a broad sense), including but not limited to: drop-down modules (drop-down words, historical searches, hidden patterns, search discoveries, hot search lists), active searches, sorting categories (sales ranking, price ranking, comment ranking), screening categories (JD Logistics, Paipai second-hand, same style in shopping malls, new products, etc.), re-search categories (category tiling, middle labels, related searches, business card tags), etc. After users click on a product module, a re-search is triggered. Some product modules do not modify the original search terms. In order to distinguish different search pages, this type of search is to manually modify the search terms to the original query + product module name (such as: men's shoes + price ranking) in data preprocessing; R is the reward function. The definition of the reward function will facilitate as many transactions between users and sellers as possible. For any time step ( ), when observing that the user is on a certain search page ht+1, the user will purchase the product with an average probability of b(ht+1). Although different users will choose different products to purchase, from a statistical point of view, the transaction price of the product corresponding to the conversion time on ht+1 will inevitably obey a specific distribution. m(ht+1) represents the transaction expectation on the search page ht+1. At the same time, the average correlation coefficient between the search result page product and the search term. If the correlation coefficient is less than 0.5, it is considered that there is a lack of relevant product supply under the search term, which affects the user experience to some extent. It is defined that a negative reward is given at this time, which is the average price of the exposed products on the current search page; then the user performs action a on state C(ht) to transfer to any state The reward definition is as follows:

[0077]

[0078] in, for The terminal state corresponding to the transaction conversion event that occurred on the query result page. From the reward function definition, we can see that a positive reward can only be obtained when a transaction occurs, and a negative reward can be obtained when the supply is insufficient. In other cases, the reward value obtained is zero. The expected transaction price of any search query result page is usually unknown. In specific applications, the actual transaction price of each transaction can be used as the reward signal for the corresponding transaction event.

[0079] Specifically, in the algorithm design process, let M=<H,S,A,R,P> For a search session decision process, Represents a collection of policies, whose elements , construct a calculation method for the cumulative reward value that can be obtained under any state s, which can be expressed by the following state value function formula:

[0080]

[0081]

[0082] At the same time, define the action value function Indicates that the MDP follows the strategy The expected return of executing action a for the current state s is:

[0083]

[0084]

[0085] Where V(s) is the value function of state s, is the discount factor, and the value function is calculated according to the historical strategy until convergence, giving the transformed state transfer matrix. At the same time, for the state s, the action value function can be calculated separately The action value of state s is calculated, and the expected return of all subsequent states of state s is output. The best recommendation item is generated according to the expectation, which can be used as an alternative for shopping guide tools (category tiling, middle label), as a page for pre-fetching or recommending to users, further shortening the user's purchase path and improving the platform traffic conversion rate. When identifying the key churn path, according to the user behavior state transition matrix, the state is used as the node of the transition path graph, and the elements with transition probability greater than 0 in the matrix are used as the edges in the transition path graph model; based on the obtained transition path graph, the behavior state and path required for each row state to the churn state are extracted to obtain the key churn path of each behavior state, which can effectively identify the key path that causes user churn. At the same time, the state value and removal effect of each state can be combined. The removal effect refers to the change in the sum of probabilities on all paths from the starting state to the conversion state after removing a certain state. The removal effect actually reflects the degree of decline in the overall conversion rate of the system after removing the channel. In turn, business operations can optimize product supply, shopping guide tools, and collaborative search algorithms to optimize search results in a targeted manner, thereby improving traffic conversion efficiency and user experience.

[0086] According to a second aspect of an embodiment of the present invention, a shopping guide strategy optimization device is provided.

[0087] Figure 4 Schematic diagram of the main modules of the shopping guide strategy optimization device according to an embodiment of the present invention. Figure 4 As shown, a shopping guide strategy optimization device 400 includes:

[0088] The path determination module 401 is used to determine the user behavior path based on the acquired user search behavior data; wherein the user behavior path includes a decision state and a decision action; the decision state includes the page state of the user, and the decision action includes the operation behavior performed by the user;

[0089] A value determination module 402 is used to perform Markov modeling on the decision action to obtain a value function, and determine the value of each decision state based on the value function;

[0090] The strategy optimization module 403 is used to optimize the search and shopping guide strategy according to the value of each decision state.

[0091] Optionally, the value determination module 402 is further configured to:

[0092] Define the key elements of the model; the key elements of the model include state space, action space, state transition probability and reward function;

[0093] Construct a state transition matrix based on the key elements of the model;

[0094] Determine the value function based on the decision action and state transition matrix.

[0095] Optionally, the device 400 further includes a loss path module, and the loss path module is used to:

[0096] The decision states in the state transfer matrix are defined as nodes, and the state transfer probabilities exceeding a preset probability threshold in the state transfer matrix are defined as edges;

[0097] Construct a transfer path graph based on nodes and edges;

[0098] The key loss paths are extracted from the transfer path diagram to optimize the search guide strategy based on the key loss paths.

[0099] Optionally, the apparatus 400 further comprises a removal effect module, the removal effect module being configured to:

[0100] The decision states in the state transfer matrix are removed one by one, and the state transfer matrix is ​​updated. The decrease in the user conversion probability is determined based on the updated state transfer matrix, so as to optimize the search guide strategy according to the decrease in the user conversion probability.

[0101] Optionally, the device 400 further includes a data cleaning module, and the data cleaning module is used to:

[0102] Clean the acquired user search behavior data, and build a user behavior operation log based on the cleaned user search behavior data;

[0103] Based on time similarity or category similarity, a complete session set is screened out from the user behavior operation log to determine the user behavior path according to the user search behavior data corresponding to the complete session set.

[0104] Optionally, the search guide strategy includes a strategy for optimizing user search results by adjusting the display and sorting of guide items; the strategy optimization module 403 is further used for:

[0105] The shopping guide items corresponding to the decision states whose values ​​exceed the preset threshold are placed in front, and the shopping guide items corresponding to the decision states whose values ​​do not exceed the preset threshold are removed.

[0106] It should be noted that the specific implementation content of the shopping guide strategy optimization device of the present invention has been described in detail in the above shopping guide strategy optimization method, so the repeated content will not be described again here.

[0107] According to a third aspect of an embodiment of the present invention, there is provided an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method provided by the first aspect of the embodiment of the present invention.

[0108] According to a fourth aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method provided by the first aspect of the embodiment of the present invention is implemented.

[0109] According to a fifth aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the method provided by the first aspect of the embodiment of the present invention is implemented.

[0110] Figure 5 An exemplary system architecture 500 is shown to which the shopping guide strategy optimization method or shopping guide strategy optimization device according to the embodiment of the present application can be applied.

[0111] like Figure 5 As shown, system architecture 500 may include terminal devices 501, 502, 503, a network 504 and a server 505. Network 504 is used to provide a medium for communication links between terminal devices 501, 502, 503 and server 505. Network 504 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0112] Users can use terminal devices 501, 502, 503 to interact with server 505 through network 504 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 501, 502, 503, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0113] The terminal devices 501 , 502 , and 503 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0114] The server 505 may be a server that provides various services, such as a backend management server (only an example) that provides support for shopping websites browsed by users using the terminal devices 501, 502, and 503. The backend management server may analyze and process the received shopping guide strategy optimization request and other data, and feed back the processing results (such as shopping guide strategy optimization results - only an example) to the terminal device.

[0115] It should be noted that the shopping guide strategy optimization method provided in the embodiment of the present application is generally executed by the server 505 , and accordingly, the shopping guide strategy optimization device is generally disposed in the server 505 .

[0116] It should be understood that Figure 5 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0117] Reference below Figure 6 , which shows a schematic diagram of the structure of a computer system 600 of a terminal device suitable for implementing an embodiment of the present application. Figure 6 The terminal device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0118] like Figure 6 As shown, the computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the computer system 600 are also stored. The CPU 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0119] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage section 608 as needed.

[0120] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit (CPU) 601, the above-mentioned functions defined in the system of the present application are executed.

[0121] It should be noted that the computer-readable medium shown in the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. Computer-readable storage media may include, for example, but are not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0122] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the above-mentioned module, program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0123] The units involved in the embodiments described in the present application may be implemented by software or by hardware. The units described may also be set in a processor, for example, it may be described as: a processor includes a path determination module, a value determination module and a strategy optimization module. Among them, the names of these modules do not constitute a limitation on the modules themselves under certain circumstances. For example, the path determination module may be described as "a module for determining a user behavior path based on the acquired user search behavior data".

[0124] As another aspect, the present application also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: determining a user behavior path based on the acquired user search behavior data; wherein the user behavior path includes a decision action and a decision state; the decision action includes the operation behavior performed by the user; the decision state includes the page state of the user; Markov modeling is performed according to the decision action to obtain a value function, and the value of each decision state is determined based on the value function; and the search guide strategy is optimized according to the value of each decision state.

[0125] The computer program product provided by the embodiment of the present invention includes a computer program. When the computer program is executed by a processor, the shopping guide strategy optimization method in the embodiment of the present invention is implemented.

[0126] The technical solution according to the embodiment of the present invention has the following advantages or beneficial effects: determining the user behavior path based on the acquired user search behavior data; wherein the user behavior path includes decision actions and decision states; decision actions include operation behaviors performed by users; decision states include page states in which users are located; Markov modeling is performed according to decision actions to obtain a value function, and the value of each decision state is determined based on the value function; the search shopping guide strategy is optimized according to the value of each decision state; thereby, the user behavior path can be combined to perform value evaluation on the decision states between different decision actions to optimize the search shopping guide strategy, thereby improving the search shopping guide order success rate and user experience.

[0127] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

[0128] It should be noted that the acquisition, storage and application of user personal information involved in the technical solution of the present disclosure are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

Claims

1. A shopping guide strategy optimization method, characterized in that: include: Determine a user behavior path based on the acquired user search behavior data; wherein the user behavior path includes a decision action and a decision state; the decision action includes an operation performed by the user; and the decision state includes a page state where the user is located; Performing Markov modeling according to the decision action to obtain a value function, and determining the value of each decision state based on the value function; The search and shopping guide strategy is optimized according to the value of each decision state.

2. The method according to claim 1, characterized in that Markov modeling is performed according to the decision action to obtain a value function, including: Defining key elements of the model; wherein the key elements of the model include state space, action space, state transition probability and reward function; Constructing a state transition matrix based on the key elements of the model; A value function is determined based on the decision action and the state transition matrix.

3. The method according to claim 2, characterized in that The method further comprises: The decision states in the state transfer matrix are defined as nodes, and the state transfer probabilities exceeding a preset probability threshold in the state transfer matrix are defined as edges; Constructing a transfer path graph based on the nodes and the edges; A key loss path is extracted from the transfer path diagram to optimize the search and shopping guide strategy based on the key loss path.

4. The method according to claim 2, characterized in that: The method further comprises: The decision states in the state transfer matrix are removed one by one, and the state transfer matrix is ​​updated. The decrease in the user conversion probability is determined based on the updated state transfer matrix, so as to optimize the search guide strategy according to the decrease in the user conversion probability.

5. The method according to claim 1, characterized in that Before determining the user behavior path based on the acquired user search behavior data, it also includes: Clean the acquired user search behavior data, and build a user behavior operation log based on the cleaned user search behavior data; A complete session set is screened out from the user behavior operation log based on time similarity or category similarity, so as to determine the user behavior path according to the user search behavior data corresponding to the complete session set.

6. The method according to claim 1, characterized in that The search and shopping guide strategy includes a strategy for optimizing user search results by adjusting the display and sorting of shopping guide items; The search guide strategy is optimized according to the value of each decision state, including: The shopping guide items corresponding to the decision states whose values ​​exceed the preset threshold are placed in front, and the shopping guide items corresponding to the decision states whose values ​​do not exceed the preset threshold are removed.

7. A shopping guide strategy optimization device, characterized in that: include: A path determination module, used to determine a user behavior path based on the acquired user search behavior data; wherein the user behavior path includes a decision state and a decision action; the decision state includes the page state of the user, and the decision action includes the operation behavior performed by the user; A value determination module, used for performing Markov modeling on the decision action to obtain a value function, and determining the value of each decision state based on the value function; The strategy optimization module is used to optimize the search and shopping guide strategy according to the value of each decision state.

8. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

9. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images