Customer complaint shop management and control method and device, equipment and medium

Abstract

The invention relates to a customer complaint shop management and control method and device, equipment and medium in the technical field of e-commerce, and the method comprises the steps: responding to an event that the transaction amount of a target shop reaches the standard, carrying out the statistics of the transaction amount of the target shop every day in a statistical time period to which the day belongs, and determining the corresponding statistical feature distribution; when the statistical feature distribution meets a short-term order blasting condition, obtaining a corresponding customer complaint estimation data set of the target shop at the current moment, including customer complaint estimation data corresponding to a plurality of estimation fields; calling a preset customer complaint estimation model to deduce a corresponding customer complaint risk degree according to the customer complaint estimation data set; and determining a customer complaint rating to which the customer complaint risk degree belongs, and executing a shop disposal process corresponding to the customer complaint rating on the target shop. According to the method and the device, the high customer complaint risk shops in the short-term explosive shops can be efficiently and accurately pre-judged, the high customer complaint risk shops are managed and controlled in time, unnecessary loss is avoided, and the customer complaint rate of the platform is greatly reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of e-commerce, and particularly relates to a customer complaint store management method and a corresponding device, computer equipment and computer readable storage medium. BACKGROUND

[0002] Under the impetus of today's digital wave, the e-commerce industry is booming, with transaction size continuously rising, becoming an important part of the global retail market. However, with the increasing frequency of e-commerce transactions, disputes between customer complaints and store transactions have become increasingly prominent, which has become a key bottleneck and potential risk that restricts the construction of a healthy e-commerce platform ecosystem and sustainable development.

[0003] For e-commerce platforms, the negative impact of customer complaint stores is all-round and far-reaching. On the one hand, customer complaint stores directly impact the shopping experience of consumers, causing consumers to encounter setbacks and dissatisfaction during the shopping process, which in turn leads to a significant decline in consumer trust in the platform. Once trust is damaged, the purchase rate and repeat rate of consumers are difficult to maintain, and when consumers cannot get timely and proper solutions to problems, they will not hesitate to turn to other e-commerce platforms to seek better shopping experiences, which poses a serious challenge to the long-term user retention of the platform. The increase in consumer attrition rate means that the platform needs to invest more resources to attract new users, increasing the cost of acquiring customers and affecting the economic benefits and market share of the platform.

[0004] On the other hand, customer complaint stores have a devastating impact on the reputation of the platform. In today's information dissemination is extremely rapid, consumer dissatisfaction can spread widely through social media, online forums, online reviews and other channels in an instant. A negative review can trigger a chain reaction, quickly spreading among potential user groups, thereby building a negative impression of the platform in the minds of potential users, leading to potential users stepping back and losing a large number of potential customer resources. In the long run, it will seriously erode the brand value accumulated by the platform over the long term, weaken the core competitiveness of the platform in fierce market competition, and hinder the long-term development of the platform.

[0005] In the traditional complaint handling mode, the e-commerce platform mainly relies on a large number of human resources, and adopts manual docking communication to audit and handle the complaint shop. The customer service personnel need to communicate with the customer and the shop one by one, understand the details of the problem, verify the situation, and then judge and handle according to the platform rules. This mode is not only inefficient, but also difficult to deal with a large number of complaint requests, and the operation cost is high. With the increasing number of complaints, the size of the customer service team also needs to be expanded, and the labor cost increases linearly, which brings a heavy burden to the platform. At the same time, manual processing also has many uncertainties, and the professional quality, processing experience and efficiency of different customer service personnel are different, which may lead to different effects of complaint handling, affecting the fairness, accuracy and efficiency of the processing, further intensifying the dissatisfaction of consumers, and falling into a vicious circle.

[0006] In view of the shortcomings of the traditional complaint handling technology, the applicant has been deeply engaged in the e-commerce field for a long time, and has been committed to solving the industry problems. After in-depth research and practice exploration, a new path is taken to explore a more efficient, intelligent, low-cost and fair platform operation scheme for handling complaint shops, so as to help e-commerce platforms break through the development bottleneck, reshape a healthy ecology, improve user experience and platform operation efficiency, and promote the healthy development of the e-commerce industry. SUMMARY

[0007] The primary purpose of the present application is to solve at least one of the above problems and provide a complaint shop management method and its corresponding device, computer equipment and computer program product.

[0008] To achieve the various purposes of the present application, the present application adopts the following technical solutions:

[0009] A complaint shop management method provided to adapt to one of the purposes of the present application, comprising the following steps:

[0010] In response to the transaction amount of the target shop reaching the standard event on the same day, the transaction amount of the target shop in the statistical period on the same day is counted to determine the corresponding statistical feature distribution;

[0011] When the statistical feature distribution meets the short-term explosive order condition, the complaint estimation data set corresponding to the target shop at the current time is obtained, and the complaint estimation data set includes the complaint estimation data corresponding to a plurality of estimation fields;

[0012] The preset complaint estimation model is called to infer the corresponding complaint risk degree according to the complaint estimation data set;

[0013] Determine the complaint rating to which the complaint risk degree belongs, and execute the shop disposal process corresponding to the complaint rating on the target shop.

[0014] In another aspect, a complaint store management device provided for one of the purposes of the present application includes an event response module, a data acquisition module, a complaint estimation module, and a store handling module. The event response module is configured to respond to a target store's daily transaction amount reaching a standard event, count the target store's daily transaction amount in a statistical period of the day, and determine a corresponding statistical feature distribution. The data acquisition module is configured to acquire a complaint estimation dataset corresponding to the target store at a current time when the statistical feature distribution meets a short-term explosive order condition, the complaint estimation dataset including complaint estimation data corresponding to a plurality of estimation fields. The complaint estimation module is configured to call a preset complaint estimation model to infer a corresponding complaint risk degree based on the complaint estimation dataset. The store handling module is configured to determine a complaint rating to which the complaint risk degree belongs, and execute a store handling process corresponding to the complaint rating on the target store.

[0015] In yet another aspect, a computer device provided for one of the purposes of the present application includes a central processing unit and a memory. The central processing unit is configured to call a computer program stored in the memory to execute the steps of the complaint store management method described in the present application.

[0016] In yet another aspect, a computer program product provided for another purpose of the present application includes a computer program / instruction that, when executed by a processor, implements the steps of the method described in any one of the embodiments of the present application.

[0017] The technical solution of the present application has multiple advantages, including but not limited to the following aspects:

[0018] The present application responds to a target store's daily transaction amount reaching a standard event, counts the target store's daily transaction amount in a statistical period of the day, determines a corresponding statistical feature distribution, and judges whether a short-term explosive order condition is met. It can be seen that the store with a sharp increase in transaction orders in a short period of time can be accurately screened out. Once such a store is found, the complaint estimation dataset of the short-term explosive order store is further acquired, ensuring the subsequent estimation of the complaint risk of such a store, so that the platform can concentrate limited resources and efforts on the stores that are most likely to raise the complaint rate of the platform and can raise the complaint rate of the store by a large margin, avoiding the situation that the store is overwhelmed by too many transactions in a short period of time compared to other stores, thereby making it easier to generate and more customer complaints.

[0019] After determining that the target store is a short-term hot-selling store, immediately obtain its corresponding customer complaint prediction data set at the current moment. This data set covers customer complaint prediction data corresponding to multiple estimation fields. Then call the preset customer complaint prediction model to quickly infer the corresponding customer complaint risk based on these data. It can be seen that real-time and rapid evaluation of customer complaint risks is achieved, which greatly shortens the time from abnormal store transaction volume to customer complaint risk identification. In the past, it may take a long time to collect and analyze a lot of data, and these data may even contain redundant data. In comparison, this method uses pre-built models and real-time data acquisition of corresponding evaluation fields to quickly derive accurate customer complaint risk using necessary data in a short period of time, so that the platform can promptly discover potential high-customer complaint stores and gain more valuable time for subsequent implementation of corresponding disposal measures.

[0020] Based on the inferred customer complaint risk, the platform determines the customer complaint rating and executes the corresponding store handling process accordingly. This tiered approach allows the platform to take varying degrees of action based on the severity of the store's customer complaint risk. This approach can quickly minimize the negative impact of high-complaint stores and prevent the generation of large numbers of customer complaint orders. This effectively reduces the sharp increase in platform customer complaint rates caused by high-complaint stores, mitigates damage to the platform's reputation and user experience, maintains the platform's normal operations, and enhances its overall competitiveness.

[0021] To sum up, the customer complaint store management method of the present invention realizes the effective management of high customer complaint risk stores among short-term explosive stores through the three key links of accurate identification, rapid evaluation and efficient disposal, significantly improves the identification timeliness and disposal efficiency of customer complaint stores, effectively guarantees the stable operation of the platform, and is of great significance to the long-term development of the platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0023] Figure 1 The network architecture of the e-commerce platform exemplified in this application;

[0024] Figure 2 This is a flowchart of a typical embodiment of the customer complaint store management method of the present application;

[0025] Figure 3 This is a functional block diagram of the customer complaint store management device of this application;

[0026] Figure 4 This is a schematic diagram of the structure of a computer device used in this application. DETAILED DESCRIPTION

[0027] Embodiments of the present application are described below in detail with reference to the accompanying drawings, examples of which are shown in the drawings, wherein the same or similar notations represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the accompanying drawings are exemplary and are for the purpose of explanation only, and are not to be construed as limiting the present application.

[0028] It should be understood that, as used here, the singular form "a", "an" and "the" include plural references unless the context clearly dictates otherwise. It should be further understood that the word "comprise" or "comprises" when used in this specification is taken to mean the presence of stated features, integers, steps, operations, elements, or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or groups thereof. It should be understood that when we say an element is "connected" or "coupled" to another element, it can be directly connected or coupled to the other element, or intervening elements can be present. In addition, the use of "connected" or "coupled" here includes wireless connection or wireless coupling. The phrase "and / or" as used herein includes all possible combinations of one or more of the associated listed items and all combinations of the items.

[0029] It should be understood that, as used here, the singular form "a", "an" and "the" include plural references unless the context clearly dictates otherwise. It should be further understood that the word "comprise" or "comprises" when used in this specification is taken to mean the presence of stated features, integers, steps, operations, elements, or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or groups thereof. It should be understood that when we say an element is "connected" or "coupled" to another element, it can be directly connected or coupled to the other element, or intervening elements can be present. In addition, the use of "connected" or "coupled" here includes wireless connection or wireless coupling. The phrase "and / or" as used herein includes all possible combinations of one or more of the associated listed items and all combinations of the items.

[0030] As Figure 1 In the network architecture as shown, the e-commerce platform 82 is deployed in the Internet to provide corresponding services to its users, and the devices 80 of the merchant users and the devices 81 of the consumer users of the e-commerce platform 82 are also connected to the Internet to use the services provided by the e-commerce platform.

[0031] An exemplary e-commerce platform 82 provides a supply and demand matching of products and / or services to the public through the Internet infrastructure, in which the products and / or services are provided as commodity information. For the purpose of simplifying the description, the concepts of commodity, product, etc. are used in the present application to refer to the products and / or services in the e-commerce platform 82, which can be physical products, digital products, tickets, service subscriptions, other offline services, etc.

[0032] Real-world entities can access the e-commerce platform 82 in the identity of users to use various online services provided by the e-commerce platform 82 to achieve the purpose of participating in the business activities implemented by the e-commerce platform 82. These entities can be natural persons, legal persons, or social organizations, etc. Corresponding to the two types of entities of merchants and consumers in the business activities, the e-commerce platform 82 correspondingly exists two types of users of merchant users and consumer users. The entities of the product circulation chain in the business activities, including manufacturers, sellers, retailers, logistics providers, etc., can use online services in the e-commerce platform 82 in the identity of merchant users, while the consumers in the business activities, including real or potential consumers, can use online services in the e-commerce platform 82 in the identity of their corresponding consumer users. In actual business activities, the same entity can act as a merchant user and a consumer user, and flexible and variable understanding should be made accordingly.

[0033] The infrastructure for deploying the e-commerce platform 82 mainly includes a backend architecture and front-end devices. The backend architecture runs various online services through a service cluster, including middleware or front-end services for platform parties, services for consumers, services for merchants, etc., to enrich and perfect its service functions. The front-end devices mainly cover terminal devices used by users as clients to access the e-commerce platform 82, including but not limited to various mobile terminals, personal computers, point-of-sale devices, etc. For example, a merchant user can use his terminal device 80 to input product information for his online store, or use the interface provided by the e-commerce platform to generate product information; a consumer user can access the web page of the online store implemented by the e-commerce platform 82 through his terminal device 81, trigger the shopping process through the shopping button provided on the web page, and call various online services provided by the e-commerce platform 82 in the shopping process, so as to achieve the purpose of shopping and ordering.

[0034] In some embodiments, the e-commerce platform 82 can be implemented by a processing facility including a processor and a memory, which stores a set of instructions that, when executed, cause the e-commerce platform 82 to perform the electronic commerce and support functions involved in this application. The processing facility can be part of a server, a client, a network infrastructure, a mobile computing platform, a cloud computing platform, a fixed computing platform, or other computing platforms, and provides electronic components of the e-commerce platform 82, merchant devices, payment gateways, application developers, marketing channels, transportation providers, customer devices, point-of-sale devices, etc.

[0035] The e-commerce platform 82 can be implemented as a cloud computing service, a software as a service (SaaS), an infrastructure as a service (IaaS), a platform as a service (PaaS), a desktop as a service (DaaS), a hosted software as a service, a mobile backend as a service (MBaaS), an information technology management as a service (ITMaaS), and the like online services. In some embodiments, various functional components of the e-commerce platform 82 can be implemented to be suitable for operation on various platforms and operating systems, for example, an administrator user of an online store enjoys the same or similar functions in various embodiments such as iOS, Android, HomonyOS, or web page, and the like.

[0036] The e-commerce platform 82 can implement respective independent stations for various merchants to run their respective online stores, and provide respective business management engine instances for the merchants to establish, maintain, and run one or more online stores in one or more independent stations. The business management engine instance can be used for content management, task automation, and data management of one or more online stores, and can configure various specific business processes of the online store through an interface or a built-in component to support the implementation of business activities. The independent station is the infrastructure of the e-commerce platform 82 with cross-border service functions, and the merchant can maintain its online store based on the independent station in a relatively centralized and autonomous manner. The independent station usually has a domain name and storage space dedicated to the merchant, and different independent stations have relative independence. The e-commerce platform 82 can provide standardized or personalized technical support for a large number of independent stations, so that the merchant user can customize a business management engine instance suitable for itself, and use this business management engine instance to maintain one or more online stores owned by the merchant.

[0037] The online store can be configured and maintained by the merchant user logging in as an administrator to its business management engine instance. With the support of various online services provided by the infrastructure of the e-commerce platform 82, the merchant user can configure various functions in its online store and review various data as an administrator, for example, the merchant user can manage various aspects of its online store, such as viewing recent activities of the online store, updating product catalog of the online store, managing orders, recent visit activities, total order activities, and the like; the merchant user can also view more detailed information about business and visitors to the online store of the merchant by obtaining reports or metrics, such as displaying sales summary of the overall business of the merchant, specific sales and engagement data of the activity sales marketing channel, and the like.

[0038] The e-commerce platform 82 can provide a communication facility and associated merchant interface for providing electronic communications and marketing, such as collecting and analyzing communication interactions between merchants, consumers, merchant devices, customer devices, point-of-sale devices, etc. using an electronic message aggregation facility, aggregating and analyzing communications, such as for increasing the potential for providing product sales, etc. For example, a consumer can have a question about a product that can generate a conversation between the consumer and a merchant (or an automated processor-based agent representing the merchant), where the communication facility is responsible for the interaction and provides the merchant with analysis on how to improve the probability of a sale.

[0039] In some embodiments, an application suitable for installation to a terminal device can be provided to serve the access needs of different users, so as to enable various users to access the e-commerce platform 82, such as a merchant back-end module of an online store in the e-commerce platform 82, etc. in the terminal device by running the application. In the process of implementing business activities through these functions, the e-commerce platform 82 can implement various functions related to the implementation of business activities as middleware or online services and open corresponding interfaces, and then implant the interface access functions into the application to realize function extension and task implementation. The business management engine can include a series of basic functions, and expose these functions to online services and / or application calls through APIs. Online services and applications use corresponding functions by remotely calling corresponding APIs.

[0040] With the support of various components of the business management engine instance, the e-commerce platform 82 can provide online shopping functions, enabling merchants to establish contact with customers in a flexible and transparent manner, and enabling consumer users to select and purchase goods online, create a product order, provide a delivery address for the goods in the product order, and complete payment confirmation of the product order. Then, the merchant can review and complete or cancel the order.

[0041] The complaint store management method of the present application can be programmed as a computer program product, deployed in a client or server for running and implementation, for example, in the exemplary application scenario of the present application, it can be implemented in the server of the e-commerce customer service platform. Therefore, through the interface opened after the running of the computer program product, human-computer interaction can be performed with the process of the computer program product through the graphical user interface to execute the method.

[0042] Please refer to Figure 2 In a typical embodiment of the complaint store management method of the present application, the following steps are included:

[0043] Step S1100, in response to the daily transaction amount of the target store reaching the standard event, counting the daily transaction amount of the target store in the statistical period to which the day belongs, and determining the corresponding statistical feature distribution;

[0044] For each shop in the e-commerce platform, the daily transaction amount of the shop can be monitored in real time, and then whether the daily transaction amount exceeds the large transaction volume is determined. When the daily transaction amount exceeds the large transaction volume, it means that the daily transaction amount of the shop at this time meets the standard, which is represented as a sudden increase in orders in this day, that is, a sharp increase in transaction orders. Therefore, the shop is taken as a target shop, and a transaction amount meeting event corresponding to the target shop is responded. In this process, the date of the day and the corresponding preset reference time point are used to expand forward and backward on the time axis, so as to determine a complete statistical period. Then, the statistical period is divided into two parts: an earlier period and a later period, with the reference time point as the boundary. Then, the statistical characteristics of the daily transaction amount of the target shop in the two sub-periods are calculated respectively, and the statistical characteristics are integrated into a unified statistical characteristic distribution according to the time sequence of the corresponding period.

[0045] In an embodiment, the large transaction volume can be obtained by multiplying the mean value of the daily transaction amount of the corresponding shop in the time period by the addition multiple, and the addition multiple is greater than one. The time period and the addition multiple can be set by those skilled in the art as needed, for example, 1.5 times of the annual level.

[0046] The reference time point refers to the specific day marked in chronological order within a period of time corresponding to the date, which can be set by those skilled in the art as needed, for example, the seventh day within ten days. The statistical characteristics include any one or any combination of the following: mean value, standard deviation, ratio of days meeting the large transaction volume to total days, total number of days, total transaction amount, and values obtained by performing four arithmetic operations on any combination of the above.

[0047] For ease of understanding, an exemplary example is given. In the e-commerce platform, when the daily transaction amount of a certain shop exceeds 1.5 times of the daily level of the shop, the date of the day and the corresponding reference time point are used to trace back 7 days (including the day) and extend 3 days (including the day), forming a statistical period of 10 days. Then, according to the daily transaction amount in the previous 7 days, the mean value and the standard deviation are calculated, and the ratio of the standard deviation to the mean value is taken as the transaction fluctuation coefficient. According to the daily transaction amount in the next 3 days, the ratio of the number of days whose transaction amount exceeds the large transaction volume to the total number of days is determined as the large transaction concentration degree. According to the time sequence, the statistical characteristics are integrated, and then the distribution characteristics of "fluctuation first and concentration later" are integrated according to the time sequence, which are taken as the statistical characteristic distribution.

[0048] In step S1200, when the statistical characteristic distribution meets the short-term order explosion condition, the target shop corresponding to the current time is obtained. The complaint estimation data set includes multiple estimated fields corresponding to the complaint estimation data.

[0049] When the statistical feature distribution hits the short-term explosive order condition, it means that the statistical feature distribution meets the short-term explosive order condition, and the target store is confirmed to be a short-term explosive order store. It can be understood that the short-term explosive order has a certain duration, so the transaction volume of the target store in the future will be more than usual. In order to avoid huge losses caused by frequent customer complaints in the future, it is necessary to estimate the complaint risk of the target store. Accordingly, the complaint estimation dataset corresponding to the target store at the current time is triggered. Specifically, in order to accurately estimate the complaint risk of the store, a plurality of key estimation fields are pre-selected. These estimation fields cover various important information of the store during the transaction process. By collecting and analyzing the data corresponding to each estimation field, the core basis for complaint estimation is obtained, thereby achieving effective estimation of customer complaints about the store. Therefore, for each estimation field, the data corresponding to the estimation field of the target store at the current time is obtained as the complaint estimation data. Finally, the complaint estimation data corresponding to each estimation field is collected to form a complete complaint estimation dataset.

[0050] The short-term explosive order condition includes a judgment condition for each statistical feature in the corresponding statistical feature distribution, which represents the transaction performance of the store in the statistical period as a short-term explosive order performance.

[0051] Step S1300, calling a pre-set complaint estimation model to infer the corresponding complaint risk degree according to the complaint estimation dataset;

[0052] The complaint estimation model is a model trained to reach a convergence state and suitable for a binary classification task. Its core learning goal is to identify and induce specific feature patterns hidden in the input complaint estimation dataset. These patterns have been verified by historical data that the store has explosive growth transactions in the short term, and the relationship between the store and the complaint risk. Training convergence means that the model has learned through a large number of training samples, and the internal parameters have been adjusted to stably and effectively capture these potential correlations, thereby possessing the ability to reliably infer risks from new input data. Therefore, the complaint estimation model can use the neuron parameters learned during training and stored internally to perform layer-by-layer, nonlinear transformation and combination on the complaint estimation dataset as input, aiming to extract and highlight the key feature patterns that best distinguish between "the store will have frequent complaints in the short term" and "the store will not have frequent complaints in the short term". Finally, the complaint risk degree is output in the output layer. For a binary classification task, the risk degree is usually mapped to a probability value between 0 and 1, which represents the likelihood of the store having frequent complaints in the short term. The higher the likelihood, the higher the complaint risk.

[0053] The specific selection of the complaint estimation model includes but is not limited to: Gradient Boosting Decision Trees (GBDT), such as XGBoost, LightGBM, CatBoost, etc. This kind of model is good at processing structured table data, can effectively capture the interaction and combination of features, has relatively low dependence on feature engineering, fast training speed, high prediction accuracy, and is very suitable for binary classification tasks in e-commerce risk control scenarios. Deep Neural Networks (DNN), especially structured Multi-Layer Perceptron (MLP). DNN can automatically learn complex nonlinear relationships and is suitable for high-dimensional features and deep interactions between features. Ensemble Models, such as Random Forest or the hybrid strategy of GBDT and DNN (Stacking / Blending), which combines multiple base models to improve generalization ability and robustness. Those skilled in the art can select appropriate model architectures to implement this binary classification task based on performance, efficiency, interpretability, data size, and other factors.

[0054] Step S1400, determining the complaint rating to which the complaint risk degree belongs, and executing the shop disposal process corresponding to the complaint rating on the target shop.

[0055] In an embodiment, different levels of complaint ratings and their corresponding value ranges can be pre-set to measure the risk level of the currently determined complaint risk degree. These complaint ratings and their corresponding value ranges can be divided by those skilled in the art as needed, for example, low risk, medium risk, and high risk, corresponding to the interval [0, 0.3), (0.3, 0.8), (0.8, 1]. Thus, the complaint rating corresponding to the value range to which the currently determined complaint risk degree belongs is determined as the complaint rating to which the complaint risk degree belongs.

[0056] In a further embodiment, step S1400, executing the shop disposal process corresponding to the complaint rating on the target shop, includes the following steps:

[0057] Step S1410, when the complaint rating is a level representing frequent complaints in the short term, the transaction permission of the target shop is removed;

[0058] When the determined complaint rating belongs to the highest level among all complaint ratings, it means that the customer service risk degree is too high at this time, and the target store is about to have frequent complaints in the short term, which is manifested as extremely high complaint frequency and extremely large total amount of complaints, far exceeding the level of normal operation state of the store in the platform, causing serious threat to the normal transaction order and reputation of the platform, and immediate measures need to be taken to remove the transaction permission of the target store, so that the target store cannot have any transaction with customers afterwards to prevent further negative impact.

[0059] Step S1420, when the complaint rating is the level representing intermittent complaints in the short term, observing the complaint performance of the target store in a preset time length, so as to re-determine the complaint rating according to the complaint performance;

[0060] When the determined complaint rating does not belong to the highest level or the lowest level among all complaint ratings, it means that the customer service risk degree is in an uncertain range at this time, and the target store is about to have intermittent complaints in the short term, which is manifested as complaint frequency and total amount of complaints being higher than the level of normal operation state of the store in the platform to a certain extent, but not far beyond, and there is a certain volatility, so that the complaint risk level of the target store needs to be further determined by observing its subsequent performance, so as to take corresponding reasonable measures, avoid unnecessary economic losses caused by unreasonable business intervention on the target store, and also reserve a certain doubt, so that the target store can be disposed of in time once it is found to be far beyond the level.

[0061] Step S1430, when the complaint rating is the level representing regular complaints in the short term, the target store does not need to be disposed of.

[0062] When the determined complaint rating belongs to the lowest level among all complaint ratings, it means that the customer service risk degree is too low at this time, and the target store is about to have regular complaints in the short term, which is manifested as low complaint frequency and low total amount of complaints, tending to the normal operation state of the store in the platform. Therefore, the target store does not need to be disposed of and can continue normal operation.

[0063] Through the disclosure of the typical embodiments and their variations, it can be understood that the present application contains many positive advantages, including but not limited to the following aspects:

[0064] The present application can accurately screen out shops with a sharp increase in transaction orders in a short period of time. Once such shops are found, further acquisition of the complaint estimation dataset of the short-term explosive shop is triggered, ensuring the subsequent estimation of the complaint risk of such shops, enabling the platform to focus limited resources and efforts on shops that are most likely to raise the platform's complaint rate and shops that can raise the complaint rate by a large margin, avoiding shops that may be overwhelmed by a large number of transactions relative to other shops, and thus more likely to produce and more customer complaints.

[0065] After determining that the target shop is a short-term explosive shop, the corresponding complaint estimation dataset of the target shop at the current time is immediately acquired, which covers the complaint estimation data corresponding to multiple estimation fields. Then, a preset complaint estimation model is called to quickly infer the corresponding complaint risk degree based on these data. It can be seen that real-time and rapid evaluation of complaint risk is achieved, greatly shortening the time consumption from abnormal transaction volume of the shop to identification of complaint risk. In the past, it may take a long time to collect and analyze a large amount of data, and even the data contains redundant data. Compared with the foregoing, the present method can quickly obtain an accurate complaint risk degree by using necessary data in a short time with the aid of a pre-constructed model and real-time data acquisition of corresponding evaluation fields, so that the platform can discover potential high-complaint shops in time, and more valuable time can be obtained for subsequent execution of corresponding disposal measures.

[0066] According to the inferred complaint risk degree, the complaint rating of the target shop is determined, and the corresponding shop disposal process is executed. This disposal method based on complaint risk degree classification enables the platform to take different disposal measures according to the severity of the complaint risk of the shop, quickly controls the negative impact of high-complaint shops within the minimum range, prevents a large number of complaint orders, effectively reduces the sharp increase in the platform's complaint rate caused by high-complaint shops, reduces the damage to the reputation and user experience of the platform, maintains the normal operation order of the platform, and improves the overall competitiveness of the platform.

[0067] In summary, the complaint shop management method of the present application realizes effective management and control of high-complaint risk shops in short-term explosive shops through three key links of accurate identification, rapid evaluation and efficient disposal, significantly improves the identification timeliness and disposal efficiency of complaint shops, and effectively guarantees the stable operation of the platform, which is of great significance for the long-term development of the platform.

[0068] In a further embodiment, step S1100, in response to the daily transaction amount of the target store meeting the threshold event, the daily transaction amount of the target store in the statistical period is counted, and the corresponding statistical feature distribution is determined, including the following steps:

[0069] Step S1110, monitoring the daily transaction amount of the target store, and determining whether the daily transaction amount exceeds the preset large transaction volume;

[0070] In the vast store system of the e-commerce platform, the daily transaction amount of each store fluctuates differently. In order to accurately screen out the stores that may be complained by customers in a short period of time, it is necessary to pay close attention to the transaction dynamics of each store and focus on the daily transaction amount change.

[0071] The large transaction volume is used to measure whether the daily transaction amount of the store has suddenly increased, so as to determine whether to trigger a series of processes aimed at determining whether the store belongs to a short-term explosive store. The skilled person in the art can set it as needed.

[0072] In one embodiment, the portraits of each store in the e-commerce platform can be pre-set, and a clustering algorithm is used to cluster all store portraits to determine a plurality of clusters, so that each cluster includes at least one store portrait belonging to the cluster, that is, each store portrait has its own cluster. For each cluster, the daily transaction amount of each store corresponding to the portrait in the same time period is obtained, and the average transaction amount of the store in the time period is calculated in units of stores, and then the average value corresponding to the average transaction amount of all stores is obtained to obtain the large transaction volume.

[0073] In another embodiment, the large transaction volume can be set by the skilled person in the art according to business experience or data analysis, for example, 1000 US dollars.

[0074] If the daily transaction amount of the target store does not exceed the large transaction volume, it is directly determined that the target store does not belong to a short-term explosive store.

[0075] Step S1120, when the transaction amount exceeds the large transaction volume, the statistical period is determined along the time axis bidirectionally from the date corresponding to the transaction amount and the reference time point corresponding thereto;

[0076] At this time, the target store shows a short-term explosive sales trend. To this end, further analysis of the target store is needed. The transaction performance of the target store in the statistical period of the day ensures reliable confirmation of whether the target store belongs to a short-term explosive sales store. The corresponding date of the day when the transaction amount exceeds the large transaction volume is set in advance. The specific number of days in a period of time is sequentially labeled as a reference time point. Those skilled in the art can set it as needed according to the disclosure, for example, the seventh day in ten days. In order to ensure the interpretability and rationality of the transaction performance, the reference time point should be close to the last day in the period of time.

[0077] It is not difficult to understand that the date of the day and its corresponding reference time point are bidirectionally expanded along the time axis. In the corresponding period of time, the date of the first day is respectively traced back to the date of the last day, thereby forming a statistical period. For ease of understanding, an exemplary embodiment is provided. The date corresponding to the day when the transaction amount exceeds the large transaction volume is October 7th. The corresponding reference time point is the seventh day in ten days. Therefore, the statistical period is [October 1st, October 10th].

[0078] Step S1130, dividing the statistical period into a preceding period and a subsequent period based on the reference time point;

[0079] It is not difficult to understand that the period between the date of the first day in the statistical period and the date corresponding to the reference time point can be divided into a preceding period, which is located before the date corresponding to the reference time point. Meanwhile, the period between the date corresponding to the reference time point and the date of the last day in the statistical period can be divided into a subsequent period, which is located after the date corresponding to the reference time point.

[0080] Step S1140, determining the statistical characteristics of the preceding period and the subsequent period based on the transaction amount of the target store in the statistical period;

[0081] In an embodiment, for the preceding period, the average value of the transaction amount of the target store in the period is calculated as the statistical characteristic of the preceding period. For the subsequent period, the total transaction amount of the sum of the transaction amount of the target store in the period is calculated. In addition, the total number of days when the transaction amount exceeds the large transaction volume is calculated. The total transaction amount and the total number of days are calculated as the statistical characteristics of the subsequent period.

[0082] Step S1150, integrating the statistical characteristics in sequence to obtain a statistical characteristic distribution.

[0083] According to the time sequence, the statistical features of the preceding period and the statistical features of the subsequent period are integrated into a statistical feature distribution. It is not difficult to understand that, since the statistical feature distribution is arranged in sequence, the short-term explosive order condition can be configured in advance, so that the statistical feature distribution and the short-term explosive order condition can be matched directly and quickly. For ease of understanding, by way of example, the statistical feature distribution is [average value: 729 US dollars; total transaction amount: 5354 US dollars; number of days when the transaction amount exceeds the large transaction volume: 3 days], and the corresponding short-term explosive order condition is [average value < 750 US dollars; total transaction amount > 3000 US dollars; number of days when the transaction amount exceeds the large transaction volume > 2 days]

[0084] In this embodiment, first, the transaction amount of the target store is monitored every day, and it is judged whether it exceeds the preset large transaction volume, which helps to accurately select the store that may have an explosive order in the short term among numerous e-commerce platform stores, and provides a clear direction for subsequent in-depth analysis. Once the transaction amount exceeds the large transaction volume, the date corresponding to the transaction amount and the reference time point are immediately determined as the basis for expanding along the time axis in both directions to determine the statistical period, which can ensure comprehensive investigation of the transaction performance of the target store within a certain time range, avoiding missing potential transaction trends by only focusing on single-day transaction data. Then, the statistical period is divided into a preceding period and a subsequent period based on the reference time point, and the statistical features of the two periods are determined respectively, and finally integrated into a statistical feature distribution in time sequence. Accordingly, the platform can intuitively understand the changes in the transaction characteristics of the target store before and after the short-term explosive order, providing a strong basis for judging whether the short-term explosive order condition is met, and thus achieving accurate identification of the short-term explosive order store, laying a solid foundation for subsequent complaint risk assessment.

[0085] In further embodiments, after the transaction permission of the target store is removed when the complaint rating is a level representing frequent complaints in the short term, the following steps are included:

[0086] Step S1411, determining a target evaluation explanation text in the set of preset complaint evaluation strategies that matches the complaint estimation data set;

[0087] The set of complaint evaluation strategies includes a plurality of evaluation explanation texts, which are used to explain in detail the specific reasons for the complaint risk degree corresponding to the level representing frequent complaints in the short term predicted by the evaluation, including a plurality of judgment conditions corresponding to the complaint estimation data set. Those skilled in the art can manually set the set of complaint evaluation strategies according to business experience, or prepare the set of complaint evaluation strategies according to the disclosure of the subsequent embodiments.

[0088] The complaint estimation data set is matched with each evaluation explanation text in the set of complaint evaluation strategies to determine the evaluation explanation text hit by the complaint estimation data set as the target evaluation explanation text.

[0089] Step S1412, according to the complaint risk degree, complaint rating and target evaluation explanation text, an evaluation report is constructed, and the evaluation report is pushed to the target store.

[0090] The complaint risk degree, complaint rating and target evaluation explanation text obtained are integrated into the evaluation report in the preset report template. The report template organizes multiple aspects of information through coherent and explanatory language expressions, ensuring the readability of the report, which can be implemented as needed by those skilled in the art.

[0091] In this embodiment, by further determining the target evaluation explanation text matched with the complaint estimation dataset after determining that the complaint rating is the short-term frequent complaint level and canceling the transaction permission of the target store, and constructing the evaluation report and pushing it to the target store, the specific reasons for the complaint risk can be analyzed in depth, and clear and explicit information can be provided for the platform and the store. In addition, the platform can take more targeted measures to rectify or optimize the store according to the detailed explanation in the evaluation report, and it is also helpful for the store itself to understand the problem and actively cooperate with the platform to improve, thereby effectively reducing the complaint risk and maintaining the transaction order and reputation of the platform.

[0092] In further embodiments, before step S1410, the target evaluation explanation text matched with the complaint estimation dataset is determined from the preset complaint evaluation strategy set, the step includes the following steps:

[0093] Step S14111, the complaint risk degree of the training sample in the preset training set is inferred by using the preset complaint estimation model, and the complaint risk degree is supplemented to the training sample;

[0094] The training set contains multiple training samples and their supervision labels. The training sample contains historical operating data corresponding to multiple dimensions of the short-term single store. The supervision label of the training sample represents whether the complaint performance of the short-term single store in its corresponding short-term single operating period meets the high complaint condition.

[0095] The training set is used to supervise and train the complaint estimation model in advance, and the complaint estimation model is trained to a convergent state, so that the complaint estimation model learns the ability to infer and output the complaint risk degree according to the input training sample in the training sample.

[0096] By supplementing the complaint risk degree inferred by the complaint estimation model to the training sample, the data dimension of the training sample can be enriched, which provides an important basis for subsequent strategy mining and lays a foundation for constructing a more accurate and more explanatory complaint evaluation strategy set.

[0097] Step S14112, the complaint analysis model is trained to a convergent state by using the supplemented training set;

[0098] Customer complaint analysis models can be selected from decision tree models such as GBDT, Random Forest, and LightGBM. The training goal of a customer complaint analysis model is to use a tree structure to learn the association rules between feature combinations in supplemented training samples and customer complaint risks, representing the judgment criteria. The customer complaint analysis model generates nodes by recursively splitting data. Model convergence means that the error between the customer complaint risk output by the model and the supervisory labels of the supplemented training samples is less than a preset threshold. A model trained to convergence forms at least one corresponding decision tree, which contains multiple nodes. Each leaf node is the last node in the corresponding decision path, in order of splitting.

[0099] Step S14113: Determine multiple corresponding evaluation explanation texts based on multiple reasoning paths of the customer complaint analysis model, and construct a customer complaint evaluation strategy set.

[0100] Filter out the training samples corresponding to the level of frequent customer complaints in a short period of time in the supplemented training set, and use them as high-risk samples. Traverse each leaf node in each decision tree corresponding to the customer complaint analysis model that has been trained to a convergent state, and determine the decision path in the decision tree where each leaf node is located, that is, the reasoning path. For each reasoning path, match the reasoning path with each training sample in the training set, determine the total number of high-risk samples in all training samples that hit the reasoning path, and then divide the total number of high-risk samples by the total number of all corresponding training samples to calculate the high-risk accuracy rate corresponding to the reasoning path. Finally, select multiple reasoning paths whose high-risk accuracy rate exceeds the preset threshold or whose high-risk accuracy rate is relatively high, and use them as evaluation explanation texts respectively, and combine them into a customer complaint evaluation strategy set.

[0101] In this embodiment, by first using the trained customer complaint prediction model to infer the training set samples and supplement the customer complaint risk, this not only enriches the data dimension of the training samples, but also provides an important basis for subsequent strategy mining. Furthermore, the customer complaint analysis model is trained using the supplemented training set, so that the model can deeply learn the association rules between feature combinations and customer complaint risks, thereby improving the model's predictive ability and interpretability. Finally, based on the reasoning path of the customer complaint analysis model, the path with high risk accuracy is screened as the evaluation explanation text, and a customer complaint evaluation strategy set is constructed, providing the platform with a detailed and explainable evaluation basis, so that the platform's operators can clearly understand the specific reasons for the customer complaint risk, so that when facing stores with high customer complaint risks, they can quickly and accurately give evaluation explanations, avoid the potential negative impact of a large number of customer complaints on the platform, and maintain the store's right to know and evaluation experience, and maintain the normal operation order and reputation of the platform.

[0102] In a further embodiment, before step S1100, responding to the event that the target store's transaction amount reaches the target amount on the same day, the following steps are included:

[0103] Step S1000: Perform a single training on the customer complaint prediction model using a preset training set to obtain a version of the customer complaint prediction model trained to a convergence state, and determine the inference reliability of the version of the model;

[0104] The customer complaint prediction model was trained once using a preset training set until the model reached convergence, thereby obtaining the version of the customer complaint prediction model generated by this training. The customer complaint prediction model can be a machine learning architecture such as a gradient boosting decision tree (e.g., the LightGBM model). The training process includes initializing model hyperparameters (e.g., tree depth, minimum number of leaf node instances) and iteratively constructing a weak classifier using the gradient boosting framework: in each iteration, the negative gradient (pseudo-residual) is calculated based on the error between the supervised label of the training sample (e.g., 0 represents no customer complaint, 1 represents a customer complaint) and the model's predicted output. The feature histogram optimization algorithm is then used to quickly determine the optimal split point to add a new decision tree. The leaf-wise growth strategy is then used to prioritize splitting nodes with the highest gain, thereby improving training efficiency while ensuring accuracy. When the error between the model's prediction result and the supervisory label is lower than the preset threshold, the model is judged to be in a converged state. At the same time, in order to evaluate the generalization ability of the model, the predicted probability of the training sample (in the range of 0 to 1, representing the customer complaint risk) is calculated to generate performance indicators such as the AUC value (by drawing the ROC curve and calculating the area under the curve) or the KS value (based on the cumulative distribution difference of the samples), which is defined as the inference reliability. The higher the value of this indicator (closer to 1), the stronger the model's ability to distinguish between customer complaints and non-customer complaints. The preset threshold can be set by technical personnel according to specific business needs to ensure that the model can be applied to the actual environment.

[0105] Step S1010: When the inference reliability meets the preset conditions, the historical business data with the relatively low estimated value of customer complaints in the training samples in the training set are eliminated;

[0106] When the determined inference reliability meets the preset condition (which can be set as required by those skilled in the art based on the disclosure herein, for example, the AUC value is greater than 0.8 or the KS value is greater than 0.5), it indicates that the current complaint estimation model is initially efficient; at this time, in order to simplify the model input and optimize subsequent training, each training sample in the training set needs to be analyzed, and the historical operation data with relatively poor complaint estimation value is removed. Specifically, the complaint estimation value is based on the feature importance weight (such as calculated by feature division gain) output by the model, and the relatively poor complaint estimation value refers to the historical operation data with a weight lower than a specified threshold, and the purpose of removing it is to reduce redundant input data while ensuring that the training set after removal can maintain the model performance index above an acceptable range, and the number and threshold of removal can be dynamically adjusted according to the scene; then, the above-mentioned iterative training is repeated using the updated training set to ensure that the model re-converges and adapts to the new data distribution, thereby achieving the balance between simplifying the model inference and performance.

[0107] Step S1020, when the inference reliability does not meet the preset condition, the preset validation set is used to verify each version of the complaint estimation model in reverse according to the training order, and the first version that meets the standard is determined as the complaint estimation model for online application;

[0108] When the inference reliability does not meet the preset condition (for example, the AUC value is lower than the preset threshold or the KS value is insufficient), it means that the performance of the model trained in the current iteration has decreased and may be over-fitted, at which time the reverse verification mechanism needs to be started to locate the reliable version. For this purpose, the preset validation set (containing historical operation data samples independent of the training set, and the supervision label is also related to the complaint event) is used to verify each complaint estimation model version obtained by training in reverse (i.e. from the latest version to the earliest version) according to the training order; when verifying, the inference reliability (such as AUC value and KS value) of each version under the validation set is calculated, and the inference reliability under the training set is compared, if the difference between the two is within the tolerance range (for example, the AUC difference is not more than 0.05), it is determined that the model version meets the standard; the verification is performed from the back to the front, until the first version that meets the tolerance requirement (i.e. the model with the most stable performance and strong generalization ability) is identified, and it is specified as the complaint estimation model for online application. In this way, the method ensures that the finally selected model has high online deployment reliability while relying on a small amount of inference data, and also avoids the risk of excessive simplification leading to large errors.

[0109] Step S1030, the historical operation data used in the training of the complaint estimation model for online application is taken as complaint estimation data, and the field to which the complaint estimation data belongs is taken as the estimation field.

[0110] The online application-based complaint estimation model determines specific input data on which the inference relies: historical operation data used in the training of the model (which has been optimized in the iterative optimization, such as entries that have not been removed in the elimination) is defined as complaint estimation data; then, the fields to which each complaint estimation data belongs are extracted and listed as estimation fields, which constitute necessary input sub-dimensions of the model in actual deployment, facilitating subsequent integration and monitoring.

[0111] In this embodiment, by first training the model using the preset training set, obtaining the model version in the convergent state and determining the inference reliability, the model can be ensured to have certain prediction ability in the initial stage, and then the model is further optimized. If the inference reliability meets the preset condition, the samples with substandard complaint estimation value in the training set are eliminated, the model input is simplified while the model performance is ensured, so that the model can more efficiently learn the key features in the subsequent iterative training, and the prediction accuracy is improved. On the contrary, if the inference reliability does not meet the condition, the model is verified from the latest version to the earliest version one by one through the reverse verification mechanism, and the first version that meets the condition is determined as the complaint estimation model for online application. In this way, prediction errors caused by overfitting or performance degradation of the model can be avoided, the stability and reliability of the model in actual application can be ensured, and the inference of customer service risk degree based on the least effective complaint evaluation data can provide strong support for subsequent complaint risk evaluation.

[0112] In a further embodiment, before step S1000, the preset training set is used to train the complaint estimation model once, a version of the complaint estimation model trained to a convergent state is obtained, and the inference reliability of the version of the model is determined, and the steps include the following steps:

[0113] Step S2000, obtaining historical operation data corresponding to multiple dimensions of a plurality of short-term explosive shop, constructing the historical operation data of the shop as a sample in units of shops, and the dimensions include merchant information dimension, disposal information dimension, shop information dimension, and shop transaction dimension;

[0114] For the pre-prepared sample set, a plurality of short-term explosive shops that meet the short-term explosive condition can be determined according to the corresponding transaction amount performance of the shops of the e-commerce platform in the historical operation process. Further, for each short-term explosive shop, the historical operation data corresponding to multiple dimensions of the shop at the moment (i.e., the current moment) when the short-term explosive condition is determined is obtained, specifically,

[0115] Historical business data in the merchant information dimension includes data corresponding to multiple fields, each of which is considered historical business data. Any number of fields include: the merchant to which the short-term explosive store belongs, the total number of stores under its name, the total number of currently operating stores, the total number of currently deregistered stores, the time since the last store deregistration, and the time since the last store registration; the total transaction amount, total customer complaint amount, percentage of customer complaint orders, total transaction orders, and total transaction days generated by all stores under the merchant's name within at least one recent time period.

[0116] The historical operations section of the disposal information dimension includes data corresponding to multiple fields, each of which serves as historical operations data. It's understandable that e-commerce platform operators will take action against stores with a high risk of customer complaints. These fields include: the merchant of the store experiencing a short-term surge in sales, the total number of stores under their name that have been disposed of and their percentage of the total, the time since the earliest store disposal, and the time since the latest store disposal.

[0117] The historical operation of the store information dimension includes data corresponding to multiple fields, each of which is considered historical operation data. Any number of fields include: the store registration days, the store's country of origin, the store's product categories, the store's historical total number of products launched, the store's historical total number of products removed from shelves, and the store's current total number of products for sale.

[0118] The store transaction history includes data corresponding to multiple fields, each of which serves as historical operating data. Any number of fields include: the time between the first transaction of a short-term hot-selling store and the current moment; the store's total transaction amount, total number of transactions, total customer complaint amount, total number of logistics information updates in transactions, total customer complaint amount and its ratio to normal transaction amount, and the transaction amount of multiple high-volume items and their ratio to the total transaction amount of all goods, compared to adjacent time periods. The month-on-month growth rate of transaction amount, month-on-month growth rate of transaction number, month-on-month growth rate of customer complaint amount, and month-on-month growth rate of the ratio of customer complaint amount to normal transaction amount are compared.

[0119] Step S210: Based on whether the customer complaint performance of each sample corresponding to the short-term high-order store meets the high customer complaint condition during the corresponding short-term high-order store operation period, a supervision label is correspondingly marked for each sample, and a sample set is constructed using all samples and their supervision labels;

[0120] The short-term hot-selling business period refers to the time period from the confirmation of the store as a short-term hot-selling store to the end of the corresponding period after a preset period of time. Those skilled in the art can set it as needed based on the disclosure here, for example, 30 days after the hot-selling store is confirmed.

[0121] The complaint performance can represent whether the short-term single shop is disposed of by manual judgment of high complaint risk during the short-term single operation period, or can be the corresponding complaint transaction amount ratio of the short-term single shop during the short-term single operation period.

[0122] By calculating the sum of the amounts of all orders of the short-term single shop during the short-term single operation period and the sum of the amounts of all orders complained by customers in these orders, the total transaction amount and the total complaint transaction amount are obtained, and the total complaint transaction amount is divided by the total transaction amount to obtain the complaint transaction amount ratio.

[0123] It is not difficult to understand that the corresponding complaint performance can be pre-configured with high complaint conditions, for example: representing whether the short-term single shop is disposed of by manual judgment of high complaint risk during the short-term single operation period, the corresponding complaint performance, and the high complaint condition for this judgment can be shop disposal; the complaint transaction amount ratio as the complaint performance, and the high complaint condition for this judgment can be 6%.

[0124] The high complaint condition is used to determine whether the complaint performance meets the standard. If it meets the standard, it means that the short-term single shop has frequent customer complaints during the short-term single operation period, and the supervision label of the corresponding training sample is marked as 1; otherwise, the supervision label of the corresponding training sample is marked as 0.

[0125] All training samples and their supervision labels are collected into a sample set.

[0126] Step S2020, according to the pre-set sample division strategy, the training set and the verification set are divided from the sample set.

[0127] According to the pre-set sample division strategy, the training set for training the model and the verification set for verifying the model effect are reasonably divided from the sample set containing multiple short-term single shops and their supervision labels which have been constructed. This division method usually follows certain proportion rules, for example, the common division proportions are 7:3 or 8:2, that is, the number of positive samples and negative samples in the training set accounts for 70% or 80% of the number of positive samples and negative samples in the whole sample set, and correspondingly, the number of positive samples and negative samples in the verification set accounts for 30% or 20%. Such division proportion can ensure that the training set has enough samples for the model to learn the relationship between shop features and complaint performance, so that the model can learn more comprehensive and accurate rules based on a larger amount of data; at the same time, the verification set also has enough samples for evaluating the performance of the model on the data not involved in the training, so as to judge whether the model has good generalization ability, can accurately predict the complaint risk of new short-term single shops, and thus provide objective and effective basis for subsequent model optimization and other operations.

[0128] In this embodiment, by obtaining multi-dimensional historical operating data of a plurality of short-term explosive single shops, constructing the multi-dimensional historical operating data as samples, labeling supervision labels to build a sample set, and dividing a training set and a validation set, various types of information related to the short-term explosive single shops can be comprehensively collected, including merchant information, disposal information, shop information, and shop transaction information, etc., thereby providing rich and multi-angle data support for the complaint estimation model. Based on these samples, supervised learning is performed, and the model can fully learn the correlation between the characteristics of the short-term explosive single shops in different dimensions and the complaint performance, thereby improving the accuracy of the complaint risk prediction of new shops. At the same time, a reasonable sample division strategy can ensure that the model has enough data to learn in the training process, and can effectively evaluate the generalization ability of the model in the validation phase, ensuring that the model can accurately estimate the complaint risk of new short-term explosive single shops in actual application, and providing a reliable basis for the complaint shop control of the platform.

[0129] Please refer to Figure 3 , a complaint shop control device provided for one of the purposes of the present application, is a functional embodiment of the complaint shop control method of the present application. On the other hand, a complaint shop control device provided for one of the purposes of the present application includes an event response module 1100, a data acquisition module 1200, a complaint estimation module 1300, and a shop disposal module 1400. The event response module 1100 is used to respond to the transaction amount reaching the target shop's daily transaction amount event, count the daily transaction amount of the target shop in the statistical period of the day, and determine the corresponding statistical feature distribution. The data acquisition module 1200 is used to acquire the corresponding complaint estimation data set of the target shop at the current time when the statistical feature distribution meets the short-term explosive single condition, and the complaint estimation data set includes complaint estimation data corresponding to a plurality of estimation fields. The complaint estimation module 1300 is used to call a preset complaint estimation model to infer the corresponding complaint risk degree according to the complaint estimation data set. The shop disposal module 1400 is used to determine the complaint rating to which the complaint risk degree belongs, and execute the shop disposal process corresponding to the complaint rating on the target shop.

[0130] In a further embodiment, the event response module 1100 comprises: a transaction volume judgment submodule for listening to the daily transaction amount of the target store and judging whether the daily transaction amount exceeds a preset large transaction volume; a time period determination submodule for, when the transaction amount exceeds the large transaction volume, expanding the statistical time period along the time axis in both directions based on the date corresponding to the transaction amount and the corresponding reference time point; a time period division submodule for dividing the statistical time period into a preceding time period and a subsequent time period based on the reference time point; a feature determination submodule for determining the statistical features of the preceding time period and the subsequent time period based on the daily transaction amount of the target store in the statistical time period; and a distribution determination submodule for integrating the statistical features in time sequence to obtain a statistical feature distribution.

[0131] In a further embodiment, the store handling module 1400 comprises: a first-level processing submodule for, when the complaint rating is a level representing frequent complaints in a short period, canceling the transaction permission of the target store; a second-level processing submodule for, when the complaint rating is a level representing intermittent complaints in a short period, observing the complaint performance of the target store in a preset time period and determining the complaint rating again based on the complaint performance; and a third-level processing submodule for, when the complaint rating is a level representing regular complaints in a short period, not handling the target store.

[0132] In a further embodiment, the first-level processing submodule is followed by: a text determination submodule for determining a target evaluation explanation text in the preset complaint evaluation strategy set that matches the complaint estimation data set; and a report pushing submodule for constructing an evaluation report based on the complaint risk degree, the complaint rating, and the target evaluation explanation text and pushing the evaluation report to the target store.

[0133] In a further embodiment, the text determination submodule is preceded by: a feature supplement submodule for supplementing a complaint risk degree of a training sample in a preset training set to the training sample by using a preset complaint estimation model to infer the complaint risk degree; a model training submodule for training a complaint analysis model to a convergent state by using the supplemented training set; and a strategy set construction submodule for determining a plurality of evaluation explanation texts corresponding to a plurality of inference paths of the complaint analysis model to construct a complaint evaluation strategy set.

[0134] In a further embodiment, the event response module 1100 includes: a version determination submodule, which is used to use a preset training set to perform a single training on the customer complaint prediction model, obtain the version of the customer complaint prediction model that has been trained to a convergence state, and determine the reasoning reliability of the version model; a data elimination submodule, which is used to eliminate historical business data whose customer complaint prediction value is relatively substandard in the training samples in the training set when the reasoning reliability meets the preset conditions; a model determination submodule, which is used to use a preset verification set to reversely verify each version of the customer complaint prediction model in the training order when the reasoning reliability does not meet the preset conditions, and determine the version that is first verified to meet the standards as the customer complaint prediction model for online application; a field determination submodule, which is used to use the historical business data used in the training of the customer complaint prediction model for online application as the customer complaint prediction data, and use the field to which the customer complaint prediction data belongs as the estimation field.

[0135] In a further embodiment, before the version determination submodule, it includes: an initial data acquisition submodule, which is used to obtain historical operating data corresponding to multiple dimensions of multiple short-term hot-selling stores, and construct their historical operating data into samples based on stores, and the dimensions include merchant information dimension, disposal information dimension, store information dimension, and store transaction dimension; a set construction submodule, which is used to mark the supervision label of each sample according to whether the customer complaint performance of the short-term hot-selling store corresponding to each sample in its corresponding short-term hot-selling operating period meets the high customer complaint condition, and construct a sample set with all samples and their supervision labels; a set division submodule, which is used to divide the training set and the validation set from the sample set according to a preset sample division strategy.

[0136] In order to solve the above technical problems, the embodiment of the present application also provides a computer device. Figure 4 As shown, a schematic diagram of the internal structure of a computer device. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. Among them, the computer-readable storage medium of the computer device stores an operating system, a database, and computer-readable instructions. The database may store a control information sequence. When the computer-readable instructions are executed by the processor, the processor may implement a customer complaint store management method. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor may execute the customer complaint store management method of the present application. The network interface of the computer device is used to connect and communicate with the terminal. Those skilled in the art will understand that Figure 4The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0137] The processor in the embodiment is configured to execute the specific functions of each module and the sub-modules thereof in Figure 3 The memory stores the program codes and various data required for executing the above-mentioned modules or sub-modules. The network interface is configured to transmit data between the user terminal and the server. The memory in the embodiment stores the program codes and data required for executing all the modules / sub-modules in the complaint store management device of the present application, and the server can call the program codes and data of the server to execute the functions of all the sub-modules.

[0138] The present application also provides a storage medium storing computer readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the complaint store management method of any embodiment of the present application.

[0139] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments of the present application can be completed by a computer program instructing related hardware. The computer program can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments of each method. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0140] In summary, the present application can efficiently and accurately predict high-complaint-risk stores in short-term explosive stores, and timely control them to avoid unnecessary losses, greatly reducing the complaint rate of the platform.

[0141] Those skilled in the art can understand that the steps, measures, and schemes in the various operations, methods, and processes discussed in the present application can be alternated, changed, combined, or deleted. Further, other steps, measures, and schemes in the various operations, methods, and processes discussed in the present application can also be alternated, changed, rearranged, decomposed, combined, or deleted. Further, the steps, measures, and schemes in the prior art with the various operations, methods, and processes disclosed in the present application can also be alternated, changed, rearranged, decomposed, combined, or deleted.

[0142] The above merely describes some embodiments of the present application, and it should be pointed out that, for those skilled in the art, some improvements and refinements can be made without departing from the principles of the present application, and these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A customer complaint store management method, characterized in that: The steps include: In response to the event that the target store's transaction amount reaches the target amount on that day, the target store's daily transaction amount within the statistical period of that day is counted to determine the corresponding statistical feature distribution; When the statistical feature distribution meets the short-term explosive order condition, obtaining a customer complaint estimation dataset corresponding to the target store at the current moment, wherein the customer complaint estimation dataset includes customer complaint estimation data corresponding to multiple estimation fields; Calling a preset customer complaint prediction model to infer the corresponding customer complaint risk based on the customer complaint prediction data set; Determine the customer complaint rating to which the customer complaint risk belongs, and execute a store handling process corresponding to the customer complaint rating for the target store.

2. The customer complaint store management method according to claim 1 is characterized in that: In response to the event that the target store's transaction amount reaches the target amount on that day, the target store's daily transaction amount within the statistical period of that day is counted and the corresponding statistical feature distribution is determined, including the following steps: Monitor the daily transaction amount of the target store and determine whether the transaction amount on that day exceeds the preset large transaction volume; When the transaction amount exceeds the large transaction volume, the statistical period is determined by bidirectionally expanding along the time axis based on the date corresponding to the transaction amount and its corresponding reference time point; Dividing the statistical period into an earlier period and a later period with the reference time point as the boundary; Determining the statistical characteristics of the preceding and succeeding time periods based on the daily transaction amounts of the target store during the statistical time period; The statistical features are integrated according to time series to obtain a statistical feature distribution.

3. The customer complaint store management method according to claim 1 is characterized in that: The store handling process corresponding to the customer complaint rating is executed for the target store, including the following steps: When the customer complaint rating is a level indicating frequent customer complaints in a short period of time, the transaction privileges of the target store are revoked; When the customer complaint rating is a level representing intermittent customer complaints in a short period of time, observing the customer complaint performance of the target store within a preset period of time, so as to re-determine the customer complaint rating based on the customer complaint performance; When the customer complaint rating is a level representing a regular customer complaint in a short period of time, there is no need to deal with the target store.

4. The customer complaint store management method according to claim 3 is characterized in that: When the customer complaint rating is a level indicating frequent customer complaints in a short period of time, after the transaction privileges of the target store are revoked, the following steps are included: Determining a target evaluation interpretation text in a preset customer complaint evaluation strategy set that matches the customer complaint estimation dataset; An evaluation report is constructed based on the customer complaint risk, customer complaint rating, and target evaluation explanation text, and the evaluation report is pushed to the target store.

5. The customer complaint store management method according to claim 4 is characterized in that: Before determining the target evaluation interpretation text that matches the customer complaint estimation data set in the preset customer complaint evaluation strategy set, the following steps are included: Using a preset customer complaint prediction model to infer the customer complaint risk of training samples in a preset training set, and adding the customer complaint risk to the training samples; Use the supplemented training set to train the customer complaint analysis model until convergence; Based on the multiple reasoning paths of the customer complaint analysis model, multiple corresponding evaluation interpretation texts are determined to construct a customer complaint evaluation strategy set.

6. The customer complaint store management method according to claim 1, characterized in that: Before responding to the event that the target store's transaction amount reaches the target amount on the day, the following steps are included: Conduct a single training of the customer complaint prediction model using a preset training set, obtain a version of the customer complaint prediction model that has been trained to a convergent state, and determine the inference reliability of this version of the model; When the reliability of the reasoning meets the preset conditions, the historical business data with relatively low estimated value of customer complaints in the training samples in the training set are eliminated; When the inference reliability does not meet the preset conditions, the preset verification set is used to reversely verify the customer complaint prediction model of each version in the training order, and the version that is first verified to meet the standards is determined as the customer complaint prediction model for online application; The historical operating data used in training the online customer complaint prediction model is used as the customer complaint prediction data, and the field to which the customer complaint prediction data belongs is used as the prediction field.

7. The customer complaint store management method according to claim 6, characterized in that: The customer complaint prediction model is trained once using a preset training set to obtain a version of the customer complaint prediction model that has been trained to a convergent state. The following steps are included before determining the inference reliability of the version of the model: Obtain historical operating data corresponding to multiple dimensions of multiple short-term hot-selling stores, and construct their historical operating data into samples based on the store. The dimensions include merchant information, disposal information, store information, and store transaction dimensions. Based on whether the customer complaint performance of each sample's corresponding short-term high-order store meets the high customer complaint condition during its corresponding short-term high-order store period, each sample is labeled with a supervision label, and a sample set is constructed with all samples and their supervision labels; A training set and a validation set are divided from the sample set according to a preset sample division strategy.

8. A customer complaint store management device, characterized in that: include: The event response module is used to respond to the event that the target store's transaction amount reaches the target amount on that day, collect statistics on the target store's daily transaction amount within the statistical period of that day, and determine the corresponding statistical feature distribution; A data acquisition module is configured to acquire a customer complaint prediction dataset corresponding to the target store at the current moment when the statistical feature distribution meets the short-term explosive order condition, wherein the customer complaint prediction dataset includes customer complaint prediction data corresponding to multiple prediction fields; A customer complaint prediction module is used to call a preset customer complaint prediction model to infer the corresponding customer complaint risk based on the customer complaint prediction data set; The store handling module is used to determine the customer complaint rating to which the customer complaint risk belongs, and execute the store handling process corresponding to the customer complaint rating for the target store.

9. A computer device comprising a central processing unit and a memory, characterized in that: The central processing unit is configured to call and run a computer program stored in the memory to execute the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that It stores a computer program implemented according to the method described in any one of claims 1 to 7 in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

Images