Intelligent optimization method and system for store operation

Abstract

The invention discloses a store operation intelligent optimization method and system. The method comprises the steps of obtaining behavior data of users in a store; according to the behavior data, the cognitive level of the user is analyzed through a preset dynamic user hierarchical model, and a plurality of user hierarchies are obtained; based on the behavior data, interest of each user level is analyzed through a preset user interest map construction model, and an interest map corresponding to each user level is constructed; according to the interest map, generating a model through a preset operation scheme, and generating a corresponding initial operation scheme; evaluating the effect of the initial operation scheme through a preset scheme effect evaluation model to obtain a scheme effect; according to the scheme effect, optimizing the initial operation scheme through a preset operation scheme optimization model to obtain an optimized operation scheme; according to the method, the operation scheme is dynamically optimized, so that market changes and user requirements can be met in time, and the competitiveness of stores is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of store operation, and in particular to a store operation intelligent optimization method and system. Background Art

[0002] Current educational product stores face the challenges of a wide variety of products, large age differences and demand differences among service objects. How to make corresponding product recommendations and activities for children of different age groups is crucial to improving store service quality and economic benefits. However, the existing store operation plan formulation method lacks in-depth mining and precise analysis of user behavior data, making it difficult to accurately grasp the user's cognitive level and interests, resulting in a lack of targeted operation plans; relying on experience judgment when formulating plans, lacking systematicity and scientificity, and unable to adapt to market changes and dynamic changes in user needs in a timely manner; using a single standard to classify users, it cannot be dynamically adjusted according to the actual situation of users, and cannot meet the ever-changing needs of users.

[0003] To solve at least one of the above problems, the present application proposes a method and system for intelligent optimization of store operations. Summary of the invention

[0004] In view of the shortcomings of the prior art, the main purpose of the present invention is to provide a method and system for intelligent optimization of store operations, which can effectively solve the problems in the background technology. The specific technical solutions of the present invention are as follows: A method for intelligent optimization of store operations, comprising: Obtain behavioral data of users in the store; According to the behavior data, the user's cognitive level is analyzed through a preset dynamic user stratification model to obtain multiple user strata; Analyze the user's interest tags based on the behavior data, analyze the interest frequency of each user level through a preset user interest graph construction model, and construct an interest graph corresponding to each user level, wherein the user level in the interest graph is a root node, the interest tags are child nodes, and the corresponding root nodes and child nodes are connected through the interest frequency; Filtering core interests from the interest graph, mapping them in a preset activity template library through a preset operation plan generation model, and generating a corresponding initial operation plan; Evaluate the effect of the initial operation plan through a preset plan effect evaluation model to obtain the plan effect; According to the effect of the plan, through the preset operation plan optimization model, the activity sequence and activity time of the initial operation plan are optimized in combination with the non-core interests in the interest graph to obtain an optimized operation plan to achieve intelligent optimization of store operations.

[0005] Specifically, the user's cognitive level is analyzed according to the behavior data through a preset dynamic user stratification model to obtain multiple user levels, including: According to the user age in the behavior data, multiple age intervals are divided according to the time axis to obtain age groups; According to the age groups, by analyzing the cognitive level of each group of users, a cognitive index value corresponding to each group of users is obtained; Combined with the cognitive index values, groups are merged when the similarity between groups is greater than a preset similarity threshold, and groups are split when the standard deviation of the cognitive index values ​​within a group is greater than a preset standard deviation threshold. The age groups are dynamically adjusted through a preset dynamic user stratification model to obtain multiple user levels.

[0006] Specifically, the user age in the behavior data is divided into multiple age intervals according to the time axis to obtain age groups, including: According to the user age in the behavior data, calculate the corresponding age density; According to the age density, a plurality of seed points are generated; According to the seed point, the time axis is divided into a plurality of age intervals to obtain initial age intervals, wherein each seed point corresponds to an initial age interval; Between every two adjacent initial age intervals, an overlapping area is set according to a preset interval length; According to the user density in the overlapping area, the users are allocated to corresponding initial age ranges to obtain age groups.

[0007] Specifically, the user's interest tags are analyzed based on the behavior data, and the interest frequency of each user level is analyzed through a preset user interest graph construction model to construct an interest graph corresponding to each user level, wherein the user level in the interest graph is a root node, the interest tags are child nodes, and the corresponding root nodes and child nodes are connected through the interest frequency, including: Based on the behavior data, keywords are extracted through the preset keyword extraction model to obtain high-frequency behavior keywords; By using a preset keyword mapping model, the high-frequency behavior keywords are mapped to corresponding tags to obtain user interest tags; According to the interest tags combined with the user levels, the interest frequency of each user level is analyzed through a preset user interest graph construction model to construct an interest graph corresponding to each user level.

[0008] Specifically, the interest frequency of each user level is analyzed according to the interest tags combined with the user level through a preset user interest graph construction model to construct an interest graph corresponding to each user level, including: According to the user level corresponding to the interest tags, the interest frequency is obtained by analyzing the user's operation frequency and duration for different interest tags; The model is constructed by using a preset user interest graph, with the user level as the root node, the interest tag as the child node, and the interest frequency as the edge weight to connect the corresponding root node and child node to obtain an initial interest graph; In combination with real-time user data, when the interest frequency change value within the same user level is greater than a preset first threshold, the initial interest graph is updated to obtain an interest graph corresponding to each user level.

[0009] Specifically, the core interests are screened out from the interest graph, mapped in a preset activity template library through a preset operation plan generation model, and a corresponding initial operation plan is generated, including: According to the weights of the edges in the interest graph, select interest tags whose interest frequency values ​​are greater than a preset second threshold as core interests; Based on the core interests, a preset operation plan generation model is mapped with a preset activity template library to obtain corresponding operation activities; The operation activities are optimized through a preset activity optimization model to obtain a corresponding initial operation plan.

[0010] Specifically, the operation activities are optimized by using a preset activity optimization model to obtain a corresponding initial operation plan, including: According to the operational activities, the activity sequence and duration are arranged through the preset activity arrangement model to obtain the initial activity arrangement; Within a preset time period, the initial activity arrangement is optimized through a preset activity optimization model to obtain a corresponding initial operation plan.

[0011] Specifically, according to the effect of the plan, the activity sequence and activity time of the initial operation plan are optimized by using a preset operation plan optimization model and combining the non-core interests in the interest graph to obtain an optimized operation plan to achieve intelligent optimization of store operations, including: Calculate the effect score of the initial operation plan based on the plan effect; When the effect rating value is less than a preset third threshold, the activity sequence and activity time of the initial operation plan are optimized through a preset operation plan optimization model in combination with non-core interests in the interest graph to obtain an optimized operation plan.

[0012] Specifically, when the effect score value is less than a preset third threshold, the activity sequence and activity time of the initial operation plan are optimized by combining the non-core interests in the interest graph through a preset operation plan optimization model to obtain an optimized operation plan, including: When the effect score is less than the preset third threshold, the candidate solutions are generated by combining the non-core interests in the interest graph through the preset strategy generation model; According to the interest graph, the activities in the initial operation plan and the candidate plan are sorted to obtain an activity sequence; Based on the activity sequence, the activity time is optimized through a preset operation plan optimization model to obtain an optimized operation plan.

[0013] A store operation intelligent optimization system, used to implement the store operation intelligent optimization method, comprising: Data acquisition module, which obtains the behavior data of users in the store; A user grading module, which analyzes the user's cognitive level through a preset dynamic user stratification model according to the behavior data to obtain multiple user levels; An interest graph construction module analyzes the user's interest tags based on the behavior data, analyzes the interest frequency of each user level through a preset user interest graph construction model, and constructs an interest graph corresponding to each user level, wherein the user level in the interest graph is a root node, the interest tag is a child node, and the corresponding root node and child node are connected through the interest frequency; An operation plan generation module selects core interests from the interest map, maps them in a preset activity template library through a preset operation plan generation model, and generates a corresponding initial operation plan; An operation plan evaluation module evaluates the effect of the initial operation plan through a preset plan effect evaluation model to obtain the plan effect; The operation plan optimization module optimizes the activity sequence and activity time of the initial operation plan according to the effect of the plan through a preset operation plan optimization model and combines the non-core interests in the interest graph to obtain an optimized operation plan to achieve intelligent optimization of store operations.

[0014] Compared with the prior art, this application has the following beneficial effects: This application combines user dynamic stratification to build a corresponding user interest map, develops personalized operation plans for users at different levels, and dynamically optimizes the operation plan based on the effect of the plan, which can more accurately understand the characteristics and needs of different user levels. The constructed interest map can reflect the changes in user interests in real time, making the operation plan more in line with the user's actual interests; through the collaborative work of multiple models, the generated operation plan is more scientific and reasonable, which can effectively improve operational efficiency and results; dynamically optimize the operation plan based on the effect of the plan, which can timely adapt to market changes and user needs, and enhance the competitiveness of the store BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a workflow diagram of a store operation intelligent optimization method in Embodiment 1 of the present invention; Figure 2 A schematic diagram of constructing an interest graph in Example 1 of the present invention; Figure 3 It is a schematic diagram of mapping the core interests and the activity template library in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the structure of a store operation intelligent optimization system in Example 2 of the present invention. DETAILED DESCRIPTION

[0016] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0017] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0018] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0019] Embodiment 1: This embodiment provides a method for intelligent optimization of store operations. Figure 1 As shown, the store operation intelligent optimization method comprises: S101. Obtaining behavior data of users in the store; S102, analyzing the user's cognitive level through a preset dynamic user stratification model according to the behavior data to obtain multiple user strata; S103, analyzing the user's interest tags based on the behavior data, analyzing the interest frequency of each user level through a preset user interest graph construction model, and constructing an interest graph corresponding to each user level, wherein the user level in the interest graph is a root node, the interest tags are child nodes, and the corresponding root nodes and child nodes are connected through the interest frequency; S104, selecting core interests from the interest graph, mapping them in a preset activity template library through a preset operation plan generation model, and generating a corresponding initial operation plan; S105, evaluating the effect of the initial operation plan through a preset plan effect evaluation model to obtain the plan effect; S106. According to the effect of the plan, the activity sequence and activity time of the initial operation plan are optimized through a preset operation plan optimization model in combination with non-core interests in the interest graph to obtain an optimized operation plan to achieve intelligent optimization of store operations.

[0020] This embodiment dynamically stratifies users, constructs corresponding interest graphs, and formulates personalized operation plans based on the interest graphs, so that the operation plans can accurately reach the target user groups; compared with the existing single operation plan formulation method, by dynamically adjusting user stratification, interest graphs, and operation plans, it can be optimized according to real-time data and plan effects, ensuring that store operations always adapt to market changes and user needs.

[0021] In this embodiment, cameras are arranged in the store to capture the user's walking path, stay area, products of interest and other behavioral data; the store's sales system is used to collect the user's purchase records, including the type, quantity, and time of purchase of the purchased products; the learning robot in the store can obtain the user's operation records and the usage time of each function, and collect the user's browsing history, search history, favorite products and other behavioral data on the online platform; by comprehensively and accurately obtaining user behavior data, a sufficient data basis is provided for analyzing user characteristics and needs.

[0022] Specifically, users are dynamically stratified based on the collected behavioral data, and users are stratified through a preset dynamic user stratification model to obtain multiple user levels. The preset dynamic user stratification model can be a cluster analysis model, a decision tree model, or other machine learning model to dynamically adjust the user stratification results. Through user stratification, differences in cognitive levels among users of different ages are taken into account. By analyzing the age information in the behavioral data and combining the user's behavioral performance in the store, such as the understanding of product information, the depth of inquiries, and the operation records on the learning robot, the dynamic user stratification model is used to divide users with different cognitive levels into levels. Accurate user stratification enables stores to formulate more targeted marketing strategies and service methods for users with different cognitive levels. For example, for children with higher cognitive levels, more difficult and professional learning activities and learning products can be provided; for children with lower cognitive levels, more basic learning content and products can be provided, thereby improving operational results and user satisfaction.

[0023] Specifically, after stratifying the users, an interest analysis is conducted on each layer of users, a model is built through a preset user interest graph, an interest graph corresponding to each user level is constructed, keywords in the user behavior data are extracted and mapped to interest tags, and then combined with user levels and the frequency and duration of user operations on interest tags, a graph that can reflect the interest preferences of different user levels is constructed; the constructed interest graph can intuitively display the interest preferences of different user levels, and help stores clearly understand the interests of target users; when planning activities, displaying goods, and pushing information, stores can more accurately meet user needs based on the interest graph, and improve user participation and purchasing intention; for example, according to the interest graph, it is found that a certain user level has a strong interest in a certain type of electronic product, and the store can increase the display and promotion of related products in that area.

[0024] At the same time, according to the constructed interest graph, the corresponding operation plan is generated. Through the preset operation plan generation model, based on the constructed interest graph, interest tags with high weight values ​​are selected as core interests, which are mapped with the preset activity template library to generate corresponding operation activities, and further optimized to obtain the initial operation plan; the generated initial operation plan can closely revolve around user interests, has high pertinence and feasibility, quickly and effectively attracts user attention, improves the effectiveness of operation activities, and greatly increases the probability of success compared to the traditional operation plan based on experience.

[0025] Specifically, after the initial operation plan is generated, the effect of the initial operation plan is evaluated through a preset plan effect evaluation model, and the effect of the initial operation plan after implementation is quantitatively evaluated from multiple dimensions using preset evaluation indicators and algorithms to obtain the plan effect. The preset plan effect evaluation model can be a hierarchical analysis method, a fuzzy comprehensive evaluation method, etc. The plan effect evaluation model of this embodiment is specifically a hierarchical analysis method, which establishes a correlation between the evaluation indicators and the initial operation plan, and constructs a judgment matrix in combination with the effect of the activity corresponding to the indicator to obtain the weight of each evaluation indicator. The comprehensive score of the plan effect is calculated by weighted summation to obtain the corresponding plan. Through objective and scientific evaluation, we can evaluate the actual effect of the initial operation plan, find out the advantages and disadvantages of the plan, and optimize the plan to avoid blindly adjusting the operation plan; according to the evaluation results of the plan effect, when the effect does not meet expectations, combined with information such as non-core interests in the interest map, use the operation plan optimization model to generate candidate plans, and optimize the activity sequence and time to obtain a better operation plan; by continuously optimizing the operation plan, we can continuously improve the store operation effect, adapt to market changes and dynamic adjustments of user needs, so that store operations always remain at a high level, and improve the store's competitiveness and profitability.

[0026] This application combines user dynamic stratification to construct a corresponding user interest map, formulates personalized operation plans for users at different levels, and dynamically optimizes the operation plans based on the results of the plans. It can more accurately understand the characteristics and needs of different user levels. The constructed interest map can reflect the changes in user interests in real time, making the operation plan more in line with the user's actual interests; through the collaborative work of multiple models, the generated operation plan is more scientific and reasonable, and can effectively improve operational efficiency and results; dynamically optimize the operation plan based on the results of the plan, and can timely adapt to market changes and user needs, thereby enhancing the competitiveness of the store.

[0027] Furthermore, according to the behavior data, the user's cognitive level is analyzed by a preset dynamic user stratification model to obtain multiple user levels, including: S201, dividing the user age in the behavior data into multiple age intervals according to the time axis to obtain age groups; S202, analyzing the cognitive level of each group of users according to the age groups to obtain a cognitive index value corresponding to each group of users; S203, combining the cognitive index values, merging when the inter-group similarity is greater than a preset similarity threshold, splitting when the standard deviation of the cognitive index values ​​within a group is greater than a preset standard deviation threshold, and dynamically adjusting the age groupings through a preset dynamic user stratification model to obtain multiple user levels.

[0028] In this embodiment, users are grouped according to their ages in the collected user behavior data, and multiple age intervals are divided according to the time axis to obtain age groups. Considering that users of different age groups often have differences in consumption habits, interests and hobbies, by grouping users by age, users with similar characteristics can be preliminarily classified into one category, which can intuitively reflect the distribution of different age stages.

[0029] Specifically, based on age groups, analyze the behavior of each group of users in the store. These behaviors can reflect the user's understanding and cognition of goods, services, etc., and quantify the user's cognitive level into a cognitive index value, which is convenient for comparing and distinguishing the cognitive levels of users in different age groups; set a series of indicators for measuring cognitive levels, and assign corresponding weights to each indicator. The cognitive index value of each user is calculated by weighted summation through the quantitative data of the user on each indicator used to measure cognitive level and the corresponding weight. For example, the weight of the time spent viewing the product manual is set to 0.3, the weight of the complexity of the questions asked to the clerk is set to 0.4, the weight of the time spent in the product area is set to 0.2, and the weight of the understanding and reaction to the activity rules is set to 0.1; according to the set indicators and weights, calculate the cognitive index value of each user, and average the cognitive index values ​​of all users in each age group to obtain the corresponding cognitive index value of each group of users; by calculating the cognitive index value, it helps stores to more accurately understand the cognitive characteristics of users of different age groups, and provide a basis for formulating targeted marketing strategies and service plans.

[0030] Specifically, after obtaining the cognitive index value corresponding to each age group, by comparing the inter-group similarity (including the cosine similarity, Euclidean distance, Pearson correlation coefficient, etc. between the cognitive index values ​​corresponding to the age groups) and the discrete degree (standard deviation) of the cognitive index values ​​within the group, the initial age grouping is optimized and adjusted using a preset dynamic user stratification model. The preset dynamic user stratification model can be a machine learning model such as a cluster analysis model and a decision tree model. The dynamic user stratification model in this embodiment is specifically a random forest model. The random forest model is trained by a large amount of historical grouping data to obtain a pre-trained dynamic user stratification model. Each age group, the inter-group similarity and the discrete degree of the cognitive index value within the group are input into the pre-trained dynamic user stratification model. When the inter-group similarity is high, it means that the two groups of users are relatively similar in cognitive level and can be merged into one group; when the standard deviation of the cognitive index value within the group is large, it means that the cognitive level differences of the users within the group are large, and they need to be further split into different levels to more accurately reflect the real situation of the users. The model outputs the user stratification results after merging or splitting; by dynamically adjusting the user stratification, the grouping results can more accurately reflect the differences in the cognitive levels of the users. This avoids the irrationality caused by simple age grouping, allowing stores to formulate more refined and personalized operation strategies for users with different cognitive levels.

[0031] Exemplarily, by calculating the cosine similarity of the cognitive index between age groups, the similarity between age groups, i.e., the inter-group similarity, is obtained, and the calculated inter-group similarity is compared with a preset similarity threshold, and the preset similarity threshold is set to 0.95. When the similarity of age groups A and B, 0.98, is greater than the threshold, the two age groups are merged into a new group. At the same time, the standard deviation of the cognitive index values ​​within each age group is calculated, and the calculated intra-group standard deviation is compared with the preset standard deviation threshold, and the preset standard deviation threshold is set to 0.8. When the standard deviation of the intra-group cognitive index values, 0.96, is greater than the threshold, it indicates that the cognitive levels of users in the group vary greatly and the group needs to be split. After the merging and splitting operations, the age groups are updated through the preset dynamic user stratification model to obtain multiple user levels.

[0032] Furthermore, the user age in the behavior data is divided into multiple age intervals according to the time axis to obtain age groups, including: S301, calculating the corresponding age density according to the user age in the behavior data; S302, generating a plurality of seed points according to the age density; S303, dividing the time axis into multiple age intervals according to the seed point to obtain initial age intervals, wherein each seed point corresponds to an initial age interval; S304, setting an overlapping area between every two adjacent initial age intervals according to a preset interval length; S305 . Allocate users in the overlapping area to corresponding initial age ranges according to user density to obtain age groups.

[0033] In this embodiment, age density is calculated based on the age of users to measure the distribution of users of different age groups in the total number of users. By calculating the age density, it is possible to understand which age groups have relatively more users and which have relatively fewer users, which provides a basis for determining seed points and dividing age intervals, and reflects the concentration trend and distribution characteristics of user ages based on the statistical frequency of users of different age groups.

[0034] Specifically, the calculated age density data is sorted to find out the age interval with higher age density, and a representative age is selected as a seed point in each age interval with higher age density. In this embodiment, the middle value of the interval is taken. For example, in the 10-15 years old interval, 12 years old is selected as the seed point; in the 15-20 years old interval, 17 years old is selected as the seed point; the time axis is divided based on the seed point, and the users are divided into different intervals according to their age. Each interval corresponds to a seed point. By making the seed point correspond to the age interval, it can be ensured that each age interval revolves around a representative age, so that users in the same interval have a certain similarity in age. With each seed point as the center, the interval is expanded according to the preset width of the initial age interval. In this embodiment, with the seed point as the center, 2 years are expanded on the left and right as the width of the initial age interval, and the initial age interval includes 10-14 years old and 15-19 years old.

[0035] At the same time, between every two adjacent initial age intervals, an overlapping area is set according to a preset interval length, and the length of the overlapping area is set according to business needs and data characteristics. By setting the overlapping area, the flexibility and accuracy of age grouping are increased, which can better reflect the continuity of user age characteristics and avoid missing some users with transitional characteristics due to simple division of age intervals; according to the user density of the overlapping area, the overlapping area is allocated to the corresponding age interval, and the user density characteristics in the overlapping area are compared with those in the adjacent initial age intervals. According to the comparison results, the users in the overlapping area are allocated to the initial age interval with more similar user density characteristics. For example, if the age range of the overlapping area is 13-16 years old, when the density of users who purchase specific products in the overlapping area is closer to the density of users in the 10-14 age range, the users in the overlapping area will be allocated to the 10-14 age range; otherwise, they will be allocated to the 15-19 age range. After allocating users in all overlapping areas, the final age grouping is obtained. By allocating according to the density of users in the overlapping areas, the age grouping results can be further optimized, making the users in each age group more similar in behavioral characteristics, which can provide a more reliable basis for stores to formulate targeted operation strategies and improve the effectiveness and accuracy of operation strategies.

[0036] Further, the user's interest tags are analyzed based on the behavior data, and the interest frequency of each user level is analyzed through a preset user interest graph construction model to construct an interest graph corresponding to each user level, wherein the user level in the interest graph is a root node, the interest tags are child nodes, and the corresponding root nodes and child nodes are connected through the interest frequency, including: S401, extracting keywords based on the behavior data through a preset keyword extraction model to obtain high-frequency behavior keywords; S402, mapping the high-frequency behavior keywords to corresponding tags through a preset keyword mapping model to obtain user interest tags; S403: According to the interest tags and the user levels, the interest frequency of each user level is analyzed through a preset user interest graph construction model to construct an interest graph corresponding to each user level.

[0037] In this embodiment, high-frequency behavioral keywords are identified from the behavioral data through a preset keyword extraction model, which is specifically a TF-IDF model. The word frequency of each word is calculated, that is, the number of times the word appears divided by the total number of words; then the inverse document frequency is calculated, which reflects the rarity of a word. It is obtained by dividing the total number of documents by the number of documents containing the word and then taking the logarithm. Finally, the word frequency and the inverse document frequency of each word are multiplied to obtain the TF-IDF value of the word. The higher the TF-IDF value of a word, the more important it is in the document and relatively uncommon in other documents. Words with higher weights are screened out as high-frequency behavioral keywords. By screening keywords, the user's behavioral information can be highly concentrated, and complex behavioral data can be simplified into a representative vocabulary set.

[0038] Specifically, according to the extracted keywords, through a preset keyword mapping model, the preset keyword mapping model can be a rule-based mapping model, a deep learning model and other models. The preset keyword mapping model in this embodiment is specifically a mapping model based on mapping rules. The high-frequency behavior keywords are mapped to corresponding tags to obtain the user's interest tags, and the keywords are mapped to a pre-defined interest tag system. The keyword mapping model establishes a corresponding mapping relationship between the keywords and the interest tags, and corresponds the specific behavior keywords to the interest tags. For example, when the high-frequency behavior keywords are "eraser" and "pencil", the high-frequency behavior keywords are corresponded to the interest tag "stationery"; combined with the analyzed user interest tags, the corresponding interest graph is constructed through the preset user interest graph construction model. The preset user interest graph construction model can be an association rule mining model, a graph embedding model, a deep learning model and other models. The user level is used as the root node of the graph, and the interest tag is used as the child node. The weight of the edge is used to represent the user's attention to the interest tag. By combining the user level and the interest tag to construct an interest graph, the interest preferences and differences of different user levels can be clearly presented, helping stores to formulate personalized operation strategies for different user groups.

[0039] Furthermore, the interest frequency of each user level is analyzed by combining the interest tags with the user level through a preset user interest graph construction model to construct an interest graph corresponding to each user level, including: S501, according to the user level corresponding to the interest tag, by analyzing the user's operation frequency and duration for different interest tags, the interest frequency is obtained; S502: construct a model through a preset user interest graph, use the user level as the root node, the interest tag as the child node, and the interest frequency as the edge weight to connect the corresponding root node and the child node to obtain an initial interest graph; S503. In combination with real-time user data, when the interest frequency change value within the same user level is greater than a preset first threshold, the initial interest graph is updated to obtain an interest graph corresponding to each user level.

[0040] In this embodiment, the operation frequency and duration of users in each user level for different interest tags are analyzed respectively to obtain the interest frequency. The user's operation behavior on the goods or content associated with different interest tags reflects the degree of his / her attention to the interest. The operation frequency reflects the user's active participation in a specific interest, and the operation duration reflects the depth of the user's focus on the interest. By comprehensively analyzing the data of these two dimensions, the user's interest in each interest tag can be quantified, and the user's operation frequency and operation duration on the corresponding interest tag are weighted and summed to obtain the interest frequency corresponding to each interest tag. By calculating the interest frequency, the correlation strength between the user level and the interest tag can be accurately measured.

[0041] Specifically, Figure 2 As shown, a model is constructed by a preset user interest graph. The user interest graph construction model in this embodiment is specifically a graph embedding model. An interest graph is constructed according to the calculated interest frequency. The user level is taken as the root node, the interest tag as the child node, and the interest frequency as the weight of the edge to obtain a corresponding interest graph. By constructing an interest graph, the interest distribution and interest intensity of different user levels are reflected. User interests are not static. With the passage of time, changes in the market environment, and the emergence of new products or services, user interests will change. By real-time monitoring of user data, the change value of the interest frequency is calculated and compared with the preset first threshold. When the change value exceeds the preset first threshold, it means that the user interest of the corresponding level has changed significantly, and the interest graph is updated. By updating the interest graph in real time, the store can always keep up with the dynamic changes of user interests and adjust the operation strategy in time.

[0042] Furthermore, the core interests are screened out from the interest graph, mapped in a preset activity template library through a preset operation plan generation model, and a corresponding initial operation plan is generated, including: S601, according to the weight of the edge in the interest graph, selecting the interest tag whose interest frequency value is greater than the preset second threshold as the core interest; S602: Based on the core interests, a preset operation plan generation model is mapped with a preset activity template library to obtain corresponding operation activities; S603: Optimize the operation activity through a preset activity optimization model to obtain a corresponding initial operation plan.

[0043] In this embodiment, according to the weight of the edge in the interest graph, the interest tags are screened, and the interest tags whose interest frequency value (corresponding to the value of interest frequency) is greater than the preset second threshold are selected as core interests. The weight of the edge in the interest graph reflects the user's attention and interest intensity to different interest tags. The preset second threshold is set according to the actual business situation and data characteristics, and all interest tags in the interest graph are traversed to obtain the weight value of each interest tag, that is, the interest frequency value. The weight value of each interest tag is compared with the preset second threshold. When the weight value of a certain interest tag is greater than the preset second threshold, it is selected as the core interest; by selecting the core interest, it can help the operator quickly grasp the main interest points of the user group and avoid distraction when formulating the operation plan. Guided by the core interest, the operation plan can be more targeted, directly hit the key needs of the user, and improve the effectiveness and attractiveness of the operation plan, thereby better attracting user participation and improving user satisfaction.

[0044] Specifically, Figure 3 As shown, interest tags with weights greater than 0.5 are selected as core interests. According to the selected core interests, the preset operation plan generation model is used to map with the preset activity template library to obtain corresponding operation activities. The preset activity template library stores a variety of designed operation activity templates suitable for different scenarios and interest points. The operation plan generation model is specifically a random forest model, which is trained through a large amount of historical data to obtain a pre-trained operation plan generation model. The core interests are input into the pre-trained operation plan generation model. The operation plan generation model searches and matches in the activity template library according to the learned mapping relationship to obtain the corresponding operation activities. By generating operation activities related to the core interests, the time and cost of designing operation activities are saved, and the generated operation activities can be guaranteed to have high feasibility and attractiveness, thereby improving the quality and success rate of the operation activities.

[0045] At the same time, the effects of the operation activities are evaluated, and the initially generated operation activities are optimized from multiple dimensions through the preset activity optimization model according to the evaluation results. The preset activity optimization model includes but is not limited to the particle swarm optimization model. Optimizing the operation activities through the activity optimization model can ensure that the operation plan has higher feasibility and effectiveness in actual implementation. The optimized operation plan can better balance the activity cost and expected benefits, improve user participation and satisfaction, and enhance the competitiveness of the activity in the market, thereby bringing better operation results and economic benefits to the store.

[0046] Furthermore, the operation activities are optimized by using a preset activity optimization model to obtain a corresponding initial operation plan, including: S701. Arrange the activity sequence and duration according to the operation activities through a preset activity arrangement model to obtain an initial activity arrangement; S702: Within a preset time period, the initial activity arrangement is optimized by using a preset activity optimization model to obtain a corresponding initial operation plan.

[0047] In this embodiment, the sequence and duration of the activities are arranged according to the matched multiple operational activities. Different operational activities have different characteristics and goals. Reasonable arrangement of the sequence and duration of the activities can make them cooperate with each other to achieve the maximum effect. The activity process is planned through a preset activity arrangement model. The preset activity arrangement model includes but is not limited to a genetic algorithm model. The genetic algorithm model is trained through a large number of historical operational activities to obtain a pre-trained activity arrangement model. The activity sequence and duration of multiple activities are arranged to achieve the best activity effect. The various sub-activities included in the operational activities and their goals are clarified, and the sub-activities are classified according to their nature. For example, publicity activities focus on the scope and speed of information dissemination, experience activities emphasize user participation and interactivity, and promotion activities focus on purchase conversion rate. The sequence and duration of activities are formulated according to different users corresponding to different times. Through reasonable initial activity arrangements, the operational activities can be carried out in an orderly manner. The various sub-activities cooperate with each other to improve the attractiveness and participation of the activities, avoid the blindness of the activities, and improve the overall effect of the operational activities.

[0048] For example, based on the rules of user online activity, online preheating promotion is started at 7 pm on weekdays and lasts for 3 days to fully cover the target user group; then offline learning experience activities are carried out on weekends for 2 days, which is in line with the characteristics of users' offline participation time; then, an online limited-time discount activity is launched at 8 pm on weekdays next week for 1 day, using users' free time after school to promote purchases; finally, after-sales feedback activities are carried out within a week after the end of the activity to consolidate user relationships. Through such arrangements, the initial activity schedule is obtained.

[0049] Specifically, based on the initial activity arrangement, the initial activity arrangement is optimized through a preset activity optimization model, which includes but is not limited to a particle swarm optimization model. The particle swarm optimization model is trained through a large amount of activity execution data to obtain a pre-trained activity optimization model. The model performs optimization steps such as sequence optimization and duration optimization on multiple activities, searches for the optimal activity arrangement plan, and obtains an optimized initial operation plan; within a preset time period, the real-time data during the activity execution process is analyzed to adjust the activity sequence, duration, resource allocation, etc. to achieve better operation results; according to the scale and nature of the operation activities, a suitable time period is determined for optimization analysis; for a one-month promotion activity, an optimization evaluation is performed on a weekly basis; during the activity execution process, various data are collected in real time, and the collected data is input into the activity optimization model. The model analyzes the reasons for low user participation or sales not meeting expectations. According to the analysis results, the activity optimization model uses an algorithm to optimize and adjust the initial activity arrangement. By optimizing the initial activity schedule within the preset time period, problems in the activity execution process can be discovered in a timely manner, and adjustments can be made based on actual conditions to make the operation plan more in line with market changes and user needs, thereby improving the success rate and operational effectiveness of the activity, maximizing the benefits of the operational activities, and bringing better economic benefits and market competitiveness to the store.

[0050] For example, when the purchase conversion rate of an online limited-time discount event is low, the model recommends increasing the discount, extending the event duration by half a day, optimizing the event page design, and increasing the loading speed; when the number of participants in an offline new product experience event is small, the model recommends adding a round of targeted publicity and promotion before the event, and adjusting the event time to a more appropriate time period on weekend afternoons; through these adjustments, an optimized initial operation plan is obtained.

[0051] Furthermore, according to the effect of the plan, the activity sequence and activity time of the initial operation plan are optimized by using a preset operation plan optimization model and combining the non-core interests in the interest graph to obtain an optimized operation plan to achieve intelligent optimization of store operations, including: S801. Calculate the effect score of the initial operation plan according to the effect of the plan; S802. When the effect score value is less than a preset third threshold, the activity sequence and activity time of the initial operation plan are optimized by combining the non-core interests in the interest graph through a preset operation plan optimization model to obtain an optimized operation plan.

[0052] In this embodiment, the effect score of the initial operation plan is calculated according to the effect of the plan, and a comprehensive effect score is calculated by collecting various types of data related to the implementation of the plan; various indicators for evaluating the effect of the operation plan are determined, including sales growth, customer flow changes, user participation, etc., and corresponding weights are assigned to each indicator according to the importance of each indicator to store operations and in combination with the store's business objectives; the collected data are weighted according to the set weights to obtain an effect score; the calculated effect score is compared with a preset third threshold value. When the effect score is less than the preset third threshold value, the initial operation plan is optimized by a preset operation plan optimization model to obtain an optimized operation plan. The preset operation plan optimization model can be a decision tree model, a reinforcement learning model, or other models. The preset operation plan optimization model can perform scientific analysis based on multiple factors, quickly generate effective optimization strategies, and improve the effect of the operation plan. The optimized operation plan can better adapt to market changes and user needs, help improve the operating efficiency and competitiveness of the store, and achieve intelligent optimization of store operations.

[0053] Specifically, when it is determined that optimization is needed, the initial operation plan and related data are input into the preset operation plan optimization model, and the model analyzes the plan to find out the key factors that affect the effectiveness of the plan; for example, the model analysis finds that the preferential strength of the promotional activities in the initial operation plan is not enough to attract users, resulting in sales growth that does not meet expectations; improper selection of event publicity channels has limited the increase in customer flow; based on the analysis results of the model, specific optimization strategies are generated; to address the problem of insufficient preferential strength of promotional activities, the model recommends increasing the discount range, launching full-reduction activities or gift strategies; for publicity channel issues, the model recommends adjusting the publicity channel combination, increasing publicity investment on social media platforms where the target user groups are active, and adjusting the initial operation plan according to the generated optimization strategy to obtain an optimized operation plan.

[0054] Furthermore, when the effect score value is less than a preset third threshold, the activity sequence and activity time of the initial operation plan are optimized by combining the non-core interests in the interest graph through a preset operation plan optimization model to obtain an optimized operation plan, including: S901, when the effect score value is less than the preset third threshold, combining the non-core interests in the interest graph and using the preset strategy generation model to generate candidate solutions; S902, sorting the activities in the initial operation plan and the candidate plans according to the interest graph to obtain an activity sequence; S903: Based on the activity sequence, the activity time is optimized by using a preset operation plan optimization model to obtain an optimized operation plan.

[0055] In this embodiment, when the effect score value is less than the preset third threshold, it indicates that the effect of the operation plan has not met expectations, and the original plan is insufficient in mining and satisfying user interests. The interest map includes core interests and non-core interests. Non-core interests also reflect the potential needs of users. These non-core interests are analyzed by a preset strategy generation model, and the pre-set rules and algorithms in the model are used to combine market dynamics, user behavior patterns and other factors to generate candidate plans that can better attract users and improve the effect of the plan; the preset strategy generation model includes but is not limited to a genetic algorithm model, and the genetic algorithm model is trained using a large amount of historical data to obtain a pre-trained strategy generation model. The model combines non-core interests to evaluate the activity effect of the corresponding activity based on mechanisms such as inheritance, mutation and selection, and generates candidate plans by mining non-core interests, which provides new ideas and directions for optimizing the operation plan, expands the coverage of operation activities, meets more diverse needs of users, and avoids conducting activities only around core interests while ignoring users' potential interests.

[0056] Specifically, according to the interest map, the activities in the initial operation plan and the candidate plan are sorted to obtain the activity sequence, which reflects the user's attention to different interests and the correlation between interests. The activities are sorted according to the interest map so that they are carried out in sequence according to the priority and correlation logic of user interests, thereby improving user participation and activity effects. Reasonable activity sorting can enhance the coherence and attractiveness of operation activities, and activities can be carried out according to the logic of user interests, making it easier for users to participate in them, thereby improving users' attention to and enthusiasm for participation in activities.

[0057] Furthermore, according to the sorted activity sequence, the activity time is optimized through a preset operation plan optimization model to obtain an optimized operation plan. The preset operation plan optimization model can be a decision tree model, a reinforcement learning model, or other models. The operation plan optimization model in this embodiment is specifically a particle swarm optimization model. The particle swarm optimization model is trained using a large amount of historical activity data to obtain a pre-trained operation plan optimization model. According to the previously determined activity sequence, the characteristics of each activity and the target user group are analyzed, and the data is input into the preset operation plan optimization model. The model optimizes and adjusts the activity time to search for the optimal solution to obtain an optimized operation plan. By optimizing the activity time, user participation and activity effect can be improved. At the same time, the reasonable arrangement of activity time also helps to smoothly connect activities and improve operational efficiency, thereby realizing intelligent optimization of store operations and bringing better economic benefits to stores.

[0058] Embodiment 2: In this embodiment, if Figure 4 , providing a store operation intelligent optimization system, used to implement the store operation intelligent optimization method, including: Data acquisition module, which obtains the behavior data of users in the store; A user grading module, which analyzes the user's cognitive level through a preset dynamic user stratification model according to the behavior data to obtain multiple user levels; An interest graph construction module analyzes the user's interest tags based on the behavior data, analyzes the interest frequency of each user level through a preset user interest graph construction model, and constructs an interest graph corresponding to each user level, wherein the user level in the interest graph is a root node, the interest tag is a child node, and the corresponding root node and child node are connected through the interest frequency; An operation plan generation module selects core interests from the interest map, maps them in a preset activity template library through a preset operation plan generation model, and generates a corresponding initial operation plan; An operation plan evaluation module evaluates the effect of the initial operation plan through a preset plan effect evaluation model to obtain the plan effect; The operation plan optimization module optimizes the activity sequence and activity time of the initial operation plan according to the effect of the plan through a preset operation plan optimization model and combines the non-core interests in the interest graph to obtain an optimized operation plan to achieve intelligent optimization of store operations.

[0059] In this embodiment, the data acquisition module includes components such as a data collection interface and a data storage unit. This module is responsible for collecting various behavioral data of users in the store, providing a data basis for user analysis, operation plan formulation, etc., and integrating data from different sources to ensure the comprehensiveness and accuracy of the data so that the system can fully understand the user's behavioral performance in the store; the user classification module includes a data preprocessing unit, a hierarchical model calculation unit and a result output unit. This module analyzes the user's cognitive level based on the behavioral data and a preset dynamic user stratification model, and divides the users into multiple different levels. More targeted operation strategies can be formulated for users at different levels to improve operational results and resource utilization efficiency.

[0060] Specifically, the interest graph construction module includes an interest tag extraction unit, a frequency calculation unit, and a graph construction unit. This module constructs an interest graph that reflects the interest characteristics of different user levels by analyzing behavioral data, so as to formulate personalized operation plans based on the user's interest distribution and preference level; the operation plan generation module includes an interest weight screening unit, an activity mapping unit, and a plan generation unit. This module generates an initial operation plan that meets the user's interest characteristics based on the interest graph and a preset operation plan generation model, so that the operation activities can accurately target the user's core interests and improve user participation and operation results.

[0061] Specifically, the operation plan evaluation module includes an evaluation index setting unit, a data collection unit and an effect evaluation unit. This module uses a preset plan effect evaluation model to objectively and comprehensively evaluate the implementation effect of the initial operation plan, and provides direction for plan optimization, so as to timely discover problems and deficiencies in the operation plan; the operation plan optimization module includes a candidate plan generation unit, an activity ranking unit and a time optimization unit. This module optimizes the initial operation plan according to the evaluation results of the operation plan, fully considers the non-core interests in the interest graph, and generates a more comprehensive and reasonable operation plan. By optimizing the sequence and time of activities, the overall effect of the operation plan is improved, the intelligent optimization of store operations is realized, and the operating efficiency of the store is improved.

[0062] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for intelligent optimization of store operations, characterized in that: include: Obtain behavioral data of users in the store; According to the behavior data, the user's cognitive level is analyzed through a preset dynamic user stratification model to obtain multiple user strata; Analyze the user's interest tags based on the behavior data, analyze the interest frequency of each user level through a preset user interest graph construction model, and construct an interest graph corresponding to each user level, wherein the user level in the interest graph is a root node, the interest tags are child nodes, and the corresponding root nodes and child nodes are connected through the interest frequency; Filtering core interests from the interest graph, mapping them in a preset activity template library through a preset operation plan generation model, and generating a corresponding initial operation plan; Evaluate the effect of the initial operation plan through a preset plan effect evaluation model to obtain the plan effect; According to the effect of the plan, through the preset operation plan optimization model, the activity sequence and activity time of the initial operation plan are optimized in combination with the non-core interests in the interest graph to obtain an optimized operation plan to achieve intelligent optimization of store operations.

2. A store operation intelligent optimization method according to claim 1, characterized in that: According to the behavior data, the user's cognitive level is analyzed through a preset dynamic user stratification model to obtain multiple user levels, including: According to the user age in the behavior data, multiple age intervals are divided according to the time axis to obtain age groups; According to the age groups, by analyzing the cognitive level of each group of users, a cognitive index value corresponding to each group of users is obtained; Combined with the cognitive index values, groups are merged when the similarity between groups is greater than a preset similarity threshold, and groups are split when the standard deviation of the cognitive index values ​​within a group is greater than a preset standard deviation threshold. The age groups are dynamically adjusted through a preset dynamic user stratification model to obtain multiple user levels.

3. A store operation intelligent optimization method according to claim 2, characterized in that: The method of dividing the user age in the behavior data into multiple age intervals according to the time axis to obtain age groups includes: According to the user age in the behavior data, calculate the corresponding age density; According to the age density, a plurality of seed points are generated; According to the seed point, the time axis is divided into a plurality of age intervals to obtain initial age intervals, wherein each seed point corresponds to an initial age interval; Between every two adjacent initial age intervals, an overlapping area is set according to a preset interval length; According to the user density in the overlapping area, the users are allocated to corresponding initial age ranges to obtain age groups.

4. A store operation intelligent optimization method according to claim 1, characterized in that: The method of analyzing the user's interest tags based on the behavior data, analyzing the interest frequency of each user level through a preset user interest graph construction model, and constructing an interest graph corresponding to each user level, wherein the user level in the interest graph is a root node, the interest tags are child nodes, and the corresponding root nodes and child nodes are connected through the interest frequency, including: Based on the behavior data, keywords are extracted through the preset keyword extraction model to obtain high-frequency behavior keywords; By using a preset keyword mapping model, the high-frequency behavior keywords are mapped to corresponding tags to obtain user interest tags; According to the interest tags combined with the user levels, the interest frequency of each user level is analyzed through a preset user interest graph construction model to construct an interest graph corresponding to each user level.

5. A store operation intelligent optimization method according to claim 4, characterized in that: The method of analyzing the interest frequency of each user level by combining the interest tags with the user level through a preset user interest graph construction model to construct an interest graph corresponding to each user level includes: According to the user level corresponding to the interest tags, the interest frequency is obtained by analyzing the user's operation frequency and duration for different interest tags; The model is constructed by using a preset user interest graph, with the user level as the root node, the interest tag as the child node, and the interest frequency as the edge weight to connect the corresponding root node and child node to obtain an initial interest graph; In combination with real-time user data, when the interest frequency change value within the same user level is greater than a preset first threshold, the initial interest graph is updated to obtain an interest graph corresponding to each user level.

6. A store operation intelligent optimization method according to claim 1, characterized in that: The core interests are selected from the interest graph, mapped in a preset activity template library through a preset operation plan generation model, and a corresponding initial operation plan is generated, including: According to the weights of the edges in the interest graph, select interest tags whose interest frequency values ​​are greater than a preset second threshold as core interests; Based on the core interests, a preset operation plan generation model is mapped with a preset activity template library to obtain corresponding operation activities; The operation activities are optimized through a preset activity optimization model to obtain a corresponding initial operation plan.

7. A store operation intelligent optimization method according to claim 6, characterized in that: The operation activities are optimized by using a preset activity optimization model to obtain a corresponding initial operation plan, including: According to the operational activities, the activity sequence and duration are arranged through the preset activity arrangement model to obtain the initial activity arrangement; Within a preset time period, the initial activity arrangement is optimized through a preset activity optimization model to obtain a corresponding initial operation plan.

8. The method for intelligent optimization of store operations according to claim 1, characterized in that: According to the scheme effect, the activity sequence and activity time of the initial operation scheme are optimized by using a preset operation scheme optimization model and combining the non-core interests in the interest graph to obtain an optimized operation scheme to achieve intelligent optimization of store operations, including: Calculate the effect score of the initial operation plan based on the plan effect; When the effect rating value is less than a preset third threshold, the activity sequence and activity time of the initial operation plan are optimized through a preset operation plan optimization model in combination with non-core interests in the interest graph to obtain an optimized operation plan.

9. A store operation intelligent optimization method according to claim 8, characterized in that: When the effect score value is less than a preset third threshold, the activity sequence and activity time of the initial operation plan are optimized by combining the non-core interests in the interest graph through a preset operation plan optimization model to obtain an optimized operation plan, including: When the effect score is less than the preset third threshold, the candidate solutions are generated by combining the non-core interests in the interest graph through the preset strategy generation model; According to the interest graph, activities in the initial operation plan and the candidate plan are sorted to obtain an activity sequence; Based on the activity sequence, the activity time is optimized through a preset operation plan optimization model to obtain an optimized operation plan.

10. A store operation intelligent optimization system, characterized in that: A method for realizing intelligent optimization of store operations as claimed in any one of claims 1 to 9, comprising: Data acquisition module, which obtains the behavior data of users in the store; A user grading module, which analyzes the user's cognitive level through a preset dynamic user stratification model according to the behavior data to obtain multiple user levels; An interest graph construction module analyzes the user's interest tags based on the behavior data, analyzes the interest frequency of each user level through a preset user interest graph construction model, and constructs an interest graph corresponding to each user level, wherein the user level in the interest graph is a root node, the interest tag is a child node, and the corresponding root node and child node are connected through the interest frequency; An operation plan generation module selects core interests from the interest graph, maps them in a preset activity template library through a preset operation plan generation model, and generates a corresponding initial operation plan; An operation plan evaluation module evaluates the effect of the initial operation plan through a preset plan effect evaluation model to obtain the plan effect; The operation plan optimization module optimizes the activity sequence and activity time of the initial operation plan according to the effect of the plan through a preset operation plan optimization model and combines the non-core interests in the interest graph to obtain an optimized operation plan to achieve intelligent optimization of store operations.

Images