Customer management method and device, equipment and storage medium

Abstract

The invention relates to the technical field of data processing, and discloses a customer management method, device and equipment and a storage medium, and the method is used for establishing a dynamic customer portrait and managing a customer according to the dynamic customer portrait. The method comprises the following steps: acquiring multi-source heterogeneous logistics data, and preprocessing the multi-source heterogeneous logistics data to obtain preprocessed data; fusing the preprocessed data by using a trained multi-modal data fusion model to obtain fused data; based on a preset customer label system, performing feature extraction on the fused data through a machine learning algorithm, and generating a dynamic customer portrait according to the extracted features; generating a personalized service strategy through a multi-objective optimization algorithm according to the dynamic customer portrait and the real-time logistics state, and storing the personalized service strategy in a preset personalized service strategy library; and monitoring customer logistics data in real time, and when a preset abnormal condition occurs, triggering an early warning mechanism and matching a corresponding personalized service strategy from the personalized service strategy library to start a corresponding customer service response process.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a customer management method, device, equipment and storage medium. Background Art

[0002] Amidst the rapid development of the modern logistics industry, customer management has become a core component for companies to enhance their competitiveness and strengthen customer loyalty. As logistics operations continue to expand and customer needs become increasingly diverse, accurately understanding customer needs, providing personalized services, and efficiently handling logistics anomalies have become key to gaining market advantage for logistics companies.

[0003] Traditional logistics customer management methods primarily rely on single-dimensional data, such as order value and delivery frequency, to categorize and manage customers through simple statistical analysis. This approach struggles to fully and deeply understand customer characteristics and needs, and is unable to adapt to a market environment characterized by dynamic changes in customer demands. This results in a lack of targeted customer service strategies, making it difficult to improve customer satisfaction and loyalty. Furthermore, when handling logistics anomalies, traditional methods often rely on manual judgment and decision-making, resulting in slow response and low efficiency, which can easily lead to customer complaints and business losses.

[0004] With the development of technologies such as the Internet of Things, big data, and artificial intelligence, the application of multimodal data in the logistics field has gradually attracted attention. Multimodal logistics data covers a variety of types, including order data, transportation trajectory data, warehousing data, and customer feedback data, and can reflect the comprehensive picture of logistics operations and customer needs from multiple perspectives. However, existing technologies still face many problems when processing multimodal logistics data. On the one hand, multi-source and heterogeneous logistics data have significant differences in format, structure, and semantics, making data collection, preprocessing, and fusion difficult, making effective integration and utilization difficult. On the other hand, existing customer profile generation methods are mostly based on static data and cannot reflect changes in customer needs in real time. At the same time, there is a lack of effective algorithmic and model support for the generation of personalized service strategies and the handling of exceptions, making it difficult to achieve fast and accurate decision-making.

[0005] Therefore, the existing technology still needs to be improved and developed. Summary of the Invention

[0006] The present invention provides a customer management method, apparatus, device and storage medium for establishing dynamic customer portraits and managing customers based on the dynamic customer portraits.

[0007] The first aspect of the present invention provides a customer management method, which includes: obtaining multi-source heterogeneous logistics data, preprocessing the multi-source heterogeneous logistics data to obtain preprocessed data; using a trained multimodal data fusion model to fuse the preprocessed data to obtain fused data; based on a preset customer tag system, extracting features from the fused data through a machine learning algorithm, and generating a dynamic customer profile based on the extracted features; obtaining real-time logistics status, and generating a personalized service strategy through a multi-objective optimization algorithm based on the dynamic customer profile and the real-time logistics status, and storing it in a preset personalized service strategy library; monitoring customer logistics data in real time, and when a preset abnormal situation occurs, triggering an early warning mechanism and matching the corresponding personalized service strategy from the personalized service strategy library to start the corresponding customer service response process.

[0008] Optionally, in a first implementation method of the first aspect of the present invention, the acquiring of multi-source heterogeneous logistics data and preprocessing of the multi-source heterogeneous logistics data to obtain preprocessed data include: acquiring multi-source heterogeneous logistics data, the multi-source heterogeneous logistics data including order data, warehousing data, transportation trajectory data, distribution data and customer feedback data; using a hash algorithm based on the unique identification code of the multi-source heterogeneous logistics data to identify and delete duplicate records in the multi-source heterogeneous logistics data to obtain non-duplicate data; performing data conversion and normalization on the non-duplicate data to obtain preprocessed data.

[0009] Optionally, in a second implementation method of the first aspect of the present invention, the pre-processed data is fused using a trained multimodal data fusion model to obtain fused data, including: constructing a multimodal model based on a Transformer architecture, and training the multimodal model to obtain a trained multimodal data fusion model; performing spatiotemporal alignment and semantic association on the pre-processed data to obtain associated data; and inputting the associated data into the trained multimodal data fusion model to obtain fused data.

[0010] Optionally, in a third implementation method of the first aspect of the present invention, the spatiotemporal alignment and semantic association of the preprocessed data to obtain associated data includes: applying a time window matching algorithm to establish a time correspondence between different modal data in the preprocessed data to obtain time-aligned data; converting spatial data in different coordinate systems in the time-aligned data to a unified coordinate system, and establishing an association relationship between the data based on spatiotemporal constraints to obtain spatiotemporal aligned data; calculating the semantic similarity of different modal data in the spatiotemporal aligned data, and establishing a semantic relationship based on a preset similarity threshold to obtain associated data.

[0011] Optionally, in a fourth implementation of the first aspect of the present invention, the preset customer label system is based on which features of the fused data are extracted through a machine learning algorithm, and a dynamic customer portrait is generated based on the extracted features, including: performing unsupervised learning on the fused data through a generative adversarial network to extract feature data; matching and mapping the extracted feature data with the labels in the preset customer label system; calculating the label value of the label corresponding to each customer based on the matching result, and weightedly fusing the calculated multiple label values ​​according to the customer label system to generate a dynamic customer portrait for each customer, and storing it in a preset personalized service strategy library.

[0012] Optionally, in a fifth implementation method of the first aspect of the present invention, the real-time logistics status is obtained, and a personalized service strategy is generated through a multi-objective optimization algorithm according to the dynamic customer portrait and the real-time logistics status, and stored in a preset personalized service strategy library, including: obtaining the real-time logistics status, associating the real-time logistics status with the corresponding dynamic customer portrait based on the logistics order number, and integrating the key features of the real-time logistics status into the dynamic customer portrait to obtain integrated data; based on the integrated data, analyzing the customer's current logistics scenario, and judging the customer's potential needs and expected services in the corresponding logistics scenario according to the historical behavioral preferences and service demand levels in the dynamic customer portrait; using a multi-objective particle swarm optimization algorithm to iteratively search and optimize the integrated data to generate a personalized service strategy, which includes a priority delivery plan, an exclusive customer service follow-up plan, a customized logistics route adjustment, and a value-added service recommendation list.

[0013] Optionally, in a sixth implementation method of the first aspect of the present invention, the real-time monitoring of customer logistics data, when a preset abnormal situation occurs, triggers an early warning mechanism and matches a corresponding personalized service strategy from the personalized service strategy library to start a corresponding customer service response process, including: real-time monitoring of customer logistics data, and according to preset abnormal rules and thresholds, judging in real time whether the customer logistics data meets the abnormal conditions; when the customer logistics data meets the abnormal conditions, judging the abnormal type of the customer logistics data, and matching the corresponding personalized service strategy from the preset personalized service strategy library according to the abnormal type and dynamic customer portrait; according to the matched personalized service strategy, starting the corresponding customer service response process, clarifying the responsible personnel, work content and time nodes of each link.

[0014] The second aspect of the present invention provides a customer management device, including: a preprocessing module for acquiring multi-source heterogeneous logistics data, preprocessing the multi-source heterogeneous logistics data, and obtaining preprocessed data; a fusion module for using a trained multimodal data fusion model to fuse the preprocessed data to obtain fused data; a first generation module for extracting features from the fused data through a machine learning algorithm based on a preset customer tag system, and generating a dynamic customer portrait based on the extracted features; a second generation module for acquiring real-time logistics status, generating a personalized service strategy through a multi-objective optimization algorithm based on the dynamic customer portrait and the real-time logistics status, and storing it in a preset personalized service strategy library; a calling module for real-time monitoring of customer logistics data, and when a preset abnormal situation occurs, triggering an early warning mechanism and matching the corresponding personalized service strategy from the personalized service strategy library to start the corresponding customer service response process.

[0015] Optionally, in a first implementation of the second aspect of the present invention, the preprocessing module includes: a first acquisition unit, used to acquire multi-source heterogeneous logistics data, the multi-source heterogeneous logistics data including order data, warehousing data, transportation trajectory data, distribution data and customer feedback data; a deletion unit, used to identify and delete duplicate records in the multi-source heterogeneous logistics data based on the unique identification code of the multi-source heterogeneous logistics data using a hash algorithm to obtain non-duplicate data; a preprocessing unit, used to perform data conversion and normalization on the non-duplicate data to obtain preprocessed data.

[0016] Optionally, in a second implementation of the second aspect of the present invention, the fusion module includes: a construction unit, used to construct a multimodal model based on the Transformer architecture, and train the multimodal model to obtain a trained multimodal data fusion model; an association unit, used to perform spatiotemporal alignment and semantic association on the preprocessed data to obtain associated data; and a fusion unit, used to input the associated data into the trained multimodal data fusion model to obtain fused data.

[0017] Optionally, in a third implementation of the second aspect of the present invention, the first generation module includes: an extraction unit, used to perform unsupervised learning on the fused data through a generative adversarial network to extract feature data; a matching unit, used to match and map the extracted feature data with labels in a preset customer label system; a first generation unit, used to calculate the label value of the label corresponding to each customer based on the matching result, and according to the customer label system, weightedly fuse the calculated multiple label values ​​to generate a dynamic customer portrait for each customer, and store it in a preset personalized service strategy library.

[0018] Optionally, in a fourth implementation method of the second aspect of the present invention, the second generation module includes: an integration unit for obtaining real-time logistics status, associating the real-time logistics status with the corresponding dynamic customer portrait based on the logistics order number, and integrating the key features of the real-time logistics status into the dynamic customer portrait to obtain integrated data; a judgment unit for analyzing the customer's current logistics scenario based on the integrated data, and judging the customer's potential needs and expected services in the corresponding logistics scenario based on the historical behavioral preferences and service demand levels in the dynamic customer portrait; a second generation unit for iteratively searching and optimizing the integrated data using a multi-objective particle swarm optimization algorithm to generate a personalized service strategy, which includes a priority delivery plan, an exclusive customer service follow-up plan, a customized logistics route adjustment, and a value-added service recommendation list.

[0019] Optionally, in a fifth implementation of the second aspect of the present invention, the calling module includes: a monitoring unit, which is used to monitor customer logistics data in real time, and determine in real time whether the customer logistics data meets the abnormal conditions according to preset abnormal rules and thresholds; a calling unit, which is used to determine the abnormal type of the customer logistics data when the customer logistics data meets the abnormal conditions, and match the corresponding personalized service strategy from a preset personalized service strategy library based on the abnormal type and dynamic customer portrait; a determination unit, which is used to start the corresponding customer service response process according to the matched personalized service strategy, and clarify the responsible personnel, work content and time nodes of each link.

[0020] The third aspect of the present invention provides a customer management device, comprising: a memory and at least one processor, wherein the memory stores computer-readable instructions, and the memory and the at least one processor are interconnected via a line; the at least one processor calls the computer-readable instructions in the memory so that the customer management device executes the various steps of the customer management method described above.

[0021] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, which, when executed on a computer, enable the computer to execute the various steps of the customer management method described above.

[0022] In the technical solution provided by the present invention, by integrating multi-source heterogeneous logistics data and applying advanced data processing and intelligent algorithm technologies, the full process of logistics customer management is realized to be intelligent and precise. First, the acquisition and preprocessing of multi-source heterogeneous data ensure the quality and availability of the original data, laying a solid foundation for subsequent analysis; the trained multimodal data fusion model breaks down data barriers, deeply mines the potential connections between data, and generates high-value fused data; secondly, the dynamic customer portrait constructed based on the preset label system and machine learning algorithm can reflect customer characteristics and demand changes in real time and comprehensively. Then, combined with the real-time logistics status, the multi-objective optimization algorithm is used to generate personalized service strategies, accurately match customer needs and enterprise resources, and improve service quality and operational efficiency. Finally, the real-time monitoring and intelligent early warning mechanism can promptly detect and respond to logistics anomalies, quickly retrieve response plans from the strategy library, and ensure customer experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A first flow chart of the customer management method provided by an embodiment of the present invention;

[0024] Figure 2 A second flow chart of the customer management method provided by an embodiment of the present invention;

[0025] Figure 3 A third flow chart of the customer management method provided by an embodiment of the present invention;

[0026] Figure 4 A fourth flow chart of the customer management method provided by an embodiment of the present invention;

[0027] Figure 5 A fifth flow chart of the customer management method provided by an embodiment of the present invention;

[0028] Figure 6 A sixth flow chart of the customer management method provided by an embodiment of the present invention;

[0029] Figure 7 A schematic diagram of the structure of a customer management device provided by an embodiment of the present invention;

[0030] Figure 8 A schematic diagram of the structure of a customer management device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] Embodiments of the present invention provide a customer management method, apparatus, device, and storage medium. The method is used to establish a dynamic customer profile and manage customers based on the dynamic customer profile.

[0032] The terms "first," "second," "third," "fourth," and the like (if any) in the description and claims of the present invention and in the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatus.

[0033] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 A first embodiment of a customer management method in an embodiment of the present invention includes:

[0034] S101. Acquire multi-source heterogeneous logistics data, pre-process the multi-source heterogeneous logistics data, and obtain pre-processed data.

[0035] It is understandable that the execution subject of the present invention may be a client management device, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0036] In this embodiment, multi-source heterogeneous logistics data covering different types such as orders, transportation tracks, warehousing, distribution, and customer feedback are collected through various channels such as the logistics information system API interface, sensor equipment, and customer terminal applications.

[0037] In this embodiment, a duplicate checking algorithm is used to remove duplicate data; statistical analysis methods are used to process missing values, such as filling numerical data with means and medians, and extracting key information from unstructured text data through natural language processing technology; outliers are identified and corrected based on methods such as box plots and the 3σ principle.

[0038] S102: Use the trained multimodal data fusion model to fuse the preprocessed data to obtain fused data.

[0039] In this embodiment, the multimodal data fusion model adopts a deep learning architecture, such as convolutional neural network (CNN), recurrent neural network (RNN) and its variants (LSTM, GRU), Transformer, etc., or an integrated learning method. During model training, a large amount of historical multi-source heterogeneous data is used as training samples, and the model parameters are optimized through the back propagation algorithm so that it can learn the mapping relationship and feature association between different modal data. When fusing the preprocessed data, the different modal data are first input into the sub-network corresponding to the model for feature extraction. For example, for image-type logistics document data, visual features are extracted through CNN, and for time series transport trajectory data, time series features are extracted through RNN. Then, the features extracted by each sub-network are fused using structures such as attention mechanism and fusion layer to generate fused data containing multimodal information. This fusion method can give full play to the advantages of different modal data, mine the hidden correlation information between data, and improve the availability and value density of data.

[0040] S103. Based on the preset customer tag system, feature extraction is performed on the fused data through a machine learning algorithm, and a dynamic customer profile is generated based on the extracted features.

[0041] In this embodiment, the preset customer tag system is constructed based on various behaviors and attributes of customers in the logistics business, covering multiple dimensions such as basic attribute tags (age, gender, region), logistics behavior tags (order frequency, order amount, common delivery address), service preference tags (delivery time requirements, logistics method preferences), and value assessment tags (consumption contribution, potential value). Machine learning algorithms (such as clustering algorithms K-Means and DBSCAN for discovering customer group characteristics, and classification algorithms SVM and random forest for predicting customer behavior tendencies) are used to extract features from the fused data, screening out key features that have a significant impact on customer profile construction from massive data. For example, by analyzing data such as the order time, product type, and delivery address of customers' historical orders, clustering algorithms are used to divide customers into different groups. Then, combined with classification algorithms, each group's acceptance and demand for different logistics services are predicted. Based on the extracted features, each customer is assigned a corresponding tag, thereby constructing a dynamic customer profile that can reflect changes in customer needs and behavior in real time.

[0042] S104. Obtain real-time logistics status, generate personalized service strategies through multi-objective optimization algorithms based on dynamic customer profiles and real-time logistics status, and store them in a preset personalized service strategy library.

[0043] In this embodiment, real-time logistics status data is acquired, including information such as the current location of the goods, estimated arrival time, transportation vehicle operating status, and inventory status changes. This data is updated in real time via IoT sensors, GPS positioning devices, and logistics information systems. Combined with dynamic customer profiles, multi-objective optimization algorithms (such as genetic algorithms, particle swarm optimization algorithms, and simulated annealing algorithms) are used to generate personalized service strategies with the goals of maximizing customer satisfaction, minimizing logistics costs, and optimizing service efficiency. This strategy takes into account customer requirements for delivery time and service quality, as well as the allocation of logistics resources (vehicles, personnel, and warehousing). For example, for customers with high requirements for delivery timeliness, dedicated vehicle delivery is prioritized, and delivery routes are optimized. For price-sensitive customers, affordable logistics solutions are selected and coupons are provided. The generated personalized service strategies are stored in a pre-set personalized service strategy library in a structured format, including information such as the strategy name, applicable scenarios, specific measures, and priority, to facilitate subsequent rapid retrieval and access.

[0044] S105. Monitor customer logistics data in real time. When a preset abnormal situation occurs, trigger the early warning mechanism and match the corresponding personalized service strategy from the personalized service strategy library to start the corresponding customer service response process.

[0045] In this embodiment, customer logistics data is continuously monitored through real-time data collection and analysis tools, and various anomaly detection rules are set, such as when the transportation time exceeds a certain percentage of the estimated time, the temperature / humidity of the goods exceeds the normal range, and the inventory quantity is lower than the safety threshold as preset abnormal conditions. Once an abnormality is detected, the early warning mechanism is immediately triggered, and relevant personnel are notified through SMS, email, system pop-up windows, etc. At the same time, according to information such as the type and severity of the abnormal situation, matching is performed in the personalized service strategy library, and the most suitable personalized service strategy is found using similarity measurement algorithms (such as cosine similarity and Euclidean distance). For example, when a delay in cargo transportation is detected, a service strategy that includes measures such as expedited vehicle dispatch, priority delivery, and timely notification to customers is matched, and the corresponding customer service response process is quickly initiated to coordinate resources from all parties to solve the problem and provide timely feedback to customers on the progress of the processing.

[0046] This embodiment provides a customer management method that integrates multi-source heterogeneous logistics data and uses advanced data processing and intelligent algorithm technologies to achieve full-process intelligence and precision in logistics customer management. First, the acquisition and preprocessing of multi-source heterogeneous data ensure the quality and availability of the original data, laying a solid foundation for subsequent analysis; the trained multimodal data fusion model breaks down data barriers, deeply mines the potential connections between data, and generates high-value fused data; secondly, the dynamic customer portrait constructed based on the preset label system and machine learning algorithm can reflect customer characteristics and demand changes in real time and comprehensively; then, combined with the real-time logistics status, the multi-objective optimization algorithm is used to generate personalized service strategies, accurately match customer needs with enterprise resources, and improve service quality and operational efficiency; finally, the real-time monitoring and intelligent early warning mechanism can promptly detect and respond to logistics anomalies, quickly retrieve response plans from the policy library, and ensure customer experience.

[0047] See also Figure 2 The second embodiment of the customer management method in the embodiment of the present invention includes:

[0048] S201. Acquire multi-source heterogeneous logistics data, which includes order data, warehousing data, transportation trajectory data, distribution data, and customer feedback data.

[0049] In this embodiment, through multiple channels such as the enterprise's internal information system (such as ERP, WMS, TMS systems), Internet of Things devices (such as vehicle-mounted GPS, warehouse sensors), and customer terminal applications (such as logistics query apps and ordering platforms), comprehensive collection of order data (including order number, product details, order time, etc.), warehouse data (covering inventory quantity, in and out records, cargo location information), transportation trajectory data (such as real-time vehicle location, driving route, speed), delivery data (including delivery personnel information, delivery status, estimated delivery time), and customer feedback data (such as evaluation content, complaint records, and consulting questions).

[0050] S202. A hash algorithm is used based on the unique identification code of the multi-source heterogeneous logistics data to identify and delete duplicate records in the multi-source heterogeneous logistics data to obtain non-duplicate data.

[0051] In this embodiment, a hash algorithm is used to process the data based on unique identifiers (e.g., unique order numbers for orders and unique barcode numbers for goods) in multi-source heterogeneous logistics data. The hash algorithm maps the unique identifiers to fixed-length hash values, quickly identifying duplicate records by comparing the hash values. Once data with identical hash values ​​is found, it is identified as a duplicate and deleted, resulting in non-duplicate data.

[0052] S203: Perform data conversion and normalization processing on the non-repetitive data to obtain pre-processed data.

[0053] In this example, non-repetitive data is converted. Unstructured data (such as textual customer feedback) is segmented and tagged with parts of speech to be converted into computer-readable structured data. Semi-structured data (such as logistics documents in JSON or XML format) is parsed into a unified relational data structure. Structured data is standardized in data type (for example, dates in different systems are formatted as "YYYY-MM-DD") and encoding. Next, normalization is performed. Numerical data (such as weight and distance) is mapped to a specific interval (such as [0, 1] or [-1, 1]) to eliminate dimensional differences between the data.

[0054] In this embodiment, unique identification codes and hash algorithms are used to accurately remove duplicates and reduce redundancy, and then data conversion and normalization are performed to unify data formats and standards, which facilitates subsequent analysis and mining, laying the foundation for improving logistics service efficiency and decision-making accuracy.

[0055] See also Figure 3 A third embodiment of a customer management method in an embodiment of the present invention includes:

[0056] S301. Construct a multimodal model based on the Transformer architecture and train the multimodal model to obtain a trained multimodal data fusion model.

[0057] In this embodiment, first, a multimodal model based on the Transformer architecture is built. The architecture consists of an encoder and a decoder, in which the multi-head attention mechanism can capture data features from different angles at the same time, and the position encoding can process the sequential information of the data. During the construction process, according to the multimodal characteristics of logistics data, a suitable input layer is designed to receive different types of pre-processed data, such as image-type logistics documents, time-series transport trajectory data, etc., which are input after adaptation processing. After the model is built, a large amount of historical logistics data containing multiple modalities is used as a training set. Through the back propagation algorithm, the optimization goal is to minimize the loss function (such as mean square error, cross entropy loss, etc., and the appropriate loss function is selected according to the fusion goal). The model parameters are continuously adjusted. After multiple rounds of iterative training, a trained multimodal data fusion model is obtained, which enables it to accurately learn the intrinsic connection and fusion mode between different modal data.

[0058] S302: Perform spatiotemporal alignment and semantic association on the preprocessed data to obtain associated data.

[0059] In this embodiment, the preprocessed data is subjected to spatiotemporal alignment and semantic association to obtain associated data, including: applying a time window matching algorithm to establish a time correspondence between different modal data in the preprocessed data to obtain time-aligned data; converting spatial data in different coordinate systems in the time-aligned data to a unified coordinate system, and establishing an association relationship between the data based on spatiotemporal constraints to obtain spatiotemporal aligned data; calculating the semantic similarity of different modal data in the spatiotemporal aligned data, and establishing a semantic relationship based on a preset similarity threshold to obtain associated data.

[0060] S303: Input the associated data into the trained multimodal data fusion model to obtain fused data.

[0061] In this embodiment, the associated data, after undergoing spatiotemporal alignment and semantic association processing, is organized according to the model's input format requirements and fed into a trained multimodal data fusion model. The model's encoder extracts and fuses features from the input data using a multi-head attention mechanism and a multi-layer neural network, mining the underlying deep information within the data. The decoder transforms and reconstructs the fused features and outputs the fused data.

[0062] In this embodiment, a multimodal model is constructed with the Transformer architecture as the core. By utilizing the powerful feature capture and cross-modal interaction capabilities of its self-attention mechanism, the complex dependencies and potential connections between multimodal logistics data can be effectively mined. By performing spatiotemporal alignment and semantic association on the pre-processed data, the standards of the data in the time, space dimensions and semantic levels are unified, eliminating the fusion barriers caused by data heterogeneity and significantly improving the consistency and availability of the data. By inputting the associated data into the trained model for fusion, highly integrated and valuable information-rich fusion data can be generated, providing accurate and comprehensive data support for subsequent logistics business analysis and customer management decisions.

[0063] See also Figure 4 A fourth embodiment of a customer management method according to an embodiment of the present invention includes:

[0064] S401. Perform unsupervised learning on the fused data through a generative adversarial network to extract feature data.

[0065] In this embodiment, a generative adversarial network (GAN) consists of a generator and a discriminator. During unsupervised learning of fused data, the generator attempts to learn the underlying distribution patterns of the fused data and generate new data that approximates the characteristics of the real data. The discriminator is responsible for distinguishing between the real fused data and the data generated by the generator. During training, the two interact with each other, with the generator continuously optimizing to produce more realistic data and the discriminator continuously improving its discriminative capabilities. With each iteration of training, the generator gradually extracts representative features from the fused data. These features cover a wide range of potential information about customers, including their logistics behavior and demand preferences.

[0066] S402: Match and map the extracted feature data with tags in a preset customer tag system.

[0067] In this embodiment, the preset customer label system is pre-set according to the logistics business scenario and customer analysis needs, including basic attribute labels (such as age, gender, region), logistics behavior labels (order frequency, order amount, common delivery address), service preference labels (delivery time requirements, logistics method preference) and other dimensions. The extracted feature data is compared with the label system one by one, and the correspondence between each feature data and the label in the label system is determined through technical means such as semantic analysis and data association to complete the mapping of feature data to labels. For example, if the feature data shows that a customer frequently places orders at night, it will be mapped to the "delivery time preference-night delivery" label to achieve effective connection between feature data and the label system, laying a structured foundation for the construction of customer portraits.

[0068] S403. Calculate the tag value of each customer's corresponding tag based on the matching results, and weight the calculated multiple tag values ​​according to the customer tag system to generate a dynamic customer profile for each customer, and store it in a preset personalized service strategy library.

[0069] In this embodiment, based on the matching results of feature data and labels, statistical analysis, weight calculation and other methods are used to calculate the specific value of the corresponding label for each customer. For example, for the "order frequency" label, the number of orders placed by the customer within a certain period of time is counted, and the corresponding numerical value is assigned in combination with factors such as the industry average level; for the "logistics method preference" label, the value is calculated based on the proportion of logistics methods selected by the customer's historical orders. All calculated label values ​​are weighted and fused according to the preset customer label system to form a dynamic customer portrait that fully reflects the customer's characteristics and needs. This portrait will be continuously updated with new fused data, reflecting changes in customer behavior and needs in real time. Finally, the generated dynamic customer portrait is stored in the preset personalized service strategy library, which is convenient for the subsequent generation of personalized service strategies based on the customer portrait, and provides customers with logistics services that better meet their needs.

[0070] In this embodiment, with the help of the powerful unsupervised learning ability of the generative adversarial network, it is possible to automatically mine the complex features hidden in the fused data, and to achieve in-depth analysis of customer information without a large amount of labeled data; the extracted features are accurately matched and mapped with the preset customer label system, so that customer information is structured and labeled for easy management and analysis; by calculating the label values ​​and integrating them to generate dynamic customer portraits, changes in customer status can be reflected in real time, providing accurate and dynamic data support for the formulation of personalized service strategies, and effectively improving the accuracy and timeliness of customer management.

[0071] See also Figure 5 A fifth embodiment of a customer management method according to an embodiment of the present invention includes:

[0072] S501. Obtain the real-time logistics status, associate the real-time logistics status with the corresponding dynamic customer profile based on the logistics order number, integrate the key features of the real-time logistics status into the dynamic customer profile, and obtain integrated data.

[0073] In this embodiment, real-time logistics status is obtained through channels such as logistics information systems and IoT devices, including information such as cargo transportation location, estimated arrival time, transportation vehicle operating status, and warehouse inventory changes. Using the logistics order number as a unique identifier, these real-time logistics statuses are associated and matched with corresponding dynamic customer profiles stored in a preset personalized service strategy library. For example, based on the logistics order number of an order, the customer profile corresponding to the order is found, and information such as the customer's historical behavior preferences and service demand level is obtained. Subsequently, key features of the real-time logistics status, such as cargo delay information and special transportation conditions, are extracted and integrated into the dynamic customer profile, supplementing and updating the customer's information at the current logistics stage, forming integrated data that includes the customer's historical information and real-time status, providing a comprehensive data foundation for subsequent analysis.

[0074] S502. Based on the integrated data, analyze the customer's current logistics scenario, and determine the customer's potential needs and expected services in the corresponding logistics scenario based on the historical behavioral preferences and service demand levels in the dynamic customer profile.

[0075] In this embodiment, based on the integrated data, natural language processing, data mining and other technologies are used to analyze the customer's current logistics scenario. For example, if the goods are in transit and there is a delay, the logistics scenario is judged to be a "transportation delay scenario"; if the goods arrive at the destination and are waiting for delivery, it is a "pending delivery scenario." At the same time, combined with historical behavioral preferences in dynamic customer portraits (such as whether expedited delivery has been selected multiple times, preferred delivery time), service demand level (ordinary customers, VIP customers) and other information, a comprehensive judgment is made on the customer's potential needs and expected services in the corresponding logistics scenario. For example, for VIP customers who have selected expedited delivery multiple times, in the transportation delay scenario, it is inferred that their potential needs are to speed up delivery, obtain the cause of the delay and the solution in a timely manner; in the pending delivery scenario, they expect to receive priority delivery arrangements and pre-delivery reminder services.

[0076] S503. Use a multi-objective particle swarm optimization algorithm to iteratively search and optimize the integrated data to generate a personalized service strategy. The personalized service strategy includes a priority delivery plan, an exclusive customer service follow-up plan, a customized logistics route adjustment, and a value-added service recommendation list.

[0077] In this embodiment, a multi-objective particle swarm optimization algorithm simulates the foraging behavior of a flock of birds, treating the integrated data as a search space, with each particle representing a potential service strategy solution. During the iterative process, the algorithm continuously updates the position and velocity of the particles based on multiple objectives, such as potential customer demand, logistics costs, and service efficiency, to search for the optimal solution. During this search, the algorithm comprehensively considers customer needs and expectations in different logistics scenarios and incorporates the actual logistics resource availability (such as vehicle scheduling and personnel arrangements). It generates personalized service strategies, including prioritized delivery plans (such as allocating dedicated vehicles and planning optimal delivery routes for customers with urgent needs), dedicated customer service follow-up plans (designating dedicated customer service representatives to communicate with customers, provide real-time feedback on logistics progress, and address customer issues), customized logistics route adjustments (adjusting transportation routes based on real-time road conditions and cargo priority), and a list of recommended value-added services (such as cargo insurance and packaging reinforcement services for high-value customers). After multiple rounds of iterative optimization, the personalized service strategy that best meets these multiple objectives is ultimately determined.

[0078] In this embodiment, by real-time association of logistics status and customer portrait data, the dynamic needs of customers in the current logistics scenario can be accurately captured, and key logistics characteristics can be integrated into the portrait to make customer information more comprehensive and timely; based on the integrated data, the logistics scenario in which the customer is located is deeply analyzed, and combined with historical behavioral preferences and service demand levels, the customer's potential needs and expected services can be accurately predicted, providing precise guidance for personalized services; finally, the multi-objective particle swarm optimization algorithm is used to perform efficient iterative search and optimization on the integrated data, comprehensively considering multiple goals such as customer satisfaction, logistics costs, and service efficiency, and quickly generating personalized service strategies covering diversified content such as priority delivery, exclusive customer service, route customization, and value-added service recommendations, effectively improving customer experience and logistics operation efficiency, achieving the best match between customer needs and corporate resources, and enhancing the competitiveness and adaptability of logistics services.

[0079] See also Figure 6 A sixth embodiment of a customer management method according to an embodiment of the present invention includes:

[0080] S601. Monitor customer logistics data in real time, and determine in real time whether the customer logistics data meets abnormal conditions according to preset abnormal rules and thresholds.

[0081] In this embodiment, various exception rules and thresholds are pre-defined. For example, a delay exception threshold is defined as a shipment exceeding the estimated time by 30%, and an insufficient inventory exception rule is defined as an inventory level falling below the safety stock level. During monitoring, the system automatically compares real-time logistics data against these pre-defined rules and thresholds. Using algorithms such as data screening and logical analysis, the system analyzes the data in real time to determine if it meets the exception criteria. If data triggers an exception rule, the system immediately identifies the anomaly, saving time for subsequent processing.

[0082] S602. When the customer logistics data meets the abnormal conditions, determine the abnormal type of the customer logistics data, and match the corresponding personalized service strategy from the preset personalized service strategy library based on the abnormal type and dynamic customer profile.

[0083] In this embodiment, when an anomaly is confirmed, the system conducts an in-depth analysis of the customer's logistics data based on pre-defined anomaly classification standards to accurately determine the type of anomaly, such as transportation anomalies (transportation interruption, route deviation), warehousing anomalies (damaged goods, wrong or missed shipments), and distribution anomalies (delivery delays, inability to contact the recipient), etc. Subsequently, combined with the customer's dynamic customer portrait, the system takes into account factors such as the customer's service demand level, historical behavioral preferences, and special service requirements, and searches from the personalized service policy library. The policy library pre-stores a variety of personalized service strategies for different anomaly types and customer characteristics. The system uses similarity calculation, condition matching and other technologies to screen out the service strategies that best suit the current anomaly and customer needs. For example, when a transportation delay anomaly occurs for high-value and time-sensitive customers, priority is given to matching strategies that include expedited scheduling, dedicated follow-up, and compensation plans.

[0084] S603. Based on the matched personalized service strategy, initiate the corresponding customer service response process and clarify the responsible personnel, work content and time nodes of each link.

[0085] In this embodiment, after determining the matching personalized service strategy, the corresponding customer service response process is initiated. This process includes detailed planning for each step and clearly defined responsibilities, such as assigning customer service specialists to communicate and explain the situation to customers, coordinators to handle logistics resource allocation, and technical personnel to investigate the cause of the anomaly. The process also specifies work content, including that customer service specialists must contact customers within 30 minutes to explain the situation and that coordinators must develop alternative transportation plans within two hours. Timelines are also set, requiring an initial response within one hour of the anomaly and a solution within four hours.

[0086] In this embodiment, through real-time monitoring of customer logistics data and dynamic comparison with preset exception rules, abnormal situations in the logistics process can be quickly captured, and timely warnings of potential risks can be achieved; personalized service strategies are accurately matched based on the exception type and dynamic customer portrait, and the customer's historical behavioral preferences and service demand levels are fully considered to ensure the pertinence and effectiveness of the response strategy; standardized customer service response processes are initiated according to the matching strategies, and the responsible personnel, work content and time nodes of each link are clarified to ensure that the exception handling work is carried out efficiently and orderly, significantly improving the stability of logistics services and customer satisfaction, while reducing corporate operating risks and enhancing the intelligent management level and emergency handling capabilities of logistics services.

[0087] The above describes the customer management method in the embodiment of the present invention. The following describes the device in the embodiment of the present invention. Figure 7 , the implementation of the customer management device in the embodiment of the present invention includes:

[0088] A preprocessing module 701 is used to obtain multi-source heterogeneous logistics data and preprocess the multi-source heterogeneous logistics data to obtain preprocessed data;

[0089] A fusion module 702 is configured to fuse the pre-processed data using a trained multimodal data fusion model to obtain fused data;

[0090] A first generating module 703 is configured to extract features from the fused data using a machine learning algorithm based on a preset customer tag system, and generate a dynamic customer profile based on the extracted features;

[0091] The second generation module 704 is used to obtain the real-time logistics status, generate a personalized service strategy based on the dynamic customer profile and the real-time logistics status through a multi-objective optimization algorithm, and store it in a preset personalized service strategy library;

[0092] The calling module 705 is used to monitor customer logistics data in real time. When a preset abnormal situation occurs, the early warning mechanism is triggered and the corresponding personalized service strategy is matched from the personalized service strategy library to start the corresponding customer service response process.

[0093] In this embodiment, the preprocessing module 701 includes: a first acquisition unit 7011, used to acquire multi-source heterogeneous logistics data, the multi-source heterogeneous logistics data including order data, warehousing data, transportation trajectory data, distribution data and customer feedback data; a deletion unit 7012, used to identify and delete duplicate records in the multi-source heterogeneous logistics data based on the unique identification code of the multi-source heterogeneous logistics data using a hash algorithm to obtain non-duplicate data; a preprocessing unit 7013, used to perform data conversion and normalization on the non-duplicate data to obtain preprocessed data.

[0094] In this embodiment, the fusion module 702 includes: a construction unit 7021, which is used to construct a multimodal model based on the Transformer architecture and train the multimodal model to obtain a trained multimodal data fusion model; an association unit 7022, which is used to perform spatiotemporal alignment and semantic association on the preprocessed data to obtain associated data; and a fusion unit 7023, which is used to input the associated data into the trained multimodal data fusion model to obtain fused data.

[0095] In this embodiment, the first generation module 703 includes: an extraction unit 7031, which is used to perform unsupervised learning on the fusion data through a generative adversarial network to extract feature data; a matching unit 7032, which is used to match and map the extracted feature data with the labels in a preset customer label system; a first generation unit 7033, which is used to calculate the label value of the corresponding label for each customer based on the matching result, and according to the customer label system, weightedly fuse the calculated multiple label values ​​to generate a dynamic customer portrait for each customer, and store it in a preset personalized service strategy library.

[0096] In this embodiment, the second generation module 704 includes: an integration unit 7041, which is used to obtain real-time logistics status, associate the real-time logistics status with the corresponding dynamic customer portrait based on the logistics order number, and integrate the key features of the real-time logistics status into the dynamic customer portrait to obtain integrated data; a judgment unit 7042, which is used to analyze the customer's current logistics scenario based on the integrated data, and judge the customer's potential needs and expected services in the corresponding logistics scenario according to the historical behavioral preferences and service demand levels in the dynamic customer portrait; a second generation unit 7043, which is used to use a multi-objective particle swarm optimization algorithm to iteratively search and optimize the integrated data to generate a personalized service strategy, which includes a priority delivery plan, an exclusive customer service follow-up plan, a customized logistics route adjustment, and a value-added service recommendation list.

[0097] In this embodiment, the calling module 705 includes: a monitoring unit 7051, which is used to monitor customer logistics data in real time, and determine in real time whether the customer logistics data meets the abnormal conditions according to preset abnormal rules and thresholds; a calling unit 7052, which is used to determine the abnormal type of the customer logistics data when the customer logistics data meets the abnormal conditions, and match the corresponding personalized service strategy from the preset personalized service strategy library according to the abnormal type and dynamic customer portrait; a determination unit 7053, which is used to start the corresponding customer service response process according to the matched personalized service strategy, and clarify the responsible personnel, work content and time nodes of each link.

[0098] In this embodiment, by integrating multi-source heterogeneous logistics data and applying advanced data processing and intelligent algorithm technologies, the entire process of logistics customer management is intelligent and precise. First, the acquisition and preprocessing of multi-source heterogeneous data ensure the quality and availability of the original data, laying a solid foundation for subsequent analysis; the trained multimodal data fusion model breaks down data barriers, deeply mines the potential connections between data, and generates high-value fused data; secondly, the dynamic customer portrait constructed based on the preset label system and machine learning algorithm can reflect customer characteristics and demand changes in real time and comprehensively. Then, combined with the real-time logistics status, the multi-objective optimization algorithm is used to generate personalized service strategies, accurately match customer needs and enterprise resources, and improve service quality and operational efficiency. Finally, the real-time monitoring and intelligent early warning mechanism can promptly detect and respond to logistics anomalies, quickly retrieve response plans from the policy library, and ensure customer experience.

[0099] Figure 7 The structure of the customer management device shown does not constitute a limitation on the customer management device, and can implement the steps of the customer management method provided by the above-mentioned method embodiments.

[0100] above Figure 7 The client management apparatus in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The client management device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0101] Figure 8 : is a structural diagram of a customer management device provided by an embodiment of the present invention. The device 800 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 810 (for example, one or more processors) and a memory 820, and one or more storage media 830 (for example, one or more mass storage devices) for storing application programs 833 or data 832. Among them, the memory 820 and the storage medium 830 can be temporary storage or permanent storage. The program stored in the storage medium 830 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the device 800. Furthermore, the processor 810 can be configured to communicate with the storage medium 830 to execute a series of instruction operations in the storage medium on the device 800.

[0102] The device 800 may also include one or more power supplies 840, one or more wired or wireless network interfaces 850, one or more input and output interfaces 860, and / or one or more operating systems 831, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc.

[0103] An embodiment of the present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to execute the steps of the customer management method.

[0104] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0105] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc., various media that can store program code.

[0106] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A customer management method, characterized in that: The customer management method comprises: Acquiring multi-source heterogeneous logistics data, and preprocessing the multi-source heterogeneous logistics data to obtain preprocessed data; Using a trained multimodal data fusion model to fuse the preprocessed data to obtain fused data; Based on a preset customer tag system, feature extraction is performed on the fused data using a machine learning algorithm, and a dynamic customer profile is generated based on the extracted features; Acquire real-time logistics status, generate personalized service strategies through a multi-objective optimization algorithm based on the dynamic customer profile and the real-time logistics status, and store the strategies in a preset personalized service strategy library; Monitor customer logistics data in real time. When a preset abnormal situation occurs, trigger the early warning mechanism and match the corresponding personalized service strategy from the personalized service strategy library to start the corresponding customer service response process.

2. The customer management method according to claim 1, characterized in that: The acquiring of multi-source heterogeneous logistics data and preprocessing of the multi-source heterogeneous logistics data to obtain preprocessed data includes: Acquire multi-source heterogeneous logistics data, including order data, warehousing data, transportation trajectory data, distribution data, and customer feedback data; Based on the unique identification code of multi-source heterogeneous logistics data, a hash algorithm is used to identify and delete duplicate records in multi-source heterogeneous logistics data to obtain non-duplicate data; Data conversion and normalization are performed on non-repeated data to obtain preprocessed data.

3. The customer management method according to claim 1, characterized in that: The method of using the trained multimodal data fusion model to fuse the preprocessed data to obtain fused data includes: Constructing a multimodal model based on the Transformer architecture and training the multimodal model to obtain a trained multimodal data fusion model; Performing spatiotemporal alignment and semantic association on the preprocessed data to obtain associated data; The associated data is input into a trained multimodal data fusion model to obtain fused data.

4. The customer management method according to claim 3, characterized in that: The performing spatiotemporal alignment and semantic association on the pre-processed data to obtain associated data includes: Applying a time window matching algorithm to establish a time correspondence between different modal data in the preprocessed data to obtain time-aligned data; Converting spatial data in different coordinate systems in the time-aligned data into a unified coordinate system, and establishing associations between the data based on spatiotemporal constraints to obtain spatiotemporal aligned data; The semantic similarity of the different modal data in the spatiotemporal alignment data is calculated, and a semantic relationship is established based on a preset similarity threshold to obtain associated data.

5. The customer management method according to claim 1, characterized in that: The method of extracting features from the fused data using a machine learning algorithm based on a preset customer tag system and generating a dynamic customer profile based on the extracted features includes: Perform unsupervised learning on the fused data through generative adversarial networks to extract feature data; Match and map the extracted feature data with the tags in the preset customer tag system; The tag value of each customer's corresponding tag is calculated based on the matching results. According to the customer tag system, the calculated multiple tag values ​​are weighted and fused to generate a dynamic customer portrait for each customer and store it in the preset personalized service strategy library.

6. The customer management method according to claim 1, characterized in that: The real-time logistics status is obtained, and a personalized service strategy is generated through a multi-objective optimization algorithm based on the dynamic customer profile and the real-time logistics status, and stored in a preset personalized service strategy library, including: Acquire real-time logistics status, associate the real-time logistics status with the corresponding dynamic customer profile based on the logistics order number, and integrate key features of the real-time logistics status into the dynamic customer profile to obtain integrated data; Based on the integrated data, analyze the customer's current logistics scenario, and determine the customer's potential needs and expected services in the corresponding logistics scenario based on the historical behavioral preferences and service demand levels in the dynamic customer profile; A multi-objective particle swarm optimization algorithm is used to iteratively search and optimize the integrated data to generate a personalized service strategy, which includes a priority delivery plan, an exclusive customer service follow-up plan, a customized logistics route adjustment, and a recommended list of value-added services.

7. The customer management method according to claim 1, characterized in that: The real-time monitoring of customer logistics data, when a preset abnormal situation occurs, triggers an early warning mechanism and matches a corresponding personalized service policy from the personalized service policy library to start a corresponding customer service response process, including: Monitor customer logistics data in real time and determine whether the customer logistics data meets abnormal conditions in real time according to preset abnormal rules and thresholds; When customer logistics data meets the abnormal conditions, determine the abnormal type of the customer logistics data and match the corresponding personalized service strategy from the preset personalized service strategy library based on the abnormal type and dynamic customer profile; According to the matching personalized service strategy, initiate the corresponding customer service response process and clarify the responsible personnel, work content and time nodes of each link.

8. A customer management device, characterized in that: include: A preprocessing module is used to obtain multi-source heterogeneous logistics data and preprocess the multi-source heterogeneous logistics data to obtain preprocessed data; A fusion module, configured to fuse the pre-processed data using a trained multimodal data fusion model to obtain fused data; A first generation module is configured to extract features from the fused data using a machine learning algorithm based on a preset customer tag system, and generate a dynamic customer profile based on the extracted features; The second generation module is used to obtain real-time logistics status, generate personalized service strategies through a multi-objective optimization algorithm based on the dynamic customer profile and the real-time logistics status, and store the strategies in a preset personalized service strategy library; The calling module is used to monitor customer logistics data in real time. When a preset abnormal situation occurs, it triggers the early warning mechanism and matches the corresponding personalized service strategy from the personalized service strategy library to start the corresponding customer service response process.

9. A customer management device, characterized in that: comprising a memory and at least one processor, wherein the memory has computer-readable instructions stored therein; The at least one processor calls the computer-readable instructions in the memory to execute each step of the customer management method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer-readable instructions stored thereon, characterized in that: When the computer-readable instructions are executed by a processor, the steps of the customer management method according to any one of claims 1 to 7 are implemented.

Images