Customer visit management method and device based on intelligent interaction

Abstract

The application provides a kind of customer visit management method and device based on intelligent interaction, it is related to computer technology field, method includes: obtaining user itinerary information;According to user itinerary information, generate visit list in combination with preset visit condition;According to visit list and user itinerary information, determine target visit object, and obtain the geographical position of corresponding target visit object, carry out visit path planning, obtain visit plan;According to visit plan, obtain the historical visit record of corresponding target visit object and target visit object information, generate the visit strategy of corresponding target visit object.The application systematically improves the whole-link visit efficiency from determining visit list to planning path, generating visit plan and visit strategy.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a customer visit management method and apparatus based on intelligent interaction. Background Technology

[0002] With the rapid development of information technology, enterprise customer relationship management (CRM) systems have become key tools for improving sales efficiency and optimizing customer service. Visit management, as an important component of CRM systems, aims to help companies standardize the customer visit process for sales personnel and gain real-time insights into frontline market dynamics and customer feedback. Early visit management relied heavily on paper records and manual summarization, but has gradually transitioned to electronic and mobile methods. Mobile devices enable the initial digital recording and uploading of visit information, providing enterprise managers with a more convenient way to access data.

[0003] Currently, in the field of enterprise customer visit management, visit routes are mainly planned manually, and electronic forms for visit content are filled out and submitted manually. Some systems may integrate simple location check-in functions to help confirm visit locations, but the core collection and summary generation of visit content still heavily relies on manual operation.

[0004] However, the above methods require frequent switching between different systems or functional modules to complete the entire process from route planning to form filling. This is cumbersome and can easily lead to low efficiency in the overall visit process. Furthermore, the visit content is highly dependent on manual input by sales staff, which is not only time-consuming and labor-intensive, but also carries the risk of data distortion due to subjective description bias or deliberate tampering. To ensure data accuracy, companies also need to invest additional manpower for on-site inspections and verifications, further increasing management costs and potentially affecting the scientific validity and accuracy of decisions made based on this data. Summary of the Invention

[0005] This invention provides a customer visit management method and apparatus based on intelligent interaction, which solves the shortcomings of the existing technology that relies on manual form submission, resulting in low efficiency of the visit process and difficulty in ensuring data authenticity. It systematically improves the efficiency of the entire visit process, from determining the visit list to planning the route, generating the visit plan and visit strategy.

[0006] This invention provides a customer visit management method based on intelligent interaction, comprising: acquiring user travel information; generating a visit list based on the user travel information and preset visit conditions; determining the target visit object based on the visit list and user travel information, acquiring the geographical location of the corresponding target visit object, planning the visit route, and obtaining a visit plan; and acquiring the historical visit records and target visit object information of the corresponding target visit object based on the visit plan, and generating a visit strategy for the corresponding target visit object.

[0007] According to the present invention, a customer visit management method based on intelligent interaction generates a visit list based on user travel information and preset visit conditions. The method includes: filtering from a customer database to obtain candidate visit targets based on preset visit conditions; the customer database is constructed prior to all customer data; inputting the candidate visit targets and user travel information into a visit target prediction model to extract features from the input candidate visit targets and user travel information and perform lexicalization on them respectively, obtaining object lexicals corresponding to the candidate visit targets and travel lexicals corresponding to the user travel information; for each travel lexical, using a self-attention mechanism to determine the attention score with each object lexical, and predicting the visit targets in the user's travel based on the corresponding attention scores to generate a visit list.

[0008] According to the present invention, a customer visit management method based on intelligent interaction is provided, which filters from a customer database according to preset visit conditions to obtain candidate visit targets. The method includes: parsing the preset visit conditions to obtain query conditions and priority rules; concatenating the query conditions into a structured query language (SQL) query statement and executing it to search the customer database to obtain candidate customers; comprehensively evaluating each candidate customer according to the candidate customers and the priority rules to obtain the corresponding candidate customer score; and selecting a target number of candidate customers from highest to lowest score according to the candidate customer score to obtain candidate visit targets.

[0009] According to the present invention, a customer visit management method based on intelligent interaction is provided. Based on a visit list and user itinerary information, target visit objects are determined, and their geographical locations are obtained. Visit path planning is then performed to obtain a visit plan. The method includes: determining target visit objects based on a visit list and obtaining their geographical locations; clustering target visit objects using a K-means clustering algorithm, combined with a preset cluster size constraint, based on the geographical locations of each target visit object and user itinerary information, to obtain clustering results; determining the inter-cluster distances based on the clustering results, and planning the visit order of clusters based on these inter-cluster distances, to obtain inter-cluster path planning results; performing intra-cluster path planning for each cluster based on the geographical locations of the target visit objects within that cluster, to obtain intra-cluster path planning results for each cluster; and determining the visit time for each visit object based on the intra-cluster path planning results, combined with the inter-cluster path planning results and the preset visit duration for each visit object, to obtain a visit plan.

[0010] According to the present invention, a customer visit management method based on intelligent interaction is provided. Based on clustering results, intra-cluster path planning is performed for each cluster according to the geographical location of each target visitor within the corresponding cluster, resulting in intra-cluster path planning results for each cluster. This includes: determining the distance between any two target visitors based on the clustering results and the geographical location of each target visitor within the corresponding cluster, thus obtaining object distance information; and selecting a starting point and constructing a dynamic planning table when the number of target visitors within the corresponding cluster is less than or equal to a preset threshold, based on the inter-cluster path planning results. The rows of the dynamic planning table represent the access status of each target visitor within the cluster, and the columns represent the currently selected target visitor. For each path in the dynamic programming table, the current node iteratively searches for unvisited nodes and selects the unvisited node with the shortest distance as the optimal predecessor node of the current node. The path in the dynamic programming table is updated, and the optimal predecessor node of each path is used as the current node. This step is repeated until a path containing all target objects within the corresponding cluster is obtained. For each path finally obtained in the dynamic programming table, the distance between adjacent path nodes and the distance between the starting and ending points of the path are determined based on the object distance information, or based on the object distance information, the geographical location of each target object within the cluster, and the preset residential location. The total path length is obtained, and the path with the shortest total path length is selected as the intra-cluster path planning result for the corresponding cluster.

[0011] According to the present invention, a customer visit management method based on intelligent interaction is provided. Based on the geographical location and user travel information of each target visitor, a K-means clustering algorithm is used, combined with a preset cluster size constraint, to cluster each target visitor to obtain a clustering result. The method includes: determining the number of clusters based on user travel information, and using the K-means clustering algorithm to obtain initial clustering results based on the geographical location of each target visitor; for the initial clustering results, detecting the number of target visitors within each initial cluster, and if the number exceeds the preset cluster size constraint, migrating and adjusting the objects within the corresponding initial cluster in order of distance from the cluster center, until all initial clusters meet the preset cluster size constraint, thus obtaining a clustering result; or, determining the number of clusters based on user travel information and initializing cluster centers; determining the distance between each target visitor and each cluster center based on the geographical location of each target visitor, and, if the preset cluster size constraint is met, assigning each target visitor to the corresponding cluster based on the distance, and updating the cluster members; re-determining the cluster centers based on the updated geographical location of the cluster members, and repeating the above steps until the iteration termination condition is met, thus obtaining a clustering result.

[0012] According to the present invention, a customer visit management method based on intelligent interaction includes: for each cluster, performing intra-cluster path planning based on the geographical location of each target visitor within the corresponding cluster, based on the clustering results, to obtain the intra-cluster path planning results for each cluster; further comprising: determining the distance between any two target visitors based on the clustering results and the geographical location of each target visitor within the corresponding cluster, to obtain object distance information; when the number of target visitors within the corresponding cluster exceeds a preset threshold, constructing a minimum spanning tree (MST) with the target visitors within the cluster as nodes, using the object distance information and the geographical location of each target visitor within the corresponding cluster; traversing the MST and calculating the total weight of each node; and so on. The degree of each node is determined, and all nodes with odd degrees are selected to form a set of odd-degree nodes. The minimum weight perfect matching algorithm is then used to find the matching method with the fewest edges and the minimum total weight of the edges in the odd-degree node set, connecting each odd-degree node to a matching edge. Based on the matching edges and the minimum weight of the MST (Minimum Weight Perfect Match), an Eulerian graph is formed, and the Eulerian circuit of the Eulerian graph is converted into a Traveling Salesman Problem (TSP) path to obtain the initial visiting order. Based on the initial visiting order, multiple neighborhood candidate paths are generated, and the total cost of each neighborhood candidate path is determined. Based on the total cost of each neighborhood candidate path, the neighborhood candidate path with the lowest total cost is selected as the optimal path for the current iteration step, and this iteration is repeated until a preset time threshold is reached. The final optimal path is then used as the intra-cluster path planning result for the corresponding cluster.

[0013] According to the present invention, a customer visit management method based on intelligent interaction is provided. Historical visit records include structured and unstructured data. Based on the visit plan, the method acquires the historical visit records and target visit information of the corresponding target visit objects, and generates a visit strategy for the corresponding target visit objects. This includes: acquiring the planned routes for each day of the user's trip according to the visit plan, obtaining the corresponding target visit objects for each day, and acquiring the historical visit records and target visit information of the corresponding target visit objects; for each day, extracting basic statistical features from the corresponding structured data based on the historical visit records of the corresponding target visit objects, segmenting the corresponding unstructured data, and performing word segmentation on each segment; determining the term frequency-inverse document frequency (TF-IDF) of each word segment to obtain text vectors, and using a sentiment analysis model to predict the sentiment score of the corresponding segment content, thus determining the sentiment score of the corresponding structured data; and based on the historical visit records of all target visit objects on the same day, using a latent Dirichlet allocation model to cluster the corresponding historical visit records by topic, and determining which historical visit records of each target visit object belong to which category? The process involves: determining the probability distribution of each topic and the visit frequency of each target audience; determining the proportion of positive, neutral, and negative reports about each target audience within a historical target timeframe based on their target audience information, and establishing a recent evaluation index; obtaining a comprehensive activity score for each target audience using principal component analysis or weighted summation based on their visit frequency and recent evaluation index; stratifying or clustering each target audience into different groups based on their basic statistical characteristics, sentiment scores, topic probability distributions, and comprehensive activity scores, resulting in corresponding user segmentation labels; these user segmentation labels characterize the user's value risk; predicting the intent of each target audience based on their basic statistical characteristics, sentiment scores, topic probability distributions, and comprehensive activity scores, and determining their potential behavior within a future target timeframe; and generating a visit strategy for each target audience based on their user segmentation labels and potential behavior, according to a pre-defined strategy template.

[0014] According to the customer visit management method based on intelligent interaction provided by the present invention, after generating a visit strategy for the corresponding target visit object, the method further includes: receiving a visit recording uploaded by a user; wherein the visit recording is obtained based on the visit strategy for the corresponding target visit object in the execution of the visit plan; performing speech recognition on the visit recording and generating visit summary content; extracting corresponding fields from the visit summary content according to preset fields to be entered and a first preset prompt word template, and entering the corresponding fields into a previously stored visit record to obtain a visit summary; wherein the visit record is generated in advance based on customer visit information input by the user, combined with preset fields to be entered and preset prompt word templates, and extracting fields corresponding to the user's visit information.

[0015] The present invention also provides a customer visit management device based on intelligent interaction, comprising: an information acquisition module for acquiring user travel information; a list generation module for generating a visit list based on the user travel information and preset visit conditions; a visit planning module for determining the target visit object based on the visit list and user travel information, acquiring the geographical location of the corresponding target visit object, planning the visit route, and obtaining a visit plan; and a strategy generation module for acquiring historical visit records and target visit object information of the corresponding target visit object based on the visit plan, and generating a visit strategy for the corresponding target visit object.

[0016] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the customer visit management method based on intelligent interaction as described above.

[0017] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent interactive customer visit management method as described above.

[0018] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the customer visit management method based on intelligent interaction as described above.

[0019] The present invention provides a customer visit management method and apparatus based on intelligent interaction. By acquiring user travel information and combining it with preset visit conditions, a visit list is generated to ensure that each object on the visit list meets core business objectives. This enables precise focusing from a large number of customers to high-value candidates. Based on the visit list and user travel information, visit routes are planned to proactively avoid time conflicts during visit planning, ensuring that the generated visit plan is feasible in reality. This avoids arranging visits that users cannot attend, greatly improving the practicality of the plan and user trust. Furthermore, based on the visit plan, corresponding historical visit records and target visit object information are integrated to generate highly customized communication strategies, making each visit more targeted and in-depth. This greatly promotes the establishment and maintenance of long-term customer relationships and systematically improves the efficiency of the entire visit chain, from determining the visit list to planning routes, generating visit plans, and visit strategies. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is one of the flowcharts of the customer visit management method based on intelligent interaction provided by the present invention; Figure 2 This is the second flowchart of the customer visit management method based on intelligent interaction provided by the present invention; Figure 3 This is a schematic diagram of the structure of the customer visit management device based on intelligent interaction provided by the present invention; Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0023] Figure 1 This is a flowchart illustrating the customer visit management method based on intelligent interaction provided by the present invention, as shown below. Figure 1 As shown, the method includes: S11, obtain user trip information; S12, Generate a visit list based on user travel information and preset visit conditions; S13. Based on the visit list and user itinerary information, determine the target visit object, obtain the geographical location of the corresponding target visit object, plan the visit route, and obtain the visit plan; S14. Based on the visit plan, obtain the historical visit records and target visit information of the corresponding target visit object, and generate the visit strategy for the corresponding target visit object.

[0024] It should be noted that the following will be combined with specific details. Figure 2 This invention describes a customer visit management method based on intelligent interaction.

[0025] Step S11: Obtain user travel information.

[0026] It should be added that the user's travel information includes the number of days away from home, the number of customers planned to be visited each day, the total number of customer visits expected, and the expected duration of each visit. These can be configured according to actual needs, and no further restrictions are imposed here.

[0027] Step S12: Generate a visit list based on the user's itinerary information and preset visit conditions.

[0028] In this embodiment, a visit list is generated based on user itinerary information and preset visit conditions. This includes: filtering from a customer database to obtain candidate visit targets based on preset visit conditions; wherein the customer database is constructed based on all customer data; inputting the candidate visit targets and user itinerary information into a visit target prediction model to extract features from the input candidate visit targets and user itinerary information and perform lexicalization on them respectively to obtain object lexicals corresponding to the candidate visit targets and itinerary lexicals corresponding to the user itinerary information; for each itinerary lexical, using a self-attention mechanism to determine the attention score with each object lexical, and predicting the visit targets of the user's itinerary based on the corresponding attention score to generate a visit list.

[0029] It should be added that by using preset visit criteria and filtering rules, the visit targets are screened to ensure that the selected candidate visit targets meet the core business objectives. Based on this, the selected candidate visit targets and user itinerary information are input into the visit target prediction model. The model is used to capture the deep semantics of the information to select the customers that best match the user's itinerary and potential intentions, ensuring that a high-quality visit list is generated that conforms to business rules and fits the user's dynamic intentions.

[0030] To elaborate further, based on preset visit conditions, candidate visit targets are obtained by filtering from the customer database. This includes: parsing the preset visit conditions to obtain query conditions and priority rules; concatenating the query conditions into a structured query language (SQL) query statement and executing it to search the customer database and obtain candidate customers; comprehensively evaluating each candidate customer based on the candidate customers and the priority rules to obtain the corresponding candidate customer score; and selecting a target number of candidate customers from highest to lowest score based on the score of each candidate customer to obtain the candidate visit targets.

[0031] It should be added that the filtering conditions can be configured with corresponding filtering fields according to business rules, such as high-value customers, customers who have not been visited for 3 months, customers located in a certain region, etc. The priority rules can be configured according to the customer's risk value and / or specific business rules, such as high-value customers having higher priority, etc. No further restrictions are made here.

[0032] Furthermore, based on the candidate customers and in conjunction with priority rules, a comprehensive evaluation is conducted on each candidate customer to obtain the corresponding score results. This includes: extracting the customer's static attributes and dynamic behaviors based on the candidate customers; whereby, the customer's static attributes include customer level, industry size, and years of cooperation, and the customer's dynamic behaviors include recent order amount, website activity, number of service requests, and number of times marketing emails are opened; based on the customer's static attributes and dynamic behaviors, and in conjunction with priority rules, determining the priority rules for matching candidate customers, and combining the rule weights previously assigned based on priority rules, obtaining the corresponding score results for the candidate customers.

[0033] Step S13: Based on the visit list and user itinerary information, determine the target visit object, obtain the geographical location of the corresponding target visit object, plan the visit route, and obtain the visit plan.

[0034] In this embodiment, based on the visit list and user itinerary information, target visit objects are determined, and the geographical locations of the corresponding target visit objects are obtained. Visit path planning is then performed to obtain a visit plan, including: determining target visit objects based on the visit list and obtaining the geographical locations of the corresponding target visit objects; clustering the target visit objects using the K-means clustering algorithm, combined with a preset cluster size constraint, based on the geographical locations of each target visit object and user itinerary information, to obtain clustering results; determining the inter-cluster distances between clusters based on the clustering results, and planning the visit order of clusters based on the inter-cluster distances, to obtain inter-cluster path planning results; performing intra-cluster path planning for each cluster based on the geographical locations of the target visit objects within that cluster, to obtain intra-cluster path planning results for each cluster; and determining the visit time for each visit object based on the intra-cluster path planning results, combined with the inter-cluster path planning results and the preset visit duration for each visit object, to obtain a visit plan.

[0035] Specifically, based on the clustering results, intra-cluster path planning is performed for each cluster according to the geographical location of each target visited object within the corresponding cluster, resulting in intra-cluster path planning results for each cluster. This includes: based on the clustering results, determining the distance between any two target visited objects according to the geographical location of each target visited object within the corresponding cluster, thus obtaining object distance information; based on the inter-cluster path planning results, when the number of target visited objects within the corresponding cluster is less than or equal to a preset threshold, selecting a starting point and constructing a dynamic planning table; wherein, the rows of the dynamic planning table represent the access status of each target visited object within the cluster, and the columns of the dynamic planning table represent the currently selected target visited object; for each path in the dynamic planning table... For the current node, iteratively search for unvisited nodes and select the unvisited node with the shortest distance as the optimal predecessor node of the current node. Update the path in the dynamic programming table and use the optimal predecessor node corresponding to each path as the current node. Repeat this step until a path containing all target visited objects within the corresponding cluster is obtained. For each path finally obtained in the dynamic programming table, determine the distance between adjacent path nodes and the distance between the start and end points of the path based on the object distance information, or based on the object distance information, the geographical location of each target visited object within the cluster, and the preset residential location. Obtain the total path length and select the path with the shortest total path length as the intra-cluster path planning result for the corresponding cluster.

[0036] It should be noted that the target visit object can be any object listed in the visit list, or a specific object selected from the visit list. The choice can be made according to actual design requirements, and no further limitations are made here. In addition, selecting the starting point includes: selecting any target visit object within the cluster as the starting point, or selecting a preset residential location as the starting point, and using other target visit objects within the cluster as the next path nodes of the starting point.

[0037] Furthermore, by identifying target objects and obtaining their geographical locations, the visit list is associated with specific geographical locations, ensuring that subsequent path planning has clear and calculable targets. K-means clustering, combined with a preset cluster size constraint, is used to cluster target objects, ensuring a controllable and balanced number of visits per cluster while reducing the computational complexity of subsequent path planning. The number of target objects within each cluster is compared to a preset threshold. For large-scale clusters where the number of target objects exceeds the threshold, an approximate solution strategy is adopted, selecting points near the cluster center as virtual starting points to quickly generate a reasonable initial visit order, avoiding blind searching in the large solution space. Neighborhood operations and local searches are performed to determine the final planned path, effectively eliminating unreasonable aspects such as path intersections and balancing efficiency and quality. For small-scale clusters where the number is less than or equal to the threshold, an optimal solution strategy is adopted for precise solutions. A dynamic programming table is used to access the shortest path of all nodes to determine the path endpoint, and backtracking from the path endpoint determines the final planned link level, ensuring that within a small scope, the path planning result is mathematically optimal and of the highest quality.

[0038] Furthermore, neighborhood operations include node swapping, node relocation, and node exchange. Node swapping breaks the connection between two non-adjacent road segments in the initial visit sequence and exchanges nodes from one segment with nodes from another. Node relocation extracts a continuous sequence of nodes from the initial visit sequence and inserts it into another position on the path. Node exchange swaps the positions of any two non-adjacent intersection points in the initial visit sequence. Specific neighborhood operations can be set according to actual design requirements and are not further limited here. Additionally, nodes represent specific target access objects in the initial visit sequence.

[0039] In one optional embodiment, based on the geographical location and user travel information of each target visited object, the K-means clustering algorithm is used in conjunction with a preset cluster size constraint to cluster each target visited object to obtain a clustering result. This includes: determining the number of clusters based on the user travel information, and using the K-means clustering algorithm to obtain an initial clustering result based on the geographical location of each target visited object; for the initial clustering result, detecting the number of target visited objects in each initial cluster, and if the number exceeds the preset cluster size constraint, migrating and adjusting the objects in the corresponding initial cluster in order of distance from the cluster center, until all initial clusters meet the preset cluster size constraint, thus obtaining the clustering result.

[0040] It should be noted that the number of clusters is determined based on user travel information to correlate the number of clusters with user travel information. This ensures that the clustering results directly serve the user's actual work arrangements, avoiding the blindness of the algorithm and ensuring that each cluster after decomposition is reasonable and feasible from a macroscopic perspective. Furthermore, the K-means algorithm is used for initial clustering to quickly group geographically close customer points into one category. In addition, by introducing a preset cluster size constraint, the target visit objects within the cluster are moved and adjusted so that the target visit objects that are moved out are the ones farthest from the center in the current cluster, while the new cluster they are assigned to is the second-best choice. This is to maintain the geographical compactness of the clusters to the maximum extent while satisfying the constraints, minimizing the additional path costs caused by the adjustment, and ensuring that the number of customers in each cluster is within an acceptable range. This adapts to different customer distributions and always outputs a balanced result that meets business requirements.

[0041] In addition, the objects in the corresponding initial clusters are migrated and adjusted in order of distance from the cluster center, including: selecting target visited objects in the corresponding initial clusters that exceed the number of the preset cluster size constraint in order of distance from the cluster center, and moving them to the nearest cluster whose number does not exceed the preset cluster size constraint.

[0042] In another optional embodiment, based on the geographical location and user travel information of each target visitor, the K-means clustering algorithm is used, combined with a preset cluster size constraint, to cluster each target visitor to obtain a clustering result. The method further includes: determining the number of clusters and initializing cluster centers based on the user travel information; determining the distance between each target visitor and each cluster center based on their geographical location, and, while satisfying the preset cluster size constraint, assigning each target visitor to the corresponding cluster based on the distance, and updating the cluster members; re-determining the cluster centers based on the updated cluster members' geographical locations, and repeating the above steps until the iteration termination condition is met to obtain the clustering result.

[0043] It should be noted that each target visited object is assigned to a corresponding cluster based on distance, including: Based on the geographical location of each of the remaining target visitors, the distance between each of the remaining target visitors and each cluster center is determined, and the cluster with the closest distance is taken as the cluster to be assigned to the corresponding remaining target visitors. It is checked whether the cluster to be assigned exceeds the preset cluster size constraint. If the cluster to be assigned does not exceed the preset cluster size constraint, the corresponding remaining target visitors are assigned to the cluster to be assigned; otherwise, the cluster with the next closest distance is taken as the cluster to be assigned to the corresponding remaining target visitors.

[0044] It should be noted that by randomly selecting cluster centers and assigning the remaining target objects to their corresponding clusters, and updating the cluster centers based on the updated clusters, the system iteratively optimizes the process to make the points within each cluster increasingly compact and the clusters increasingly separated. Combined with a maximum number of iterations, this ensures a stable result is obtained within a finite timeframe under any circumstances, avoiding infinite loops and enhancing the system's robustness. Furthermore, by determining the distance between the remaining target objects and the cluster centers, clusters are assigned to each remaining target object. It is then checked whether the assigned cluster exceeds a preset cluster size constraint, ensuring that each target object is successfully and appropriately assigned, thus preventing task overload.

[0045] Furthermore, based on the geographical locations of all target visited objects within each updated cluster, the cluster center of the corresponding updated cluster is updated, including: for each updated cluster, determining the geometric center based on the geographical locations of all target visited objects within the corresponding updated cluster, and updating the cluster center of the corresponding updated cluster based on the geometric center.

[0046] In addition, based on the clustering results, intra-cluster path planning is performed for each cluster according to the geographical location of each target visited object within the corresponding cluster, resulting in intra-cluster path planning results for each cluster. This also includes: based on the clustering results, determining the distance between any two target visited objects for each cluster according to the geographical location of each target visited object within the corresponding cluster, obtaining object distance information; based on the inter-cluster path planning results, when the number of target visited objects within a corresponding cluster exceeds a preset threshold, constructing a minimum spanning tree (MST) with the minimum total weight of edges connecting all nodes within the cluster, using the object distance information and the geographical location of each target visited object within the corresponding cluster as nodes; traversing the MST, counting the degree of each node, and selecting the nodes with the minimum total weight. Given nodes with odd degrees, we obtain a set of odd-degree nodes. Using the minimum weight perfect matching algorithm, we find the matching method with the fewest edges and the minimum total weight of the edges in the odd-degree node set, connecting each odd-degree node to a matching edge. Based on the matching edges and the minimum weight of the perfect matching (MST), we form an Eulerian graph and convert the Eulerian circuit of the Eulerian graph into a Traveling Salesman Problem (TSP) path to obtain the initial visiting order. Based on the initial visiting order, we generate multiple neighborhood candidate paths and determine the total cost of each neighborhood candidate path. Based on the total cost of each neighborhood candidate path, we select the neighborhood candidate path with the lowest total cost as the optimal path for the current iteration step and repeat the iteration until a preset time threshold is reached. The final optimal path is then used as the intra-cluster path planning result for the corresponding cluster.

[0047] It should be added that the Eulerian circuit is used to represent a closed-loop path that traverses each edge once and returns to the starting point; the Eulerian circuit of the Eulerian graph is converted into a Traveling Salesman Problem (TSP) path to obtain the initial visiting order, which includes: traversing along the Eulerian circuit, and when encountering a node that has been visited repeatedly, skipping the repeated node until the next unvisited node is encountered, thereby eliminating the repeated nodes in the path and obtaining an initial visiting order that contains all nodes without repetition, which is used as the initial visiting order.

[0048] In addition, the total cost of each neighborhood candidate path is determined, including: for each neighborhood candidate path, the path penalty cost is determined based on the corresponding neighborhood candidate path and the penalty coefficient of each edge of the path, and the corresponding total cost is determined based on the actual distance of the path, so as to fully consider the negative impact of high penalty edges when evaluating costs.

[0049] It is worth noting that the penalty coefficient of each edge needs to be dynamically adjusted according to the preset penalty coefficient adjustment rules. Specifically, for edges that appear more frequently than the first preset frequency threshold in a consecutive preset number of iterations, when the edge is determined to be high-cost, its penalty coefficient is increased by the first preset amount to guide the algorithm to avoid such edges in subsequent searches in order to explore path combinations that have not been fully explored; for edges that appear less frequently than the second preset frequency threshold in a consecutive preset number of iterations, its penalty coefficient is decreased by the second preset amount to avoid excessive penalty leading to path performance deterioration. The second preset frequency threshold is less than the first preset frequency threshold.

[0050] Furthermore, the first preset frequency threshold, the second preset frequency threshold, the first preset amplitude, and the second preset amplitude can be designed according to actual design requirements, and are not further limited here.

[0051] Step S14: Based on the visit plan, obtain the historical visit records and target visit information of the corresponding target visit object, and generate the visit strategy for the corresponding target visit object.

[0052] Specifically, historical visit records include structured and unstructured data. Based on the visit plan, historical visit records and target visit information for the corresponding target audience are obtained, and a visit strategy for the target audience is generated. This includes: obtaining the planned routes for each day of the user's trip according to the visit plan, identifying the corresponding target audience for each day, and obtaining the historical visit records and target visit information for the corresponding target audience; for each day, based on the historical visit records of the corresponding target audience, extracting basic statistical features from the corresponding structured data, segmenting the corresponding unstructured data, and performing word segmentation on each segment; determining the term frequency-inverse document frequency (TF-IDF) of each word segment to obtain text vectors to capture frequently mentioned topics by customers, such as price, delivery, and new features, and using a sentiment analysis model to predict the sentiment score of the corresponding segment content, thus determining the sentiment score of the corresponding structured data; based on the historical visit records of all target audiences on the same day, using a latent Dirichlet allocation model, clustering the corresponding historical visit records by topic, determining the probability distribution of each target audience's historical visit records belonging to each topic, and identifying each target audience. The process involves: determining the frequency of visits; identifying the proportion of positive, neutral, and negative reports about each target visitor within a historical target timeframe based on their target visitor information, and establishing a recent evaluation index; using principal component analysis or weighted summation to obtain a comprehensive activity score for each target visitor based on their visit frequency and recent evaluation index, thus measuring the overall value and activity of the customer; stratifying or clustering each target visitor based on their basic statistical characteristics, sentiment scores, topic probability distributions, and comprehensive activity scores to divide them into different groups, resulting in corresponding user segmentation labels; these user segmentation labels characterize the value risk of users; predicting the intent of each target visitor based on their basic statistical characteristics, sentiment scores, topic probability distributions, and comprehensive activity scores to determine their potential behavior within the future target timeframe; and generating visit strategies for each target visitor based on their user segmentation labels and potential behavior, according to a preset strategy template, enabling differentiated visit strategies for different customers and improving flexibility and scalability.

[0053] It should be added that structured data includes visit frequency, visit duration, visit participants, last visit date, transaction records, purchased products / services, contract amounts, complaint records, service request forms, etc., while unstructured data includes historical visit minutes (text), transcripts of meeting recordings, email correspondence, chat logs, etc. Additionally, target visit information includes corporate client information and individual client information, which can be obtained from internet websites. Corporate client information includes the corresponding target's official website news (new product launches, strategic partnerships, financing updates), financial reports, executive changes, industry analysis reports, social media reviews, recruitment information (indicating business expansion directions), competitor activities, etc. Individual client information includes public social media activity, interest tags, publicly available professional information, and activities participated in.

[0054] In an optional embodiment, before extracting the basic statistical features from the corresponding structured data based on the historical visit records of the corresponding target visited object for each day, the method includes: extracting entities from the unstructured data using a named entity recognition algorithm, and associating the corresponding unstructured data with the structured data using entity linking technology.

[0055] Furthermore, based on unstructured data, entities are extracted using named entity recognition algorithms, including: using named entity recognition algorithms to extract key entities such as company names, person names, and product names from unstructured data.

[0056] In addition, after extracting entities from unstructured data using named entity recognition algorithms, the process includes: deduplicating, correcting errors, and / or filling missing values ​​in the identified entities; where error correction includes standardizing date formats, currency units, etc., and filling missing values ​​includes using the mean, median, or a predictive model; the resulting data is stored in a data warehouse or data lake to form a panoramic view with the customer as the unique identifier, which does not contain multidimensional information.

[0057] In one alternative embodiment, reference Figure 2 After generating the visit strategy for the corresponding target visitee, the process also includes: receiving visit recordings uploaded by users; wherein the visit recordings are obtained based on the visit strategy for the corresponding target visitee in the visit plan; performing speech recognition on the visit recordings and generating visit summary content; extracting corresponding fields from the visit summary content according to preset fields to be entered and a first preset prompt word template, and entering the corresponding fields into the corresponding visit record to obtain the visit summary; wherein the visit record is generated by first extracting fields corresponding to the user's visit information based on the customer visit information input by the user, combined with preset fields to be entered and a second preset prompt word template.

[0058] It should be noted that, based on the preset fields to be entered and the first preset prompt word template, the corresponding fields are extracted from the visit summary content and entered into the corresponding visit record. This includes: inputting the visit summary content, the preset fields to be entered and the first preset prompt word template into the large language model to obtain the visit summary output by the large language model.

[0059] In addition, the preset fields to be entered include the fields to be generated and their corresponding descriptions, so that the model can understand the fields to be identified and the content to be summarized. The first preset prompt word template and the second preset prompt word template can be set according to actual design requirements, so that the unique business logic of the enterprise or the dynamically changing market demand can be adjusted and customized in real time, improving the flexibility of the visit summary generation mechanism, realizing personalized customization, avoiding the cost loss caused by repeated model training, ensuring that the actual needs of the enterprise to respond quickly to the market are met, and at the same time, the visit summary is automatically generated through the large language model to retain the original data and editing records, avoiding information confusion or tampering.

[0060] Furthermore, a human-computer dialogue interaction system based on the full-visit process agent has been built, enabling users to complete the entire visit process through natural language dialogue interaction without switching between multiple systems, significantly simplifying the process, realizing intelligent management of the entire customer visit process, and improving enterprise operational efficiency and market competitiveness.

[0061] In addition, before receiving the visit recordings uploaded by the user, the process includes: receiving customer visit information input by the user; wherein the customer visit information is obtained based on the target visit object selected in the visit plan; based on the customer visit information, combined with the preset field information to be entered and the second preset prompt word template, extracting the fields of the corresponding user visit information, generating the corresponding visit record and storing it.

[0062] In summary, this invention, by acquiring user travel information and combining it with preset visit conditions, generates a visit list to ensure that each object on the visit list aligns with core business objectives. This enables precise focusing from a vast pool of customers to high-value candidates. Furthermore, based on the visit list and user travel information, visit routes are planned to proactively avoid time conflicts during visit planning, ensuring the generated visit plan is feasible in reality. This avoids scheduling visits that users cannot attend, significantly enhancing the plan's practicality and user trust. Additionally, based on the visit plan, corresponding historical visit records and target visit object information are integrated to generate highly customized communication strategies, making each visit more targeted and in-depth. This greatly promotes the establishment and maintenance of long-term customer relationships, systematically improving the efficiency of the entire visit process from determining the visit list to planning routes, generating visit plans, and implementing visit strategies.

[0063] The intelligent interactive customer visit management device provided by the present invention is described below. The intelligent interactive customer visit management device described below and the intelligent interactive customer visit management method described above can be referred to in correspondence.

[0064] Figure 3 A schematic diagram of a customer visit management device based on intelligent interaction is shown. The device includes: Information acquisition module 31 acquires user travel information; The list generation module 32 generates a visit list based on the user's itinerary information and preset visit conditions; The visit planning module 33 determines the target visit object based on the visit list and user itinerary information, obtains the geographical location of the corresponding target visit object, plans the visit route, and obtains the visit plan; The strategy generation module 34, based on the visit plan, obtains the historical visit records and target visit information of the corresponding target visit object, and generates the visit strategy for the corresponding target visit object.

[0065] In this embodiment, the list generation module 32 includes: a filtering unit, which filters from a customer database according to preset visit conditions to obtain candidate visit objects; wherein the customer database is constructed based on all customer data; a visit object prediction unit, which inputs the candidate visit objects and user itinerary information into a visit object prediction model to extract features from the input candidate visit objects and user itinerary information and perform lexicalization on them respectively to obtain object lexicals corresponding to the candidate visit objects and itinerary lexicals corresponding to the user itinerary information; for each itinerary lexical, a self-attention mechanism is used to determine the attention score with each object lexical, and based on the corresponding attention score, the visit objects of the user's itinerary are predicted to generate a visit list.

[0066] Furthermore, the filtering unit includes: a parsing subunit, which parses the preset visit conditions to obtain query conditions and priority rules; a statement generation and execution subunit, which concatenates the query conditions into a Structured Query Language (SQL) query statement and executes it to search the customer database and obtain candidate customers; a comprehensive evaluation subunit, which performs a comprehensive evaluation of each candidate customer based on the candidate customers and the priority rules to obtain the corresponding candidate customer's score result; and an object filtering subunit, which selects a target number of candidate customers from highest to lowest score result to obtain candidate visit objects.

[0067] Furthermore, the comprehensive evaluation sub-unit is used to: extract static attributes and dynamic behaviors of candidate customers; whereby static attributes include customer level, industry size, and years of cooperation, and dynamic behaviors include recent order amount, website activity, number of service requests, and number of times marketing emails are opened; based on the static attributes and dynamic behaviors of customers, combined with priority rules, the priority rules for matching candidate customers are determined, and combined with the rule weights previously assigned based on priority rules, the scoring results of the corresponding candidate customers are obtained.

[0068] In this embodiment, the visit planning module 33 includes: a location acquisition unit, which determines the target visit object based on the visit list and acquires the geographical location of the corresponding target visit object; a clustering unit, which clusters the target visit objects based on the geographical location of each target visit object and user travel information using the K-means clustering algorithm, combined with a preset cluster size constraint, to obtain the clustering result; an inter-cluster planning unit, which determines the inter-cluster distance between each cluster based on the clustering result and plans the visit order of the clusters based on the inter-cluster distance to obtain the inter-cluster path planning result; an object distance determination unit, which determines the distance between any two target visit objects for each cluster based on the geographical location of each target visit object within the corresponding cluster, based on the clustering result, to obtain object distance information; and an intra-cluster planning unit, which, based on the inter-cluster path planning result, determines, when the number of target visit objects within the corresponding cluster is less than or equal to a preset number threshold, any target visit object within the cluster as the starting point, or a preset residential location as the starting point, and other target visit objects within the cluster as the next path nodes of the starting point, to construct a dynamic planning table; wherein, the rows of the dynamic planning table represent the visit status of each target visit object within the cluster, and the columns of the dynamic planning table represent the currently selected target visit... The process involves several steps: For each path in the dynamic programming table, target objects not included in the corresponding path are selected as candidate nodes for the next node of that path. For each candidate node, the distance between it and its preceding path nodes is determined based on object distance information. The preceding path node with the smallest distance is selected as the optimal predecessor node of the candidate node. The paths in the dynamic programming table are updated based on each candidate node and its optimal predecessor node. This process is repeated iteratively until a path containing all target objects within the corresponding cluster is obtained. For each path finally obtained in the dynamic programming table, the distance between adjacent path nodes and the distance between the start and end points of the path are determined based on object distance information, or based on object distance information, the geographical location of each target object within the cluster, and the preset residential location. The total path length is obtained, and the path with the smallest total path length is selected as the intra-cluster path planning result for the corresponding cluster. The intra-cluster planning unit is also used to determine, when the number of target objects within the corresponding cluster exceeds a preset threshold, to use Christofides' algorithm based on object distance information and the geographical location of each target object within the corresponding cluster. The algorithm, combined with a guided local search strategy, determines the optimal path and obtains the intra-cluster path planning results for the corresponding cluster. Based on the inter-cluster path planning results and the intra-cluster path planning results of each cluster, and combined with the preset visit duration of each visited object, the visit time of each visited object is determined, and the visit plan is obtained.

[0069] In an optional embodiment, the clustering unit includes: a clustering subunit, which determines the number of clusters based on user travel information and, in conjunction with the geographical location of each target visited object, uses the K-means clustering algorithm to obtain multiple initial clusters; and a cluster update subunit, which checks the number of target visited objects within each initial cluster, and, based on the existence of initial clusters exceeding a preset cluster size constraint, selects target visited objects within the corresponding initial cluster that exceed the preset cluster size constraint in order of distance from the cluster center, and moves them to the nearest cluster whose number does not exceed the preset cluster size constraint, thereby obtaining the clustering result.

[0070] In another optional embodiment, the clustering unit further includes: a cluster center selection subunit, which determines the number of clusters based on user travel information, and randomly selects target visitors from each target visitor according to the number of clusters, as the cluster center of the corresponding cluster; and a clustering module, which, based on the cluster centers of each cluster and the geographical locations of the remaining target visitors, assigns clusters to each of the remaining target visitors according to a preset allocation strategy, obtaining updated clusters, and updates the cluster centers of the corresponding updated clusters based on the geographical locations of all target visitors within each updated cluster, and iterates this step again based on the updated cluster centers of each cluster until the maximum number of iterations is reached. The resulting clusters are used as the clustering results. The preset allocation strategy is used to: determine the distance of each of the remaining target visited objects from the cluster center based on their geographical locations, and take the closest cluster as the cluster to be allocated to the corresponding remaining target visited object; check whether the cluster to be allocated exceeds the preset cluster size constraint. If the cluster to be allocated does not exceed the preset cluster size constraint, then the corresponding remaining target visited object is allocated to the cluster to be allocated; otherwise, the next closest cluster is taken as the cluster to be allocated to the corresponding remaining target visited object, and this step is repeated until the corresponding remaining target visited object is allocated to the corresponding cluster to be allocated.

[0071] Furthermore, the cluster center selection sub-unit includes: for each updated cluster, determining the geometric center based on the geographical location of all target visited objects within the corresponding updated cluster, and updating the cluster center of the corresponding updated cluster based on the geometric center.

[0072] In addition, the intra-cluster planning unit is also used for: constructing a minimum spanning tree (MST) with the minimum total weight of edges connecting all nodes in the cluster, based on object distance information and the geographical location of each target visited object in the corresponding cluster, using the target visited objects in the cluster as nodes; traversing the MST, counting the degree of each node, and filtering out all nodes with odd degrees to obtain a set of odd-degree nodes, and using the minimum weight perfect matching algorithm to find the matching method with the fewest number of edges and the minimum total weight of edges in the set of odd-degree nodes, so that each odd-degree node is connected to a matching edge; forming an Eulerian graph based on the matching edges and the MST, and converting the Eulerian circuit of the Eulerian graph into a Traveling Salesman Problem (TSP) path to obtain the initial visiting order; generating multiple neighborhood candidate paths based on the initial visiting order, and determining the total cost of each neighborhood candidate path; selecting the neighborhood candidate path with the lowest total cost as the optimal path for the current iteration step based on the total cost of each neighborhood candidate path, and repeating the iteration until a preset time threshold is reached, and taking the final optimal path as the intra-cluster path planning result of the corresponding cluster.

[0073] It should be added that the Eulerian circuit is used to represent a closed-loop path that traverses each edge once and returns to the starting point; the Eulerian circuit of the Eulerian graph is converted into a Traveling Salesman Problem (TSP) path to obtain the initial visiting order, which includes: traversing along the Eulerian circuit, and when encountering a node that has been visited repeatedly, skipping the repeated node until the next unvisited node is encountered, thereby eliminating the repeated nodes in the path and obtaining an initial visiting order that contains all nodes without repetition, which is used as the initial visiting order.

[0074] In addition, the total cost of each neighborhood candidate path is determined, including: for each neighborhood candidate path, the path penalty cost is determined based on the corresponding neighborhood candidate path and the penalty coefficient of each edge of the path, and the corresponding total cost is determined based on the actual distance of the path, so as to fully consider the negative impact of high penalty edges when evaluating costs.

[0075] It is worth noting that the penalty coefficient of each edge needs to be dynamically adjusted according to the preset penalty coefficient adjustment rules. Specifically, for edges that appear more frequently than the first preset frequency threshold in a consecutive preset number of iterations, when the edge is determined to be high-cost, its penalty coefficient is increased by the first preset amount to guide the algorithm to avoid such edges in subsequent searches in order to explore path combinations that have not been fully explored; for edges that appear less frequently than the second preset frequency threshold in a consecutive preset number of iterations, its penalty coefficient is decreased by the second preset amount to avoid excessive penalty leading to path performance deterioration. The second preset frequency threshold is less than the first preset frequency threshold.

[0076] Specifically, the strategy generation module 34 includes: an information acquisition unit, which acquires the planned routes for each day of the user's trip according to the visit plan, obtains the corresponding target visit objects for each day, and acquires the historical visit records and target visit object information for the corresponding target visit objects; a word segmentation unit, which extracts the basic statistical features from the corresponding structured data based on the historical visit records of the corresponding target visit objects for each day, segments the corresponding unstructured data, and performs word segmentation on the content of each segment; a sentiment scoring unit, which determines the term frequency-inverse document frequency (TF-IDF) of each word segment to obtain text vectors to capture topics frequently mentioned by customers, such as price, delivery, and new features, and uses a sentiment analysis model to predict the sentiment score of the corresponding segment content, and determines the sentiment score of the corresponding structured data; a probability prediction unit, which uses a latent Dirichlet allocation model to cluster the corresponding historical visit records by topic based on the historical visit records of all target visit objects on the same day, and determines the probability distribution of the historical visit records of each target visit object belonging to each topic, and determines the visit frequency of each target visit object; and an index determination unit, which determines the index based on the target visit object information of the corresponding target visit object. The system comprises several modules: a recent evaluation index, an activity score, and a strategy generation unit. The first module determines the recent evaluation index by analyzing the proportion of positive, neutral, and negative reports about the target audience within a historical target timeframe. The second module generates a visit strategy based on the visit frequency and recent evaluation index of each target audience, using principal component analysis or weighted summation to obtain a comprehensive activity score that measures the overall value and activity of the customer. The third module stratifies or clusters the target audience based on their basic statistical characteristics, sentiment scores, topic probability distributions, and comprehensive activity scores, creating corresponding user segmentation labels. These user segmentation labels characterize the user's value risk. The fourth module predicts the intent of each target audience based on their basic statistical characteristics, sentiment scores, topic probability distributions, and comprehensive activity scores, determining their potential behavior within a future target timeframe. The fifth module generates a visit strategy based on the user segmentation labels and potential behaviors of the target audience, using a pre-defined strategy template. This allows for differentiated visit strategies for different customers, improving flexibility and scalability.

[0077] In an optional embodiment, the strategy generation module 34 further includes an entity association unit, which, before extracting the basic statistical features from the corresponding structured data based on the historical visit records of the corresponding target visited object for each day, extracts entities from the unstructured data using a named entity recognition algorithm, and associates the corresponding unstructured data with the structured data using entity linking technology.

[0078] Furthermore, the entity association unit is used to extract key entities such as company names, personal names, and product names from unstructured data using named entity recognition algorithms.

[0079] In addition, the strategy generation module 34 also includes: data processing, which, after extracting entities from unstructured data using a named entity recognition algorithm, performs deduplication, error correction, and / or fills in missing values ​​for the identified entities; where error correction includes unifying date formats, currency units, etc., and filling in missing values ​​includes filling with the mean, median, or prediction model; and a customer profiling unit, which stores the obtained data in a data warehouse or data lake to form a panoramic view with the customer as the unique identifier, without containing multi-dimensional information.

[0080] In an optional embodiment, the device further includes: an audio acquisition module, which receives a visit recording uploaded by a user after generating a visit strategy for the corresponding target visitee; wherein the visit recording is obtained based on the visit strategy for the corresponding target visitee in the visit plan; a speech recognition module, which performs speech recognition on the visit recording and generates a visit summary; and a summary generation module, which extracts corresponding fields from the visit summary based on preset fields to be entered and a first preset prompt word template, and enters the corresponding fields into the corresponding visit record to obtain a visit summary; wherein the visit record is generated by first extracting fields corresponding to the user's visit information based on the customer visit information input by the user, combined with the preset fields to be entered and the second preset prompt word template.

[0081] It should be noted that the summary generation module is used to input the visit summary content, preset fields to be entered, and the first preset prompt word template into the large language model to obtain the visit summary output by the large language model.

[0082] In addition, the device also includes: an information receiving module, which receives customer visit information input by the user before receiving the visit recording uploaded by the user; wherein the customer visit information is obtained based on the selection of the target visit object in the visit plan; and a record generation module, which extracts the fields of the corresponding user visit information according to the customer visit information, combined with the preset field information to be entered and the second preset prompt word template, generates the corresponding visit record and stores it.

[0083] In summary, this embodiment of the invention acquires user travel information through an information acquisition module, and generates a visit list by combining it with preset visit conditions through a list generation module. This ensures that each object on the visit list meets core business objectives, achieving precise focusing from a large number of customers to high-value candidates. The visit planning module plans visit routes based on the visit list and user travel information, proactively avoiding time conflicts and ensuring the generated visit plan is feasible in reality. This avoids scheduling visits that users cannot attend, greatly improving the plan's practicality and user trust. Finally, the strategy generation module integrates historical visit records and target visit object information based on the visit plan to generate highly customized communication strategies, making each visit more targeted and in-depth. This significantly promotes the establishment and maintenance of long-term customer relationships, systematically improving the efficiency of the entire visit process from determining the visit list to planning routes, generating visit plans, and implementing visit strategies.

[0084] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include a processor 410, a communications interface 420, a memory 430, and a communication bus 440. The processor 410, communications interface 420, and memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a customer visit management method based on intelligent interaction. This method includes: acquiring user travel information; generating a visit list based on the user travel information and preset visit conditions; determining the target visit object based on the visit list and user travel information, acquiring the geographical location of the corresponding target visit object, planning the visit route, and obtaining a visit plan; and acquiring historical visit records and target visit object information based on the visit plan, and generating a visit strategy for the corresponding target visit object.

[0085] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0086] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the customer visit management method based on intelligent interaction provided by the above methods. The method includes: obtaining user travel information; generating a visit list based on the user travel information and preset visit conditions; determining the target visit object based on the visit list and user travel information, obtaining the geographical location of the corresponding target visit object, planning the visit route, and obtaining a visit plan; and obtaining the historical visit records and target visit object information of the corresponding target visit object based on the visit plan, and generating a visit strategy for the corresponding target visit object.

[0087] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the customer visit management method based on intelligent interaction provided by the above methods. The method includes: acquiring user travel information; generating a visit list based on the user travel information and preset visit conditions; determining the target visit object based on the visit list and user travel information, acquiring the geographical location of the corresponding target visit object, planning the visit route, and obtaining a visit plan; and acquiring historical visit records and target visit object information of the corresponding target visit object based on the visit plan, and generating a visit strategy for the corresponding target visit object.

[0088] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0089] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A customer visit management method based on intelligent interaction, characterized in that, include: Obtain user travel information; Based on the user's itinerary information and preset visit conditions, a visit list is generated; Based on the visit list and the user itinerary information, the target visit object is determined, and the geographical location of the corresponding target visit object is obtained. The visit route is planned to obtain the visit plan. Based on the visit plan, obtain the historical visit records and target visit information of the corresponding target visit object, and generate the visit strategy for the corresponding target visit object.

2. The customer visit management method based on intelligent interaction according to claim 1, characterized in that, Based on the user's travel information and preset visit conditions, a visit list is generated, including: Based on the preset visit criteria, candidate visit targets are obtained by filtering from the customer database; the customer database is constructed based on all customer data in advance. The candidate visit object and the user itinerary information are input into the visit object prediction model to extract the features of the input candidate visit object and user itinerary information and perform lexicalization respectively to obtain the object lexical corresponding to the candidate visit object and the itinerary lexical corresponding to the user itinerary information. For each of the aforementioned trip terms, a self-attention mechanism is used to determine the attention score with each object term, and based on the corresponding attention score, the objects to be visited in the user's trip are predicted to generate a visit list.

3. The customer visit management method based on intelligent interaction according to claim 2, characterized in that, Based on the preset visit criteria, candidate visit targets are obtained by filtering from the customer database, including: The preset visit conditions are parsed to obtain the query conditions and priority rules; The query conditions are concatenated into a Structured Query Language (SQL) query statement and executed to search the customer database and obtain candidate customers. Based on the candidate customers and in conjunction with the priority rules, a comprehensive evaluation is performed on each candidate customer to obtain the corresponding candidate customer's score result; Based on the rating results of each candidate customer, select the target number of candidate customers in descending order of their scores to obtain the candidate visit targets.

4. The customer visit management method based on intelligent interaction according to claim 1, characterized in that, Based on the visit list and the user itinerary information, the target visitees are determined, and the geographical locations of the corresponding target visitees are obtained. Visit route planning is then performed to obtain a visit plan, including: Based on the visit list, identify the target visitors and obtain the geographical locations of the corresponding target visitors; Based on the geographical location of each target visitor and the user's travel information, the K-means clustering algorithm is used, combined with a preset cluster size constraint, to cluster each target visitor and obtain the clustering result; Based on the clustering results, the inter-cluster distances between each cluster are determined, and the cluster visiting order is planned according to the inter-cluster distances to obtain the inter-cluster path planning results; Based on the clustering results, intra-cluster path planning is performed for each cluster according to the geographical location of each target visit object within the corresponding cluster, resulting in intra-cluster path planning results for each cluster. Based on the intra-cluster path planning results of each cluster, and combined with the inter-cluster path planning results and the preset visit duration of each visited object, the visit time of each visited object is determined, and a visit plan is obtained.

5. The customer visit management method based on intelligent interaction according to claim 4, characterized in that, Based on the clustering results, intra-cluster path planning is performed for each cluster according to the geographical location of each target visited object within the corresponding cluster, resulting in intra-cluster path planning results for each cluster, including: Based on the clustering results, for each cluster, the distance between any two target objects is determined according to the geographical location of each target object within the corresponding cluster, thus obtaining object distance information; Based on the inter-cluster path planning results, when the number of target visited objects within the corresponding cluster is less than or equal to a preset threshold, a starting point is selected and a dynamic planning table is constructed; wherein, the rows of the dynamic planning table represent the access status of each target visited object within the cluster, and the columns of the dynamic planning table represent the currently selected target visited object; For the current node of each path in the dynamic programming table, iteratively search for unvisited nodes, select the unvisited node with the smallest distance as the optimal predecessor node of the current node, update the path in the dynamic programming table, and take the optimal predecessor node corresponding to each path as the current node. Repeat this step until a path containing all target visited objects in the corresponding cluster is obtained. For each path finally obtained in the dynamic planning table, the distance between adjacent path nodes and the distance between the starting point and the ending point of the path are determined based on the object distance information, or based on the object distance information, the geographical location of each target visited object in the cluster and the preset residential location, to obtain the total path length. The path with the minimum total path length is selected as the intra-cluster path planning result of the corresponding cluster.

6. The customer visit management method based on intelligent interaction according to claim 4, characterized in that, Based on the geographical location of each target visitor and the user's travel information, the K-means clustering algorithm, combined with a preset cluster size constraint, is used to cluster each target visitor, resulting in the following clustering results: Based on the user travel information, the number of clusters is determined, and combined with the geographical location of each target visit object, the K-means clustering algorithm is used to obtain the initial clustering results; For the initial clustering results, the number of target visited objects within each initial cluster is detected. If the number exceeds the preset cluster size constraint, the objects within the corresponding initial cluster are migrated and adjusted in order of distance from the cluster center, until all initial clusters satisfy the preset cluster size constraint, thus obtaining the clustering result; or, Based on the user's itinerary information, determine the number of clusters and initialize the cluster centers; Based on the geographical location of each target visitor, the distance between each target visitor and each cluster center is determined, and under the condition of satisfying the preset cluster size constraint, each target visitor is assigned to the corresponding cluster according to the distance, and the cluster members are updated; The cluster centers are re-determined based on the updated geographical locations of the cluster members, and the above steps are repeated until the iteration termination condition is met to obtain the clustering result.

7. The customer visit management method based on intelligent interaction according to claim 4, characterized in that, Based on the clustering results, for each cluster, intra-cluster path planning is performed according to the geographical location of each target visited object within the corresponding cluster, resulting in intra-cluster path planning results for each cluster, which also includes: Based on the clustering results, for each cluster, the distance between any two target objects is determined according to the geographical location of each target object within the corresponding cluster, thus obtaining object distance information; Based on the inter-cluster path planning results, when the number of target visited objects within the corresponding cluster is greater than the preset number threshold, based on the object distance information and the geographical location of each target visited object within the corresponding cluster, a minimum spanning tree (MST) is constructed with the target visited object within the cluster as a node, connecting all nodes within the cluster and having the minimum total weight of the edges. Traverse the MST, count the degree of each node, and filter out all nodes with odd degrees to obtain a set of odd-degree nodes. Then, use the minimum weight perfect matching algorithm to find the matching method with the fewest number of edges and the smallest total weight of the edges in the set of odd-degree nodes, so that each odd-degree node is connected to a matching edge. Based on the matching edges and the MST, an Eulerian graph is formed, and the Eulerian circuit of the Eulerian graph is converted into a Traveling Salesman Problem (TSP) path to obtain the initial visiting order; Based on the initial visit order, multiple neighborhood candidate paths are generated, and the total cost of each neighborhood candidate path is determined. Based on the total cost of each neighborhood candidate path, the neighborhood candidate path with the lowest total cost is selected as the optimal path for the current iteration step, and the iteration is repeated until a preset time threshold is reached. The final optimal path is then used as the intra-cluster path planning result for the corresponding cluster.

8. The customer visit management method based on intelligent interaction according to claim 1, characterized in that, The historical visit records include structured data and unstructured data; Based on the visit plan, historical visit records and target visit information of the corresponding target visit object are obtained, and a visit strategy for the corresponding target visit object is generated, including: Based on the visit plan, obtain the planned route for each day during the user's trip, obtain the corresponding target visit object for each day, and obtain the historical visit records and target visit object information for the corresponding target visit object; For each day, based on the historical visit records of the corresponding target, the basic statistical features in the corresponding structured data are extracted, and the corresponding unstructured data is segmented, and the content of each segment is segmented into words. The term frequency-inverse document frequency (TF-IDF) of each word segment is determined to obtain the text vector. Then, the sentiment analysis model is used to predict the sentiment score of the corresponding segment content, and the sentiment score of the corresponding structured data is determined. Based on the historical visit records of all target visitors on the same day, the corresponding historical visit records are clustered by topic using the latent Dirichlet assignment model, and the probability distribution of the historical visit records of each target visitor belonging to each topic is determined, as well as the visit frequency of each target visitor is determined. Based on the target audience information of the corresponding target audience, determine the proportion of positive, neutral and negative reports about the corresponding target audience in the historical target time period, and determine the recent evaluation index; Based on the visit frequency and recent evaluation index of each target visitee, principal component analysis or weighted summation is used to obtain the comprehensive activity score of the corresponding target visitee; Based on the basic statistical characteristics, sentiment scores, topic probability distributions, and comprehensive activity scores of each target visitor, the target visitors are stratified or clustered to divide them into different groups, resulting in corresponding user stratification labels; wherein, the user stratification labels are used to characterize the value risk of users; Based on the basic statistical characteristics, sentiment scores, topic probability distributions, and comprehensive activity scores of each target visitor, the intent of each target visitor is predicted to determine the potential behavior of the corresponding target visitor in the future target time. Based on the user segmentation tags of the target visit object and the potential behaviors of the target visit object, a visit strategy corresponding to the target visit object is generated according to a preset strategy template.

9. The customer visit management method based on intelligent interaction according to claim 1, characterized in that, After generating the visit policy for the corresponding target visited object, the following is also included: Receive visit recordings uploaded by users; wherein, the visit recordings are obtained based on the visit strategy for the corresponding target visit object in the visit plan; The recorded visit was subjected to speech recognition, and a visit summary was generated. Based on the preset fields to be entered and the first preset prompt word template, the corresponding fields are extracted from the visit summary content and entered into the previously stored visit records to obtain the visit summary; wherein, the visit records are generated in advance based on the customer visit information input by the user, combined with the preset fields to be entered and the preset prompt word template, and the fields corresponding to the user visit information are extracted.

10. A customer visit management device based on intelligent interaction, characterized in that, include: The information acquisition module retrieves user travel information; The list generation module generates a visit list based on the user's itinerary information and preset visit conditions; The visit planning module determines the target visit object based on the visit list and the user's itinerary information, obtains the geographical location of the corresponding target visit object, plans the visit route, and obtains the visit plan. The strategy generation module obtains historical visit records and target visit information of the corresponding target visit object according to the visit plan, and generates a visit strategy for the corresponding target visit object.

Images