Object classification processing method, device, computer equipment and storage medium

Abstract

This application relates to an object classification processing method, apparatus, computer equipment, and storage medium. Applicable to the fields of mapping and transportation, the method includes: determining an object classification model with classification accuracy to be analyzed; obtaining historical object features of multiple objects in at least three target historical periods; inputting the historical object features of each object in each historical period into the object classification model with classification accuracy to be analyzed for classification, thereby predicting the predicted category of each object in the next historical period, thereby obtaining the predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods; and determining the classification accuracy of the object classification model in the current period based on the predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods. This method can improve the precision of classification accuracy.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an object classification processing method, apparatus, computer equipment, and storage medium. Background Art

[0002] With the development of computer technology, machine learning has found an increasingly broad range of applications, including transportation, gaming, and autonomous driving. For example, in the transportation sector, for scenarios requiring classification, machine learning can be used to train a corresponding classification model, which can then be used to implement the classification.

[0003] Currently, since classification accuracy is an important indicator of a classification model, it is usually necessary to analyze the classification accuracy of the classification model. However, the current method of analyzing the accuracy of the classification model is not accurate enough, resulting in low precision of the classification accuracy. Summary of the Invention

[0004] Based on this, it is necessary to provide an object classification processing method, device, computer equipment, computer-readable storage medium and computer program product that can improve the accuracy of classification in order to address the above technical problems.

[0005] On the one hand, the present application provides an object classification processing method. The method includes: determining an object classification model for classification accuracy to be analyzed; obtaining historical object features of multiple objects in at least three target historical periods; the at least three target historical periods are at least three consecutive historical periods selected from the period before the current period; for each object in each target historical period, inputting the historical object features of each object into the object classification model for classification to predict the predicted category of each object in the next period of the target historical period, obtaining the predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods; the at least two adjacent historical periods are target historical periods that are closer to the current period among the at least three target historical periods; based on the predicted category of each object in the current period and the predicted category of each object in the at least two adjacent historical periods, determining the classification accuracy of the object classification model in the current period.

[0006] On the other hand, the present application also provides an object classification processing device. The device includes: a model determination module for determining an object classification model for classification accuracy to be analyzed; a feature acquisition module for acquiring historical object features of multiple objects in at least three target historical periods; the at least three target historical periods are at least three consecutive historical periods selected from the period before the current period; a category acquisition module for inputting the historical object features of each object in each historical period into the object classification model for classification, so as to predict the predicted category of each object in the next period of the historical period, and obtain the predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods; the at least two adjacent historical periods are target historical periods that are closer to the current period among the at least three target historical periods; and an accuracy determination module for determining the classification accuracy of the object classification model in the current period based on the predicted category of each object in the current period and the predicted category of each object in the at least two adjacent historical periods.

[0007] In some embodiments, the at least two adjacent historical periods include a first historical period and a second historical period, the first historical period is adjacent to the current period and is before the current period, and the second historical period is adjacent to the first historical period and is before the first historical period; the accuracy determination module is also used to: obtain the predicted category of each of the objects in the current period to obtain the current predicted category of each of the objects; obtain the predicted category of each of the objects in the first historical period to obtain the first predicted category of each of the objects; obtain the predicted category of each of the objects in the second historical period to obtain the second predicted category of each of the objects; based on the current predicted category, the first predicted category and the second predicted category of each of the objects, determine the classification accuracy of the object classification model in the current period.

[0008] In some embodiments, the accuracy determination module is further used to: count the number of objects among the multiple objects that meet the first category condition to obtain the first object number; the first category condition includes: the actual category of the object in the current period is the first preset category, and the current predicted category of the object is the first preset category; count the number of objects among the multiple objects that meet the second category condition to obtain the second object number; the second category condition includes: the actual category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the first preset category; based on the first object number and the second object number, determine the classification accuracy of the object classification model in the current period.

[0009] In some embodiments, the accuracy determination module is further used to: count the number of objects among the multiple objects that meet the third category condition to obtain the third object number; the third category condition includes: the actual category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the actual category of the object in the first historical period is the second preset category, and the first predicted category of the object is the first preset category; based on the first object number and the second object number, statistics are performed to obtain a positive statistical value; the positive statistical value is positively correlated with the first object number and the second object number; based on the first object number and the third object number, statistics are performed to obtain a negative statistical value; the negative statistical value is positively correlated with the first object number and the third object number; based on the positive statistical value and the negative statistical value, determine the classification accuracy of the object classification model in the current period; the classification accuracy is positively correlated with the positive statistical value, and the classification accuracy is negatively correlated with the negative statistical value.

[0010] In some embodiments, the accuracy determination module is further used to: count the number of objects among the multiple objects that meet the fourth category conditions to obtain the fourth object number; the fourth category conditions include: the actual category of the object in the current period is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; statistics are performed based on the first object number, the second object number and the fourth object number to obtain a positive statistical value; the positive statistical value is negatively correlated with the fourth object number.

[0011] In some embodiments, the negative statistical value includes a first negative statistical value, and the classification accuracy includes a first classification accuracy; the accuracy determination module is also used to: count the number of objects among the multiple objects that meet the fifth category conditions to obtain the fifth object number; the fifth category conditions include: the actual category of the object in the first historical period is the first preset category, the first predicted category of the object is the second preset category, the actual category of the object in the second historical period is the first preset category, and the second predicted category of the object is the second preset category; statistics are performed based on the first object number, the third object number and the fifth object number to obtain a first negative statistical value; the first negative statistical value is positively correlated with the fifth object number; based on the positive statistical value and the first negative statistical value, the first classification accuracy of the object classification model in the current period is determined.

[0012] In some embodiments, the accuracy determination module is further used to: count the number of objects among the multiple objects that meet the fourth category conditions to obtain the fourth object number; the fourth category conditions include: the actual category of the object in the current period is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; based on the first object number, the third object number, the fourth object number and the fifth object number, statistics are performed to obtain a first negative statistical value; the first negative statistical value is negatively correlated with the fourth object number.

[0013] In some embodiments, the accuracy determination module is further used to: count the number of objects among the multiple objects that meet the sixth category conditions to obtain the sixth object number; the sixth category conditions include: the actual category of the object in the current period is the second preset category, the current predicted category of the object is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; based on the first object number, the third object number, the fourth object number, the fifth object number and the sixth object number, statistics are performed to obtain a first negative statistical value; the first negative statistical value is negatively correlated with the sixth object number.

[0014] In some embodiments, the negative statistical value includes a second negative statistical value, and the classification accuracy includes a second classification accuracy; the accuracy determination module is also used to: perform statistics based on the first number of objects, the second number of objects, and the third number of objects to obtain a second negative statistical value; the second negative statistical value is positively correlated with the second number of objects; the determination of the classification accuracy of the object classification model in the current cycle based on the positive statistical value and the negative statistical value includes: determining the second classification accuracy of the object classification model in the current cycle based on the positive statistical value and the second negative statistical value.

[0015] In some embodiments, the device further includes: a first model determination module, which is used to determine the object classification model with the classification accuracy to be analyzed as the object classification model to be trained when the classification accuracy is less than the accuracy threshold; a second model determination module, which is used to train the object classification model to be trained to obtain a new object classification model with the classification accuracy to be analyzed, and return to the step of inputting the historical object features of each of the objects into the object classification model for classification until the classification accuracy reaches the accuracy threshold; and a third model determination module, which is used to determine the object classification model with the classification accuracy to be analyzed when the classification accuracy reaches the accuracy threshold as the trained object classification model.

[0016] In some embodiments, the second model determination module is further used to: input the object features of the training object in the first historical period into the object classification model to be trained for classification, and obtain the predicted category of the training object in the current period; based on the difference between the predicted category of the training object in the current period and the actual category of the training object in the current period, adjust the parameters of the object classification model to be trained to obtain a new object classification model for classification accuracy to be analyzed.

[0017] In some embodiments, the current cycle is the current promotion cycle for the target service, and the classification category of the trained object classification model is either retention or loss; the device is also used to: obtain object features of multiple candidate objects in the current promotion cycle; input the object features of each candidate object in the current promotion cycle into the trained object classification model for classification, and obtain the predicted category of each candidate object in the next promotion cycle of the target service; select a target object with a predicted category of retention from each candidate object; and push promotional content related to the target service to the target object in the next promotion cycle.

[0018] In another aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the above-mentioned object classification processing method when executing the computer program.

[0019] In another aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps in the above-mentioned object classification processing method when executed by a processor.

[0020] On the other hand, the present application further provides a computer program product, which includes a computer program that implements the steps in the above-mentioned object classification processing method when executed by a processor.

[0021] The object classification processing method, apparatus, computer device, storage medium, and computer program product described above determine an object classification model for classification accuracy to be analyzed by obtaining historical object features of multiple objects in at least three target historical periods, where the at least three target historical periods are at least three consecutive historical periods selected from the period immediately preceding the current period. For each object, the historical object features in each historical period are input into the object classification model for classification, thereby predicting the predicted category of each object in the next historical period. The predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods are obtained, where the at least two adjacent historical periods are target historical periods that are closer to the current period among the at least three target historical periods. Based on the predicted category of each object in the current period and the predicted category of each object in the at least two adjacent historical periods, the classification accuracy of the object classification model in the current period is determined. Thus, the classification accuracy is determined by combining the predicted categories of the objects in multiple consecutive periods (including the current period), thereby improving the precision of the classification accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A diagram of an application environment of an object classification processing method in some embodiments;

[0023] Figure 2 is a flowchart of an object classification processing method in some embodiments;

[0024] Figure 3 Schematic diagram of each cycle in some embodiments;

[0025] Figure 4 A schematic diagram of a principle for calculating the number of second objects in some embodiments;

[0026] Figure 5 A schematic diagram of pushing promotional content to a target object in some embodiments;

[0027] Figure 6 is a flowchart of an object classification processing method in some embodiments;

[0028] Figure 7 is a flowchart of an object classification processing method in some embodiments;

[0029] Figure 8 is a structural block diagram of an object classification processing device in some embodiments;

[0030] Figure 9 is a diagram of the internal structure of a computer device in some embodiments;

[0031] Figure 10 1 is a diagram of the internal structure of a computer device in some embodiments. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0033] The object classification processing method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. The terminal 102 communicates with the server 104 via a network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or placed on the cloud or other servers. A variety of applications can be installed on the terminal 102, for example, an instant messaging application, a third-party payment application, a video viewing application, or a vehicle service application. Mini-programs can be embedded in the applications installed on the terminal 102, for example, an instant messaging application or a third-party payment application has a mini-program embedded in it, and the mini-programs include but are not limited to at least one of a ride-hailing mini-program, a food delivery mini-program, or a vehicle service mini-program.

[0034] Specifically, server 104 may determine an object classification model for classification accuracy analysis by obtaining historical object features for multiple objects in at least three target historical periods, where the at least three target historical periods are at least three consecutive historical periods selected from the period immediately preceding the current period. For each object, the historical object features for each historical period are input into the object classification model for classification, thereby predicting the predicted category of each object in the next period of the historical period. The predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods are obtained, where the at least two adjacent historical periods are target historical periods that are closer to the current period among the at least three target historical periods. Based on the predicted category of each object in the current period and the predicted category of each object in the at least two adjacent historical periods, the classification accuracy of the object classification model in the current period is determined. Server 104 may determine an object classification model with a classification accuracy greater than an accuracy threshold as a trained object classification model, and use the trained object classification model to predict the category of the candidate object in the next period of the current period. The server 104 may send the category to which the candidate object belongs in the next period of the current period to the terminal 102, or the server 104 may determine the target object from among the candidate objects based on the category to which the candidate object belongs in the next period of the current period, and push the corresponding content to the terminal of the target object, such as the terminal 102, in the next period of the current period.

[0035] Terminal 102 may include, but is not limited to, various desktop computers, laptops, smartphones, tablet computers, in-vehicle terminals, IoT devices, and portable wearable devices. IoT devices may include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices may include smart watches, smart bracelets, head-mounted devices, etc. Server 104 may be implemented as a standalone server or a server cluster consisting of multiple servers.

[0036] The object classification processing method provided in this application can be applied to the map field. For example, the object classification processing method proposed in this application can be used to obtain an object classification model with high classification accuracy for classifying users in map-type application software platforms, so as to classify users in map-type application software platforms and provide users with more appropriate map-related functional services based on their categories.

[0037] The object classification processing method provided in this application can be based on artificial intelligence. For example, the object classification model can be a neural network model. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that the machines have the functions of perception, reasoning and decision-making.

[0038] Artificial intelligence (AI) technology is a comprehensive discipline encompassing a wide range of fields, encompassing both hardware and software technologies. Foundational AI technologies generally include sensors, specialized AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, speech processing, natural language processing, as well as machine learning / deep learning, autonomous driving, and smart transportation.

[0039] Machine learning (ML) is a multidisciplinary field that encompasses probability theory, statistics, approximation theory, convex analysis, and algorithmic complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is at the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications span all areas of AI. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.

[0040] With the research and progress of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, smart customer service, Internet of Vehicles, automatic driving, smart transportation, etc. It is believed that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0041] The solutions provided in the embodiments of this application involve technologies such as artificial neural networks of artificial intelligence, and are specifically described through the following embodiments:

[0042] In some embodiments, as Figure 2As shown, a method for object classification processing is provided, which can be executed by a terminal or a server, or by a terminal and a server together. Figure 1 Taking the server 104 in the example as an example, the following steps are included:

[0043] Step 202: Determine the object classification model whose classification accuracy is to be analyzed.

[0044] The object may be any type of living or inanimate object. Living objects include but are not limited to at least one of a natural person, an animal, or a plant. Inanimate objects include but are not limited to at least one of a table or a computer.

[0045] An object classification model is a model used to classify objects, that is, to identify the category of an object. The object classification model can be a neural network model, and the object classification model whose classification accuracy is to be analyzed can be trained or untrained. The object classification model can be a binary classification model or a multi-classification model. Multi-classification refers to at least three classifications. The object classification model can be, for example, an LR (Logistic Regression) binary classification model. The logistic regression model introduces the Sigmoid function into the linear regression model, maps the continuous output values ​​of the uncertain range of the linear regression to the range of (0, 1), and converts the linear regression model into a probabilistic prediction model.

[0046] The object classification model can be matched with the services provided by the application, and each service in the application can be matched with an object classification model. Services include, but are not limited to, services in independent applications or services in sub-applications. Independent applications can also be called parent applications. The parent application is an application that hosts sub-applications and provides an operating environment for the implementation of sub-applications. The parent application is a native application. A native application is an application that can run directly on the operating system. Sub-applications can run in the parent application. Sub-applications run in the operating environment of the parent application. The parent application includes, but is not limited to, at least one of a social application, a dedicated application that specifically supports sub-applications, a file management application, an email application, or a game application.

[0047] For example, a sub-application can be a travel service applet in an instant messaging software. A travel service applet is a mini-program for providing travel services, which include but are not limited to at least one of discounted refueling services, car wash services, car moving code services, or designated driver services. Each service can have a matching object classification model. For example, the discounted refueling service has a matching object classification model, and the car wash service also has a corresponding object classification model. The object can be a user registered in the application. For example, when the application is an application that provides travel services to car owners, the object can be the car owner registered on the application.

[0048] The input to the object classification model is the object features in the first period, and the output of the object classification model is the predicted category of the object in the second period. The first and second periods are adjacent periods, and the first period precedes the second period. A period can be, for example, a promotion period, which refers to the period during which content is pushed to an object. A promotion period can be, for example, a marketing period, or other periods, which are not limited here.

[0049] An application's services can have multiple marketing cycles. For example, a discounted refueling service can have multiple marketing cycles, with marketing conducted in each cycle. The object classification model can be a model used to identify the category of an object in the current marketing cycle. Specifically, the object classification model can be used to predict the category of an object in the next marketing cycle after the current marketing cycle based on the object's features in the current marketing cycle. For example, an object classification model can be trained to match the current marketing cycle of the discounted refueling service, and the trained object classification model can be used to identify the category of an object in the next marketing cycle after the current marketing cycle.

[0050] Object characteristics reflect the characteristics of an object. The characteristics of the same object can be different or the same during different marketing cycles. Object characteristics include at least one of object attribute characteristics or object interaction characteristics. Object attribute characteristics are characteristics derived from object attribute data, which include at least one of gender, age, or region. Object interaction attribute characteristics are characteristics derived from object interaction data, which refers to the data generated by the object within the application. Object interaction data may include at least one of activity attribute data, recharge attribute data, or coupon attribute data. Activity attribute data includes, but is not limited to, at least one of the following: number of active days, active duration, number of active functions, and the number of days between registration and the current time. Recharge attribute data includes, but is not limited to, at least one of the following: recharge amount, consumption amount, number of recharges, number of days between recharges, and the number of days between the first recharge and the current time. Coupon attribute data includes, but is not limited to, at least one of function clicks, coupon redemption information, coupon usage information, or coupon expiration information. Coupon claim information includes at least one of the type, quantity, number of times, or value of coupons claimed; coupon usage information includes at least one of the type, quantity, or value of coupons used; and coupon expiration information includes at least one of the type, quantity, or value of expired coupons. Coupons may include at least one of gift packages and gift certificates.

[0051] The application can include functional models corresponding to each service, and the functional models are used to provide corresponding services. For example, the functional module corresponding to the preferential refueling service is the preferential refueling functional module. The category of the object can be determined based on the interaction between the object and the functional module. Take the preferential refueling service as an example, and the marketing cycle of the preferential refueling service includes T-3, T-2, T-1 and T, where T-3, T-2, T-1 and T are four consecutive marketing cycles, and T is the current cycle. Figure 3As shown, a schematic diagram of each cycle is shown. In the preferential refueling service scenario, the object classification model can be set to identify categories including retention and churn. Among them, if the object logs in to the preferential refueling function module in the T-1 cycle and does not log in to the preferential refueling function module in the T cycle, the object's category in the T cycle is churn. If the object logs in to the preferential refueling function module in the T-1 cycle and also logs in to the preferential refueling function module in the T cycle, the object's category in the T cycle is retention. If the object logs in to the preferential refueling function module in the T-2 cycle and does not log in to the preferential refueling function module in the T-1 cycle, the object's category in the T-1 cycle is churn. If the object logs in to the preferential refueling function module in the T-2 cycle and also logs in to the preferential refueling function module in the T-1 cycle, the object's category in the T-1 cycle is retention. If the object logs in to the preferential refueling function module in the T-3 cycle and does not log in to the preferential refueling function module in the T-2 cycle, the object's category in the T-2 cycle is churn. If the subject logs into the preferential refueling function module in the T-3 period and also logs into the preferential refueling function module in the T-2 period, the category of the subject in the T-2 period is retention.

[0052] Specifically, the server may obtain an object classification model to be trained, which may be a trained model or an untrained model. The server may train the object classification model to be trained and determine the trained object classification model as the object classification model for which classification accuracy is to be analyzed.

[0053] In some embodiments, the server may obtain object features of one or more objects in a first historical period (where "multiple" refers to at least two objects), input the object features of the first historical period into an object classification model to be trained, predict the category to which the object belongs in the current period, adjust parameters of the object classification model based on the difference between the predicted category of each object and the object's actual category in the first historical period, and determine the object classification model with adjusted parameters as the object classification model for classification accuracy analysis. The first historical period and the current period are two consecutive periods, and the first historical period precedes the current period.

[0054] Step 204 , obtaining historical object features of the plurality of objects in at least three target historical periods; the at least three target historical periods are at least three consecutive historical periods selected from the period preceding the current period.

[0055] The at least three target historical periods are at least three consecutive historical periods selected from the period preceding the current period. The at least three target historical periods include the period preceding the current period, which is the period adjacent to and preceding the current period. For example, T-3, T-2, T-1, and T are four consecutive periods. If the current period is T, then T-3, T-2, and T-1 belong to the at least three target historical periods. Historical object features refer to the object features possessed by an object in a historical period.

[0056] Specifically, the server can obtain the historical object characteristics of multiple objects in each of the three target historical periods. The three target historical periods are three consecutive historical periods selected from the previous period of the current period. For example, the current period is T, and the three target historical periods are T-3, T-2 and T-1 respectively, where T-1 is the previous period of the current period.

[0057] In some embodiments, the three target historical periods are, in order, the third historical period, the second historical period, and the first historical period. The third historical period precedes and is adjacent to the second historical period, the second historical period precedes and is adjacent to the first historical period, and the first historical period precedes and is adjacent to the current period. For example, the current period is T, the third historical period is T-3, the second historical period is T-2, and the first historical period is T-1. The server can obtain the object features of each object in the first historical period, the object features of the second historical period, and the object features of the third historical period to obtain the historical object features of each object.

[0058] In step 206, the historical object features of each object in each target historical period are input into the object classification model for classification to predict the predicted category of each object in the next period of the target historical period, and the predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods are obtained; the at least two adjacent historical periods are target historical periods that are closer to the current period among the at least three target historical periods.

[0059] The predicted category is the category identified by the object classification model. The at least two adjacent historical periods are target historical periods that are closer to the current period among the at least three target historical periods. For example, the current period is T, the at least three target historical periods are three target historical periods, and the three target historical periods are T-3, T-2, and T-1. The at least two adjacent historical periods are two adjacent historical periods. Since T-2 and T-1 are closer to T among T-3, T-2, and T-1, the two adjacent historical periods are T-2 and T-1, respectively.

[0060] Specifically, the at least three target historical periods include a third historical period, a second historical period, and a first historical period, and the at least two adjacent historical periods include the second historical period and the first historical period. The server may determine the object feature of the object in the first historical period as the object's first historical object feature, determine the object feature of the object in the second historical period as the object's second historical object feature, and determine the object feature of the object in the third historical period as the object's third historical object feature. The server may input the object's first historical object feature into an object classification model for category identification, predict the category to which the object belongs in the current period, and determine the predicted category to which the object belongs in the current period as the current predicted category. The server may input the object's second historical object feature into the object classification model for category identification, predict the category to which the object belongs in the first historical period, and determine the predicted category to which the object belongs in the first historical period as the first predicted category. The server may input the object's third historical object feature into the object classification model for category identification, obtain the category to which the object belongs in the second historical period, and determine the predicted category to which the object belongs in the second historical period as the second predicted category. Because the first and second historical periods belong to adjacent historical periods, the first predicted category and the second predicted category belong to the predicted category of the adjacent historical period.

[0061] Step 208 : Determine the classification accuracy of the object classification model in the current period based on the predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods.

[0062] The classification accuracy is used to reflect the accuracy of the classification performed by the object classification model, that is, the classification accuracy is used to reflect the accuracy of the predicted category. The greater the classification accuracy, the higher the accuracy of the category predicted by the object classification model.

[0063] The classification accuracy in the current cycle reflects the accuracy of the object classification model's prediction of the category to which the object belongs in the current cycle. The greater the classification accuracy in the current cycle, the more accurate the object classification model's prediction of the category to which the object belongs in the current cycle.

[0064] Specifically, the classification accuracy of the object classification model in the current period is affected by at least one of the predicted category or the actual category of the object in the adjacent historical period. The server may determine the classification accuracy of the object classification model in the current period based on the current predicted category, the first predicted category, and the second predicted category of each object.

[0065] In some embodiments, for each object, the server can obtain the real category to which the object belongs in the current period, the real category to which the object belongs in the first historical period, and the real category to which the object belongs in the second historical period, and count the number of objects based on the current predicted category, first predicted category, second predicted category, and each real category of each object to obtain the classification accuracy of the object classification model in the current period.

[0066] In some embodiments, after obtaining the classification accuracy, the server may adjust the parameters of the object classification model based on the classification accuracy, and classify the object based on the object classification model after the parameter adjustment. For example, the parameters of the object classification model may be continuously adjusted until the classification accuracy of the object classification model is greater than or equal to an accuracy threshold. The object classification model with a classification accuracy greater than or equal to the accuracy threshold is then used to classify the object. For example, the object features of the object in the current cycle are input into the object classification model to predict the category to which the object will belong in the next cycle after the current cycle. The accuracy threshold can be preset, for example, 90%.

[0067] In some embodiments, for multiple services provided in an application, such as a discounted refueling service, the steps in the above embodiments can be used to generate a target object classification model for each of the multiple services. The target object classification model refers to an object classification model with an accuracy greater than or equal to an accuracy threshold. The server can store the target object classification models corresponding to each service. The terminal can send a content promotion request for a specific object and a specific service in a specific application to the server. The server stores the target object classification model corresponding to the feature service of the specific application. In response to the content promotion request, the server can obtain the object features of the specific object in the current cycle, input the object features of the specific object in the current cycle into the target object classification model for classification, obtain the category to which the specific object belongs in the next cycle after the current cycle, determine the content to be pushed to the specific object based on the obtained category, and push the determined content to the specific object.

[0068] In the above-mentioned object classification processing method, an object classification model for analyzing classification accuracy is determined by obtaining historical object features of multiple objects in at least three target historical periods, where the at least three target historical periods are at least three consecutive historical periods selected from the period immediately preceding the current period. For each object, the historical object features of each object in each historical period are input into the object classification model for classification to predict the predicted category of each object in the next period of the historical period. The predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods are obtained, where the at least two adjacent historical periods are target historical periods that are closer to the current period among the at least three target historical periods. Based on the predicted category of each object in the current period and the predicted category of each object in the at least two adjacent historical periods, the classification accuracy of the object classification model in the current period is determined. Thus, the classification accuracy is determined by combining the predicted categories of the objects in multiple consecutive periods (including the current period), thereby improving the precision of the classification accuracy.

[0069] In some embodiments, at least two adjacent historical periods include a first historical period and a second historical period, the first historical period is adjacent to and before the current period, and the second historical period is adjacent to and before the first historical period; based on the predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods, determining the classification accuracy of the object classification model in the current period includes: obtaining the predicted category of each object in the current period to obtain the current predicted category of each object; obtaining the predicted category of each object in the first historical period to obtain the first predicted category of each object; obtaining the predicted category of each object in the second historical period to obtain the second predicted category of each object; determining the classification accuracy of the object classification model in the current period based on the current predicted category, the first predicted category and the second predicted category of each object.

[0070] The current predicted category of an object refers to the predicted category of the object in the current period. The first predicted category of an object refers to the predicted category of the object in the first historical period. The second predicted category of an object refers to the predicted category of the object in the second historical period.

[0071] Specifically, the server can obtain the true category of each object in each target historical period, count the number of objects based on the true category and predicted category of each object, and determine the classification accuracy of the object classification model in the current period based on the counted number. The true category includes a first preset category and a second preset category. The first preset category is different from the second preset category. For example, the first preset category is churn, and the second preset category is retention. The first preset category can be represented by 1, and the second preset category can be represented by 0, that is, 1 can be used to represent churn, and 0 can be used to represent retention.

[0072] In this embodiment, since the classification accuracy of the object classification model in the current period may be affected by the category to which the objects in the historical periods adjacent to the current period belong, the classification accuracy of the object classification model in the current period is determined based on the current prediction category, the first prediction category and the second prediction category of each object, thereby improving the precision of the classification accuracy of the object classification model in the current period.

[0073] In some embodiments, based on the current predicted category, the first predicted category, and the second predicted category of each object, determining the classification accuracy of the object classification model in the current period includes: counting the number of objects that meet the first category condition among multiple objects to obtain the first object number; the first category condition includes: the true category of the object in the current period is the first preset category, and the current predicted category of the object is the first preset category; counting the number of objects that meet the second category condition among multiple objects to obtain the second object number; the second category condition includes: the true category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the true category of the object in the first historical period is the first preset category, and the first predicted category of the object is the first preset category; based on the first object number and the second object number, determining the classification accuracy of the object classification model in the current period.

[0074] Specifically, the server may perform statistical calculation based on the number of the first objects and the number of the second objects to obtain the classification accuracy of the object classification model in the current cycle. The statistical calculation includes at least one of addition, subtraction, multiplication, or division.

[0075] In some embodiments, the server may select objects from the plurality of objects that satisfy the conditions that the actual category of the object in the second historical period is the first preset category and the second predicted category of the object is the second preset category to form a first object group, select objects from the plurality of objects that satisfy the conditions that the actual category of the object in the first historical period is the first preset category and the first predicted category of the object is the first preset category to form a second object group, and count the number of the same objects in the first object group and the second object group to obtain the second object number. For example, the first preset category is 1 and the second preset category is 0. Figure 4 As shown, region A is the objects whose real category is 1 and whose second predicted category is 0 in the second historical period, that is, region A is the first object group, region B is the objects whose real category is 1 and whose second predicted category is 1 in the first historical period, that is, region B is the second object group, and region C is the objects that meet the second category conditions, that is, the number of objects in region C is the second object number.

[0076] In some embodiments, the server may count the number of objects among a plurality of objects that meet a third category condition to obtain a third object number, where the third category condition includes: the object's actual category in the second historical period is the first preset category, the object's second predicted category is the second preset category, the object's actual category in the first historical period is the second preset category, and the object's first predicted category is the first preset category. The server may perform statistical calculations based on the first object number, the second object number, and the third object number to obtain the classification accuracy of the object classification model in the current period. The statistical calculations include at least one of addition, subtraction, multiplication, or division.

[0077] In this embodiment, since the process of counting the number of second objects is related to the category of the object in the first historical period and the second historical period, the classification accuracy of the object classification model in the current period is determined based on the number of first objects and the number of second objects. Therefore, the classification accuracy reflects the influence of the category of the object in the first historical period and the second historical period, thereby improving the precision of the classification accuracy.

[0078] In some embodiments, determining the classification accuracy of the object classification model in the current period based on the first object number and the second object number includes: counting the number of objects that meet the third category condition among multiple objects to obtain the third object number; the third category condition includes: the actual category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the actual category of the object in the first historical period is the second preset category, and the first predicted category of the object is the first preset category; statistics are performed based on the first object number and the second object number to obtain a positive statistical value; the positive statistical value is positively correlated with the first object number and the second object number; statistics are performed based on the first object number and the third object number to obtain a negative statistical value; the negative statistical value is positively correlated with the first object number and the third object number; based on the positive statistical value and the negative statistical value, determining the classification accuracy of the object classification model in the current period; the classification accuracy is positively correlated with the positive statistical value, and the classification accuracy is negatively correlated with the negative statistical value.

[0079] The positive statistic is positively correlated with the number of the first object and the number of the second object. The negative statistic is positively correlated with the number of the first object and the number of the third object. The classification accuracy is positively correlated with the positive statistic, and negatively correlated with the negative statistic.

[0080] A positive correlation means that, all other things remaining unchanged, two variables change in the same direction. When one variable changes from large to small, the other also changes from large to small. It's understandable that a positive correlation here means the changes are in the same direction, but it doesn't necessarily mean that a slight change in one variable necessarily requires a change in the other. For example, you could set variable b to 100 when variable a is between 10 and 20, and 120 when variable a is between 20 and 30. In this way, both a and b change in the same direction: when a increases, b also increases. However, within the range of a between 10 and 20, b doesn't need to change. A negative correlation means that, all other things remaining unchanged, two variables change in opposite directions: when one variable changes from large to small, the other changes from small to large. It's understandable that a negative correlation here means the changes are in opposite directions, but it doesn't necessarily mean that a slight change in one variable necessarily requires a change in the other.

[0081] Specifically, the server can select objects from the multiple objects that meet the conditions: the actual category of the object in the first historical period is the second preset category, and the first predicted category of the object is the first preset category, to form a third object group, count the number of objects that are the same in the first object group and the third object group, and obtain the third object number.

[0082] In some embodiments, the server may count the number of objects among the multiple objects that meet the seventh category condition to obtain the number of seventh objects, where the seventh category condition includes: the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category. The server may perform statistics based on the number of the first objects, the number of the second objects, and the number of the seventh objects to obtain a positive statistical value. The positive statistical value is positively correlated with the number of the seventh objects. The server may perform statistics based on the number of the first objects, the number of the third objects, and the number of the seventh objects to obtain a negative statistical value. The negative statistical value is positively correlated with the number of the seventh objects.

[0083] In some embodiments, the classification accuracy is positively correlated with the positive statistic and negatively correlated with the negative statistic. The server may calculate a ratio of the positive statistic to the negative statistic and use the calculated ratio to determine the classification accuracy of the object classification model in the current cycle.

[0084] In this embodiment, since the process of counting the number of third objects is related to the category of the object in the first historical period and the second historical period, the negative statistical value obtained by counting the number of first objects and the number of third objects is used to determine the classification accuracy of the object classification model in the current period based on the positive statistical value and the negative statistical value, which reflects the influence of the category of the object in the first historical period and the second historical period, and improves the precision of the classification accuracy.

[0085] In some embodiments, statistics are performed based on the number of first objects and the number of second objects to obtain a positive statistical value, which includes: counting the number of objects that meet the fourth category condition among multiple objects to obtain the fourth object number; the fourth category condition includes: the actual category of the object in the current period is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; statistics are performed based on the number of first objects, the number of second objects, and the number of fourth objects to obtain a positive statistical value; the positive statistical value is negatively correlated with the number of fourth objects.

[0086] Specifically, the server may select objects from the multiple objects that meet the conditions that the actual category of the object in the first historical period is the first preset category and the first predicted category of the object is the second preset category to form a fourth object group. The server may select objects from the multiple objects that are the actual category of the object in the current period to form a fifth object group. The server may count the number of objects that are the same in the fourth object group and the fifth object group to obtain the fourth object number. The server may select objects from the multiple objects that meet the conditions that the actual category of the object in the current period is the first preset category and the current predicted category of the object is the first preset category to form a first sub-object group, select objects from the multiple objects that meet the conditions that the actual category of the object in the current period is the first preset category and the current predicted category of the object is the second preset category to form a second sub-object group, and combine the first sub-object group with the second sub-object group to obtain a fifth object group.

[0087] In some embodiments, the server may count the number of objects among the multiple objects that meet an eighth category condition to obtain an eighth object number, where the eighth category condition includes: the actual category of the object in the current period is the first preset category, the current predicted category of the object is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category. The server may count the number of objects among the multiple objects that meet a ninth category condition to obtain a ninth object number, where the ninth category condition includes: the actual category of the object in the current period is the first preset category, the current predicted category of the object is the second preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category. The server may sum the eighth object number and the ninth object number to obtain a fourth object number.

[0088] In some embodiments, the server may perform statistics based on the first number of objects, the second number of objects, the fourth number of objects, and the seventh number of objects to obtain a positive statistical value. Specifically, the server may sum the first number of objects, the seventh number of objects, and the second number of objects, and perform a difference calculation between the sum and the fourth number of objects to obtain a positive statistical value. For example, the calculation method of the positive statistical value S1 is: S1 = (TP t +FN t-1 )-(TP t|t-1 +FN t|t-1 )+TP t-1|t-2 Among them, S1 represents the positive statistical value, TP t Indicates the number of first objects, FN t-1 Indicates the number of the seventh object, (TP t|t-1 +FN t|t-1 ) represents the fourth object number, TP t-1|t-2 Indicates the number of second objects, TP t|t-1 Indicates the eighth object number, FN t|t-1 Indicates the ninth object quantity.

[0089] In this embodiment, since the process of counting the number of the fourth objects is related to the category of the objects in the first historical period, statistics are performed based on the number of the first objects, the number of the second objects, and the number of the fourth objects to obtain positive statistical values, thereby improving the accuracy of the classification accuracy obtained based on the positive statistical values.

[0090] In some embodiments, the negative statistical value includes a first negative statistical value, and the classification accuracy includes a first classification accuracy; statistics based on the first object number and the third object number are performed to obtain the negative statistical value, including: counting the number of objects that meet the fifth category condition among multiple objects to obtain the fifth object number; the fifth category condition includes: the actual category of the object in the first historical period is the first preset category, the first predicted category of the object is the second preset category, the actual category of the object in the second historical period is the first preset category, and the second predicted category of the object is the second preset category; statistics based on the first object number, the third object number and the fifth object number are performed to obtain a first negative statistical value; the first negative statistical value is positively correlated with the fifth object number; based on the positive statistical value and the negative statistical value, determining the classification accuracy of the object classification model in the current period includes: determining the first classification accuracy of the object classification model in the current period based on the positive statistical value and the first negative statistical value.

[0091] Specifically, the negative statistic may include a first negative statistic, and the classification accuracy may include the first classification accuracy. The server may determine the first classification accuracy based on the positive statistic and the first negative statistic, wherein the first classification accuracy is positively correlated with the positive statistic, and the first classification accuracy is negatively correlated with the first negative statistic. For example, the server may calculate a ratio of the positive statistic to the first negative statistic, and determine the calculated ratio as the first classification accuracy. The server may perform a sum operation based on the first number of objects, the third number of objects, and the fifth number of objects to obtain the first negative statistic.

[0092] In some embodiments, the server may perform a statistical operation based on the first number of objects, the third number of objects, the fifth number of objects, and the seventh number of objects to obtain a first negative statistical value. The first negative statistical value is positively correlated with the seventh number of objects. Specifically, the server may perform a sum operation on the first number of objects, the third number of objects, the fifth number of objects, and the seventh number of objects to obtain the first negative statistical value.

[0093] In some embodiments, the server may count the number of objects among the plurality of objects that satisfy the conditions that: the actual category of the object in the first historical period is the second preset category, and the first predicted category of the object is the first preset category, to obtain a tenth number of objects. The first negative statistical value is positively correlated with the tenth number of objects, and the server may sum the first number of objects, the third number of objects, the fifth number of objects, the seventh number of objects, and the tenth number of objects to obtain the first negative statistical value.

[0094] In some embodiments, the server may count the number of objects among the plurality of objects that satisfy: the actual category of the object in the current period is the second preset category, and the current predicted category of the object is the first preset category, to obtain an eleventh number of objects. The first negative statistical value is positively correlated with the eleventh number of objects, and the server may sum the first number of objects, the third number of objects, the fifth number of objects, the seventh number of objects, the tenth number of objects, and the eleventh number of objects to obtain the first negative statistical value.

[0095] In this embodiment, since the process of counting the number of the fifth objects is related to the categories of the objects in the first historical period and the second historical period, statistics are performed based on the number of the first objects, the number of the third objects, and the number of the fifth objects to obtain a first negative statistical value, thereby improving the accuracy of the first classification accuracy obtained based on the first negative statistical value.

[0096] In some embodiments, statistics are performed based on the first number of objects, the third number of objects, and the fifth number of objects to obtain a first negative statistical value; the first negative statistical value is positively correlated with the fifth number of objects, including: counting the number of objects that meet the fourth category condition among multiple objects to obtain the fourth number of objects; the fourth category condition includes: the actual category of the object in the current period is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; statistics are performed based on the first number of objects, the third number of objects, the fourth number of objects, and the fifth number of objects to obtain a first negative statistical value; the first negative statistical value is negatively correlated with the fourth number of objects.

[0097] Specifically, the server may perform a sum operation on the first object quantity, the third object quantity, and the fifth object quantity, and perform a difference calculation between the summation result and the fourth object quantity to obtain a first negative statistical value.

[0098] In some embodiments, the server may sum the first object number, the third object number, the fifth object number, the seventh object number, the tenth object number, and the eleventh object number, and perform a difference calculation between the sum result and the fourth object number to obtain a first negative statistical value.

[0099] In this embodiment, since the process of counting the number of the fourth objects is related to the category of the objects in the first historical period, statistics are performed based on the number of the first objects, the number of the third objects, the number of the fourth objects, and the number of the fifth objects to obtain a first negative statistical value, thereby improving the accuracy of the first classification accuracy obtained based on the first negative statistical value.

[0100] In some embodiments, statistics are performed based on the first number of objects, the third number of objects, the fourth number of objects, and the fifth number of objects to obtain a first negative statistical value, including: counting the number of objects that meet the sixth category condition among multiple objects to obtain the sixth number of objects; the sixth category condition includes: the actual category of the object in the current period is the second preset category, the current predicted category of the object is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; statistics are performed based on the first number of objects, the third number of objects, the fourth number of objects, the fifth number of objects, and the sixth number of objects to obtain a first negative statistical value; the first negative statistical value is negatively correlated with the sixth number of objects.

[0101] Specifically, the server may perform a sum operation based on the first object number, the third object number, and the fifth object number to obtain a first summation result. For example, the server may perform a sum operation on the first object number, the third object number, the fifth object number, the seventh object number, the tenth object number, and the eleventh object number to obtain a first summation result. The server may perform a sum operation on the fourth object number and the sixth object number to obtain a second summation result, and perform a difference calculation on the first summation result and the second summation result to obtain a first negative statistical value. For example, the calculation formula of the first negative statistical value S2 is: S2=(TP t +FP t +FN t-1 +FP t-1 )-(TP t|t-1 +FN t|t-1 +FP t|t-1 )+(FN t-1|t-2 +FP t-1|t-2 ). Among them, S2 represents the first negative statistical value, FP t Indicates the number of eleven objects, FP t-1 Indicates the number of the tenth object, FN t-1|t-2 Indicates the number of fifth objects, FP t-1|t-2 Indicates the number of third objects, FP t|t-1 Indicates the sixth object quantity.

[0102] In this embodiment, since the process of counting the number of the sixth object is related to the category of the object in the first historical period, statistics are performed based on the number of the first objects, the number of the third objects, the number of the fourth objects, the number of the fifth objects, and the number of the sixth objects to obtain a first negative statistical value, thereby improving the accuracy of the first classification accuracy obtained based on the first negative statistical value.

[0103] In some embodiments, the negative statistical value includes a second negative statistical value, and the classification accuracy includes the second classification accuracy; performing statistics based on the first number of objects and the third number of objects to obtain the negative statistical value includes: performing statistics based on the first number of objects, the second number of objects, and the third number of objects to obtain the second negative statistical value; the second negative statistical value is positively correlated with the second number of objects; based on the positive statistical value and the negative statistical value, determining the classification accuracy of the object classification model in the current cycle includes: determining the second classification accuracy of the object classification model in the current cycle based on the positive statistical value and the second negative statistical value.

[0104] Specifically, the second classification accuracy is negatively correlated with the second negative statistical value. The second negative statistical value is positively correlated with the second number of objects. The server may perform statistics based on the first number of objects, the second number of objects, and the third number of objects to obtain the second negative statistical value. For example, the server may perform a sum calculation based on the first number of objects, the second number of objects, and the third number of objects to obtain the second negative statistical value.

[0105] In some embodiments, the negative statistical value includes a second negative statistical value, and the second negative statistical value is negatively correlated with the fourth object number. The server may perform statistics based on the first object number, the second object number, the third object number, and the fourth object number to obtain the second negative statistical value. Specifically, the server may sum the first object number, the second object number, and the third object number, and perform a difference calculation between the summed value and the fourth object number to obtain the second negative statistical value.

[0106] In some embodiments, the second negative statistical value is negatively correlated with the sixth number of objects, and the server may perform statistics based on the first number of objects, the second number of objects, the third number of objects, the fourth number of objects, and the sixth number of objects to obtain the second negative statistical value. Specifically, the server may sum the first number of objects, the second number of objects, and the third number of objects, perform a difference calculation on the summed value and the fourth number of objects, and perform a difference calculation on the difference calculated value and the sixth number of objects to obtain the second negative statistical value.

[0107] In some embodiments, the second negative statistical value is positively correlated with the number of the seventh object. The server may perform statistics based on the number of the first object, the number of the second object, the number of the third object, the number of the fourth object, the number of the sixth object, and the number of the seventh object to obtain the second negative statistical value. Specifically, the server may perform a sum calculation on the number of the first object, the number of the second object, the number of the third object, and the number of the seventh object, perform a difference calculation on the sum result and the number of the fourth object, and perform a difference calculation on the difference result and the number of the sixth object to obtain the second negative statistical value.

[0108] In some embodiments, the second negative statistical value is positively correlated with the tenth number of objects. The server may sum the first number of objects, the second number of objects, the third number of objects, the seventh number of objects, and the tenth number of objects, perform a difference calculation on the summed value and the fourth number of objects, and perform a difference calculation on the difference calculated and the sixth number of objects to obtain the second negative statistical value.

[0109] In some embodiments, the server can count the number of objects among the multiple objects that meet the conditions that the actual category of the object in the current cycle is the first preset category and the current predicted category of the object is the second preset category, and obtain the number of the twelfth object. The second negative statistical value is positively correlated with the number of the twelfth object. The server can sum the number of the first object, the number of the second object, the number of the third object, the number of the seventh object, the number of the tenth object, and the number of the twelfth object, perform a difference calculation on the result of the sum and the number of the fourth object, and perform a difference calculation on the result of the difference calculation and the number of the sixth object to obtain the second negative statistical value. For example, the calculation formula of the second negative statistical value S3 is: S3=(TP t +FN t +FN t-1 +FP t-1 )-(TP t|t-1 +FN t|t-1 +FP t|t-1 )+(TP t-1|t-2 +FP t-1|t-2 ). Among them, S3 represents the second negative statistical value, FN t Indicates the number of twelfth objects, FP t-1 Indicates the number of ten objects, FP t|t-1 Indicates the number of sixth objects, FP t-1|t-2 Indicates the number of third objects.

[0110] In some embodiments, the server may calculate the ratio of the positive statistic to the first negative statistic to obtain a first classification accuracy, for example, the first classification accuracy P t|t-1,t-2 Formula (1):

[0111]

[0112] Among them, P t|t-1,t-2 Indicates the first classification accuracy, P in formula (1) t|t-1,t-2 To reflect the proportion of objects whose predicted category is the first preset category in the current period under the influence of the first historical period and the second historical period, the P in formula (1) is t|t-1,t-2 It can also be called the precision rate, that is, the accuracy of the first classification can be the precision rate. The precision rate refers to the proportion of samples that are actually positive among the samples predicted to be positive. Prediction of positive means that the predicted category is the first preset category.

[0113] In some embodiments, the server may calculate the ratio of the positive statistic to the second negative statistic to obtain a second classification accuracy, for example, the second classification accuracy R t|t-1,t-2 Formula (2):

[0114]

[0115] Among them, R t|t-1,t-2 Represents the second classification accuracy, R in formula (2) t|t-1,t-2 It reflects the proportion of objects whose predicted category is the first preset category in the objects whose real category is the first preset category in the current period under the influence of the first historical period and the second historical period. Therefore, R in formula (2) t|t-1,t-2 It can also be called recall, that is, the second classification accuracy can be called recall. Recall refers to the proportion of samples that the model successfully predicted among the true positive samples. True positive means that the true class is the first preset class.

[0116] Let's take an example to illustrate the meaning of each parameter in formula (1) and formula (2). The first preset category is the loss label, which is represented by "1", and the second preset category is the retention label, which is represented by "0". The label refers to the category, then:

[0117] TP t (First object number): The number of objects that are predicted to be “1” in period T and the number of objects that are actually “1” in period T, i.e., TP t is the number of objects whose true category is “1” in T period and whose predicted category is “1” in T period;

[0118] FN t (Number of objects): The number of objects whose actual value is "1" in period T and whose predicted value is "0" in period T, i.e., FN t is the number of objects whose true category is “1” within T cycles and whose predicted category is “0” within T cycles;

[0119] FP t (Number of objects): The number of objects whose actual value is "0" in period T and whose predicted value is "1" in period T, i.e. FP t is the number of objects whose true category is “0” within T cycles and whose predicted category is “1” within T cycles;

[0120] FN t-1 (Seventh object number): The number of objects that are actually "1" in period T-1 and the number of objects that are predicted to be "0" in period T-1, that is, FN t-1 is the number of objects whose true category is “1” in the T-1 period and whose predicted category is “0” in the T-1 period;

[0121] FP t-1(Number of Tenth Objects): The number of objects whose actual value is “0” in period T-1 and whose predicted value is “0” in period T-1 are intersected, i.e., FP t-1 is the number of objects whose true category is “0” in the T-1 period and whose predicted category is “1” in the T-1 period;

[0122] TP t|t-1 (Eighth object number): The number of objects that are actually "1" in period T-1 and predicted to be "0" after intersection with the set of objects that are actually "1" and predicted to be "1" in period T, that is, TP t|t-1 is the number of objects that simultaneously meet the following four conditions: the true category in period T-1 is "1", the predicted category in period T-1 is "0", the true category in period T is 1, and the predicted category in period T is "1";

[0123] TP t-1|t-2 (Second number of objects): The number of objects whose actual value is "1" in period T-2 and whose value is predicted to be "0" in period T-2, and whose value is then intersected with the object whose actual value is "1" and whose value is predicted to be "1" in period T-1, i.e., TP t-1|t-2 is the number of objects that simultaneously meet the following four conditions: the true category in the T-2 period is "1", the predicted category in the T-2 period is "0", the true category in the T-1 period is "1", and the predicted category in the T-1 period is "1";

[0124] FN t|t-1 (Ninth object number): The number of objects whose actual value is "1" in period T-1 and whose predicted value is "0" in period T-1 after intersection with the object set whose actual value is "1" and whose predicted value is "0" in period T, i.e. FN t|t-1 is the number of objects that simultaneously meet the following four conditions: the true category is "1" in the T-1 period, the predicted category is "0" in the T-1 period, the true category is "1" in the T period, and the predicted category is "0" in the T period;

[0125] FN t-1|t-2 (Fifth object number): The number of objects that are actually "1" in period T-2 and predicted to be "0" after intersection with the set of objects that are actually "1" in period T-1 and predicted to be "0", that is, FN t-1|t-2 is the number of objects that simultaneously meet the following four conditions: the true category in the T-2 period is "1", the predicted category in the T-2 period is "0", the true category in the T-1 period is "1", and the predicted category in the T-1 period is "0";

[0126] FP t|t-1 (Sixth object number): The number of objects whose actual value is "1" in period T-1 and whose predicted value is "0" in period T-1 after intersection with the object set whose actual value is "0" in period T and whose predicted value is "1", i.e. FP t|t-1 The number of objects that simultaneously meet the following four conditions: the true category is "1" in the T-1 period, the predicted category is "0" in the T-1 period, the true category is "0" in the T period, and the predicted category is "1" in the T period.

[0127] FP t-1|t-2 (Third object number): The number of objects that are actually "1" in period T-2 and predicted to be "0" after intersection with the set of objects that are actually "0" in period T-1 and predicted to be "1", i.e. FP t-1|t-2 It is the number of objects that simultaneously meet the following four conditions: the true category in the T-2 period is "1", the predicted category in the T-2 period is "0", the true category in the T-1 period is "0", and the predicted category in the T-1 period is "1".

[0128] In this embodiment, the second classification accuracy of the object classification model in the current period is determined based on the positive statistical value and the second negative statistical value. Since the process of obtaining the second negative statistical value involves the categories of objects in the first historical period and the second historical period, it reflects the influence of the categories of objects in the first historical period and the second historical period on the second classification accuracy, thereby improving the precision of the second classification accuracy.

[0129] In some embodiments, the method further includes: when the classification accuracy is less than the accuracy threshold, determining the object classification model of the classification accuracy to be analyzed as the object classification model to be trained; training the object classification model to be trained to obtain a new object classification model of the classification accuracy to be analyzed, and returning to the step of inputting the historical object features of each object into the object classification model for classification until the classification accuracy reaches the accuracy threshold; determining the object classification model of the classification accuracy to be analyzed when the classification accuracy reaches the accuracy threshold as the trained object classification model.

[0130] The accuracy threshold can be preset or set as needed, for example, 85% or 90%.The object classification model to be trained is a model that needs to be trained or needs further training.

[0131] Specifically, after obtaining the classification accuracy of the object classification model in the current period, the server can compare the classification accuracy with an accuracy threshold. If the classification accuracy is determined to be less than the accuracy threshold, the object classification model is further trained. For example, the object classification model whose classification accuracy is to be analyzed can be determined as the object classification model to be trained. The object classification model to be trained is then trained. After training, the historical object features for each object in each historical period are returned and input into the object classification model for classification until the classification accuracy is equal to or greater than the accuracy threshold. The object classification model whose classification accuracy is equal to or greater than the accuracy threshold is determined to be the trained object classification model. The classification accuracy can include a first classification accuracy and a second classification accuracy. A classification accuracy less than the accuracy threshold means that at least one of the first and second classification accuracies is less than the accuracy threshold. For example, if the first classification accuracy is the precision rate and the second classification accuracy is the recall rate, a classification accuracy less than the accuracy threshold means that at least one of the precision rate and the recall rate is less than the accuracy threshold.

[0132] In some embodiments, after obtaining a trained object classification model, the server can use the trained object classification model to predict the category of an object in the next period after the current period. For example, the server can obtain the object features of the object in the current period, input the object features of the current period into the trained object classification model, and predict the category of the object in the next period after the current period.

[0133] In this embodiment, when the classification accuracy is less than the accuracy threshold, the object classification model with the classification accuracy to be analyzed is determined as the object classification model to be trained, and the object classification model to be trained is trained to obtain a new object classification model with the classification accuracy to be analyzed. When the classification accuracy is greater than the accuracy threshold, the object classification model with the classification accuracy to be analyzed is determined as the trained object classification model, thereby improving the classification accuracy of the object classification model.

[0134] In some embodiments, training the object classification model to be trained to obtain a new object classification model with classification accuracy to be analyzed includes: inputting the object features of the training object in the first historical period into the object classification model to be trained for classification, and obtaining the predicted category of the training object in the current period; based on the difference between the predicted category of the training object in the current period and the actual category of the training object in the current period, adjusting the parameters of the object classification model to be trained to obtain a new object classification model with classification accuracy to be analyzed.

[0135] The training object is the object to which the object features used by the training object classification model belong. There can be one or more training objects.

[0136] Specifically, the server can input the object features of the training object in the first historical period into the object classification model to be trained, obtain the predicted category of the training object in the current period, adjust the parameters of the object classification model to be trained based on the difference between the predicted category of the training object in the current period and the actual category of the training object in the current period, and determine the object classification model after the parameter adjustment as the new object classification model for classification accuracy to be analyzed.

[0137] In some embodiments, the server can obtain a target object set, which includes multiple objects, and can divide the target object set into a training object set and a test object set. The division can be random, for example, the target object set can be randomly divided into a training object set and a test object set according to a specific ratio. The specific ratio refers to the ratio between the number of training objects and the number of test objects. The specific ratio can be set or pre-set as needed. For example, the specific ratio of the number of training objects to the number of test objects is a:(1-a), where a represents the ratio between the number of training objects and the total number, and the total number is the number of objects included in the target object set. For example, if a is 0.8, the specific ratio is 0.8:0.2, that is, 8:2.

[0138] In some embodiments, the server can input the object features of the training object in the first historical period into the object classification model to be trained to obtain the predicted category of the training object in the current period, adjust the parameters of the object classification model to be trained based on the difference between the predicted category of the training object in the current period and the actual category of the training object in the current period, and obtain the object classification model after adjusting the parameters, obtain the object features of multiple test objects in the first historical period and input them into the object classification model after adjusting the parameters, and obtain the predicted category of each test object in the multiple test objects in the current period, and count the proportion of test objects in the multiple test objects whose predicted categories are consistent with the actual categories. When the calculated proportion is greater than the preset proportion (for example, 90%), the training is suspended, and the object classification model after adjusting the parameters is determined as the new object classification model for the classification accuracy to be analyzed. Otherwise, the training is continued.

[0139] In this embodiment, since the object features of the current period are relatively similar to the object features of the first historical period, the model is trained using the object features of the historical period closest to the current period, so that the trained object classification model is suitable for predicting the category of the next period based on the object features of the current period.

[0140] In some embodiments, the current cycle is the current promotion cycle for the target service, and the classification category of the trained object classification model is either retention or loss; the method also includes: obtaining object features of multiple candidate objects in the current promotion cycle; inputting the object features of each candidate object in the current promotion cycle into the trained object classification model for classification, and obtaining the predicted category of each candidate object in the next promotion cycle of the target service; selecting a target object with a predicted category of retention from each candidate object; and pushing promotional content related to the target service to the target object in the next promotion cycle.

[0141] The promotion cycle refers to the period during which content is pushed to an object. A promotion cycle can be, for example, a marketing cycle. The target service can be any service, including but not limited to at least one of a discounted refueling service, a car wash service, a car relocation service, or a designated driver service. The trained object classification model is used to identify whether an object is retained or churned. The candidate object can be any object. The candidate object can be the same or different from the object used to determine the classification accuracy of the object classification model in the current cycle. When the target service is a service provided by a target application, the candidate object can be a user registered with the target application. The target application can be any application. The next marketing cycle refers to the marketing cycle that is adjacent to and after the current marketing cycle. The predicted category of the candidate object is either retention or churn. For a period T, retention means that the object logged into the discounted refueling function module in period T-1 and did not log in during period T. Churn means that the object logged in to the discounted refueling function module in period T-1 and also logged in during period T. The target object is a candidate object with a predicted category of retention. Promotional content includes, but is not limited to, any one of advertisements or virtual resources. Virtual resources include, but are not limited to, at least one of coupons or red envelopes. When the promotion period is a marketing period, promotional content may also be referred to as marketing content.

[0142] Specifically, for each candidate object, the server can input the object features of the current marketing cycle into the trained object classification model for classification, and determine the classified category as the predicted category of the candidate object in the next marketing cycle of the target service. After obtaining the predicted category of each candidate object, the server can determine the candidate object with the predicted category of retention from each candidate object, and determine the candidate object with the predicted category of retention as the target object. In the next marketing cycle, the server can push promotional content related to the target service to the target object. For example, for a discounted refueling service, a coupon for refueling the vehicle can be pushed to the target object. Figure 5 , which shows a schematic diagram of sending promotional content to the terminal of the target object.

[0143] In this embodiment, the object features of each candidate object in the current marketing cycle are input into the trained object classification model for classification, and the predicted category of each candidate object in the next marketing cycle of the target service is obtained. The target object with the predicted category is selected from each candidate object to be retained, and in the next marketing cycle, promotional content related to the target service is pushed to the target object. Since the trained object classification model can predict the category more accurately, the accuracy of the screened target objects is improved, the precision of marketing is further improved, and the marketing cost is reduced.

[0144] The present application also provides an application scenario, which applies the above-mentioned object classification processing method.

[0145] like Figure 6 Specifically, the object classification processing method is applied in the application scenario as follows:

[0146] Step 602, for each object among the multiple objects, obtain the object characteristics of the object in the first marketing cycle of the target service, obtain the first object characteristics, obtain the object characteristics of the object in the second marketing cycle of the target service, obtain the second object characteristics, obtain the object characteristics of the object in the third marketing cycle of the target service, obtain the third object characteristics, obtain the object characteristics of the object in the current marketing cycle of the target service, and obtain the current object characteristics.

[0147] The first, second, and third marketing cycles are historical marketing cycles. The current, first, second, and third marketing cycles are consecutive cycles. The first marketing cycle is adjacent to and precedes the current marketing cycle. The second marketing cycle is adjacent to and precedes the first marketing cycle. The third marketing cycle is adjacent to and precedes the second marketing cycle. The duration of each marketing cycle can be the same or different. For example, each marketing cycle is 7 days long, the third marketing cycle is from January 1 to January 7, the second marketing cycle is from January 8 to January 14, the first marketing cycle is from January 15 to January 21, and the current marketing cycle is from January 22 to January 28.

[0148] Step 604: Obtain an object classification model to be trained, and determine a training object from various objects.

[0149] Step 606: Input the first object feature of the training object into the object classification model to be trained for classification, obtain the predicted category of the object in the current marketing cycle, obtain the training predicted category of the object, adjust the model parameters of the object classification model to be trained based on the training predicted category of the object, and obtain the object classification model whose classification accuracy is to be analyzed.

[0150] Step 608: Input the first object feature of each object into the object classification model of the classification accuracy to be analyzed for classification, obtain the predicted category of each object in the current marketing cycle, obtain the current predicted category of each object; input the second object feature of each object into the object classification model of the classification accuracy to be analyzed for classification, obtain the predicted category of each object in the first marketing cycle, obtain the first predicted category of each object; input the third object feature of each object into the object classification model of the classification accuracy to be analyzed for classification, obtain the predicted category of each object in the second marketing cycle, obtain the second predicted category of each object.

[0151] Step 610: Count the number of objects that meet a first category condition among multiple objects to obtain a first object number; the first category condition includes: the actual category of the object in the current period is the first preset category, and the current predicted category of the object is the first preset category; count the number of objects that meet a second category condition among multiple objects to obtain a second object number; the second category condition includes: the actual category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the first preset category; based on the first object number and the second object number, determine the classification accuracy of the object classification model in the current period.

[0152] Step 612 , determining whether the classification accuracy reaches the accuracy threshold, if not, executing step 614 , if yes, executing step 616 .

[0153] Step 614 : Determine the object classification model whose classification accuracy is to be analyzed as the object classification model to be trained, and return to step 606 .

[0154] In step 616, the object classification model whose classification accuracy is to be analyzed is determined to be the trained object classification model, and the current object features of the object are input into the trained object classification model for classification to obtain the predicted category of the object in the next marketing cycle of the current marketing cycle.

[0155] Step 618: Select a target object whose predicted category is retained from each object.

[0156] Step 620: Push marketing content related to the target service to the target object in the next marketing cycle of the current marketing cycle.

[0157] In this embodiment, the categories of objects in two historical periods adjacent to the current marketing cycle (i.e., the first historical period and the second historical period) are used to determine the classification accuracy of the object classification model, and the object classification model is further trained based on the classification accuracy, thereby improving the classification accuracy of the trained object classification model.

[0158] The object classification processing method provided in this application can be applied to scenarios related to preferential refueling services in travel services. In this scenario, the object classification model is a binary classification model, the category is a churn warning scene label, and the churn warning scene label is any one of the churn labels or retention labels, such as Figure 7 As shown, the object classification processing method provided in this application may include eight stages, namely, data processing stage, sample construction stage, model training and testing stage, evaluation label acquisition stage, second-order recursive confusion matrix construction stage, model evaluation stage, model prediction stage, and preferential refueling vehicle owner recommendation stage.

[0159] During the data processing phase, the owner's log data is obtained (the owner's log data is anonymized after obtaining the owner's consent) and processed into owner scene labels and owner feature data. The labels are constructed as follows:

[0160] For the label of period T: If the car owner logs in the preferential refueling function module in period T-1, but does not log in the preferential refueling function module in period T, it means that the car owner is a lost car owner in the preferential refueling module in period T, and the car owner's loss warning scene label Y in period T is set. t Marked as 1; if the car owner logs in to the preferential refueling function module in the T-1 period and also logs in to the preferential refueling function module in the T period, it means that the car owner is a retained car owner in the preferential refueling module in the T period, and the car owner's churn warning scenario label in the T period is Y t Marked as 0;

[0161] For the label of period T-1: If the car owner logs in to the preferential refueling function module in period T-2, but does not log in to the preferential refueling function module in period T-1, it means that the car owner is a lost car owner in the preferential refueling function module in period T-1, and the car owner's loss warning scene label Y in period T-1 is set. t-1 Marked as 1; if the car owner logs in to the preferential refueling function module in period T-2 and also logs in to the preferential refueling function module in period T-1, it means that the car owner is a retained car owner in the preferential refueling function module in period T-1, and the car owner's churn warning scenario label in period T-1 is Y t-1 Marked as 0;

[0162] For the label of T-2 period: If the car owner logs in the preferential refueling function module in T-3 period, but does not log in the preferential refueling function module in T-2 period, it means that the car owner is a churned car owner in the preferential refueling function module in T-2 period, and the car owner’s churn warning scene label Y in T-2 period will be set. t-2Marked as 1; if the car owner logs in to the preferential refueling function module in period T-3 and also logs in to the preferential refueling function module in period T-2, it means that the car owner is a retained car owner in the preferential refueling function module in period T-2, and the car owner's churn warning scenario label in period T-2 is Y t-2 The mark is 0. Among them, T-3, T-2, T-1, and T are consecutive cycles, and T is the current cycle.

[0163] The features are constructed as follows:

[0164] The owner log data is processed to obtain the owner's features (feature) {X t-i |i=0,1,2,3}. Among them, X t-3 represents the owner's characteristics in period T-3, X t-2 represents the owner's characteristics in period T-2, X t-1 represents the owner's characteristics in period T-1, X t represents the owner characteristics of the car owner in period T.

[0165] In the sample construction stage: the owner characteristics of period T-1 are X t-1 and the owner tag Y of period T t Construct the owner sample data S according to the owner ID (userid) matching tThe sample data is divided into training samples and test samples, and the sample data (both training and test samples) are divided into sparse features and dense features. Sparse features are processed using one-hot encoding, while dense features are processed using PCA (Principal Component Analysis), decorrelation, normalization, and feature discretization. Principal component analysis is a data dimensionality reduction algorithm that maps n-dimensional features to k-dimensional features. These k-dimensional features are newly orthogonal features, also known as principal components, and are reconstructed from the original n-dimensional features. The key idea of ​​one-hot encoding is to use an N-bit state register to encode N states. Each state is represented by its own independent register bit, and only one bit is active at any given time. For example, the one-hot encoding for the gender feature ["male", "female"] is: male -> 1 0; female -> 0 1. The one-hot encoding of the gas coupon type feature ["No. 92", "No. 95", "No. 98"] is: No. 92->1 0 0; No. 95->0 1 0; No. 98->0 0 1; the one-hot encoding of the gas coupon discount amount feature ["15 yuan", "25 yuan", "30 yuan", "50 yuan"] is: 15 yuan->1 0 0 0; 25 yuan->0 1 0 0; 30 yuan->0 0 1 0; 50 yuan->0 0 0 1. When a sample is ["male", "No. 95", "15 yuan"], the result of the one-hot encoding of the complete feature digitization is: [1, 0, 0, 1, 0, 1, 0, 0, 1]. The processed sparse features, dense features and car owner classification labels are randomly divided into training samples S according to a certain ratio. t train and test sample S t test Construct prediction sample: Use the feature X of T cycle t As prediction samples, the prediction samples are divided into sparse features and dense features. Among them, sparse features are processed with onehot, and dense features are processed with PCA decorrelation, normalization, and feature discretization. Construct the full sample of period T-1: Input the owner feature X of period T-2 t-2 and T-1 owner tag Y t-1 , build the full sample data S of the car owners in period T-1 according to the car owner identification matching t-1 Build the full sample of period T-2: Input the owner characteristics X of period T-3 t-3 and T-2 owner tag Y t-2 , build the full sample data S of the T-2 period car owners according to the car owner identification matching t-2 .

[0166] Model training and testing phase: input training sample S t train and test sample S t test A binary classification model is used to train and test the training and test samples of each scenario (e.g., the scenario of preferential refueling service). If the evaluation indicators (recall rate, precision rate, AUC (Area Under Curve) and other indicators) of period T meet the evaluation effect, the model weight vector W is saved respectively.

[0167] Obtain evaluation label stage: input model weight vector W, input T period full sample S t Using a binary classification algorithm and dividing labels according to a threshold of 0.5 (where the probability is greater than or equal to 0.5, it is recorded as 1; less than 0.5 is recorded as 0), the model evaluation index sequence E of the full sample of period T is obtained. t Similarly, input the full sample S of period T-1 t-1 , using the binary classification algorithm and dividing the labels according to the threshold of 0.5, the model evaluation index sequence E of the full sample of period T-1 is obtained t-1 . Input the full sample S of period T-2 t-2 , using a binary classification algorithm and dividing labels according to a threshold of 0.5, we obtain the model evaluation index sequence E for the full sample of period T-2 t-2 The model evaluation metric sequence includes the labels predicted by the model for each sample.

[0168] In the second-order recursive confusion matrix construction phase, a second-order recursive confusion matrix is ​​constructed. The second-order recursive confusion matrix includes the above formulas (1) and (2). Confusion matrix: A specific matrix used to visualize the performance of supervised learning algorithms. It summarizes the records in the data set according to the two criteria of the actual category and the classification judgment made by the classification model. Each column represents the predicted value and each row represents the actual category. Recursive confusion matrix: The calculation indicators (recall rate, precision rate) in the confusion matrix at time t are affected by the confusion matrix at time t-1. Second-order recursive confusion matrix: The calculation indicators (recall rate, precision rate) in the confusion matrix at time t are affected by the confusion matrices at time t-1 and time t-2.

[0169] Model evaluation phase: Input the model evaluation index sequence E of the full sample of period T t , the model evaluation index sequence E of the full sample of period T-1 t-1 , model evaluation index sequence E of the full sample of period T-2 t-2 , T period user label data Y t , T-1 period user label data Y t-1 , T-2 period user label data Y t-1Substituting into formulas (1) and (2), we can obtain the recall rate R of period T under the influence of period T-2 and period T-1 respectively. t|t-1,t-2 , precision rate P t|t-1,t-2 , and re-evaluate the model. If the model does not achieve the target performance (general experience, recall greater than or equal to 90%, precision greater than or equal to 85%), repeat steps 3 to 6 (i.e., model training and testing phase to model evaluation phase) until the model achieves the target performance.

[0170] In the model prediction phase, input the prediction sample X t , and the model W in the model training and testing phase. Using the binary classification algorithm, substitute the prediction sample X t And model W, get the predicted probability, and divide the label according to the threshold of 0.5. The probability greater than or equal to 0.5 is recorded as 1; less than 0.5 is recorded as 0. Complete the entire model training prediction.

[0171] In the stage of recommending owners with discounted refueling, the owner ID marked as 1 obtained by the model prediction stage is input, and the travel service mini program is used as a channel to distribute coupons for discounted refueling to these owner IDs, and the owner is notified of the distribution information via SMS or instant messaging.

[0172] Operational activities often have the situation where multiple periods of effects are superimposed under the same operational activity. Taking the refueling activity as an example, the operational strategy sets the coupon usage period to be valid within 21 days, and the activity period is set to 7 days, and each activity period carries out continuous operational activities (that is, after the 7-day activity ends, the operational activities of the next activity period will begin, and the effective period of each coupon is 21 days). Then the coupon validity period of period T will include the activity periods of periods T-1 and T-2, so that the operational activities of periods T-1 and T-2 have a positive impact on the operational activities of the current period (T). The sample data of periods T-1 and T-2 often affect the model effect of the current period (T), affect the evaluation of the current model by the sample data of the current period (T), affect the effect of the model prediction, and cannot accurately reflect the contribution of the current operational activity to the current model effect. In the embodiment of the present application, a second-order recursive confusion matrix is ​​used in the model evaluation, which can more accurately reflect the impact of removing the operational activity data of periods T-1 and T-2 on the model, and realize an accurate evaluation method of the model effect. For marketing activities of car networking refueling discounts, the method of issuing discount gas coupons or gift redemption to qualified car owners is often adopted. However, since the same operation activity often has the situation of overlapping effects of multiple periods of activities, that is, the model effect of period T is affected by periods T-1 and T-2, this solution uses rigorous mathematical derivation to derive the probability calculation model (such as formula (1) and formula (2)) of the recall rate and precision rate of period T under the influence of periods T-1 and T-2, and the impact on the model effect, so that the accurate evaluation method of the model effect can more accurately attribute the effect of the current operation activity to the current operation strategy. Formula (1) and formula (2) are proposed based on the second-order recursive confusion matrix. The calculation formulas obtained through rigorous data reasoning require creative labor to achieve.

[0173] The object classification processing method provided in this application can also be applied to the evaluation of model effects in multiple business scenarios and multiple activity cycles. It can effectively distinguish the model effects of each business scenario and each activity cycle, and can accurately reflect the effects of each business scenario and each activity cycle model. The object classification processing method provided in this application can be combined with marketing recommendation activities in various scenarios, can be equipped with a variety of machine learning algorithms and deep learning algorithms, and can be suitable for a variety of activity scenarios, with good scenario scalability. The object classification processing method provided in this application can be promoted and applied to business modules such as preferential refueling, preferential car washing, car moving code business, and designated driver business of travel services.

[0174] It should be understood that, although the steps in the flowcharts of the above embodiments are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0175] Based on the same inventive concept, embodiments of the present application also provide an object classification processing device for implementing the object classification processing method involved above. The implementation solution provided by this device is similar to the implementation solution described in the above method. Therefore, the specific limitations of one or more object classification processing device embodiments provided below can be referred to the limitations of the object classification processing method above, and will not be repeated here.

[0176] In some embodiments, as Figure 8 As shown, an object classification processing device is provided, including: a model determination module 802, a feature acquisition module 804, a category acquisition module 806 and an accuracy determination module 808, wherein:

[0177] A model determination module 802 is used to determine the object classification model whose classification accuracy is to be analyzed;

[0178] The feature acquisition module 804 is configured to acquire historical object features of a plurality of objects in at least three target historical periods; the at least three target historical periods are at least three consecutive historical periods selected from the period immediately preceding the current period;

[0179] Category obtaining module 806 is configured to input the historical object features of each object in each historical period into an object classification model for classification, thereby predicting the predicted category of each object in the next historical period, and obtaining the predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods; the at least two adjacent historical periods being target historical periods that are closest to the current period among the at least three target historical periods;

[0180] The accuracy determination module 808 is configured to determine the classification accuracy of the object classification model in the current period based on the predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods.

[0181] In some embodiments, at least two adjacent historical periods include a first historical period and a second historical period, the first historical period is adjacent to and before the current period, and the second historical period is adjacent to and before the first historical period; the accuracy determination module is further used to: obtain the predicted category of each object in the current period to obtain the current predicted category of each object; obtain the predicted category of each object in the first historical period to obtain the first predicted category of each object; obtain the predicted category of each object in the second historical period to obtain the second predicted category of each object; based on the current predicted category, the first predicted category and the second predicted category of each object, determine the classification accuracy of the object classification model in the current period.

[0182] In some embodiments, the accuracy determination module is further used to: count the number of objects that meet the first category condition among multiple objects to obtain the first object number; the first category condition includes: the actual category of the object in the current period is the first preset category, and the current predicted category of the object is the first preset category; count the number of objects that meet the second category condition among multiple objects to obtain the second object number; the second category condition includes: the actual category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the first preset category; based on the first object number and the second object number, determine the classification accuracy of the object classification model in the current period.

[0183] In some embodiments, the accuracy determination module is further used to: count the number of objects that meet the third category conditions among multiple objects to obtain the third object number; the third category conditions include: the actual category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the actual category of the object in the first historical period is the second preset category, and the first predicted category of the object is the first preset category; statistics are performed based on the number of first objects and the number of second objects to obtain a positive statistical value; the positive statistical value is positively correlated with the number of first objects and the number of second objects; statistics are performed based on the number of first objects and the number of third objects to obtain a negative statistical value; the negative statistical value is positively correlated with the number of first objects and the number of third objects; based on the positive statistical value and the negative statistical value, determine the classification accuracy of the object classification model in the current period; the classification accuracy is positively correlated with the positive statistical value, and the classification accuracy is negatively correlated with the negative statistical value.

[0184] In some embodiments, the accuracy determination module is further used to: count the number of objects among multiple objects that meet the fourth category conditions to obtain the fourth object number; the fourth category conditions include: the actual category of the object in the current period is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; statistics are performed based on the first object number, the second object number and the fourth object number to obtain a positive statistical value; the positive statistical value is negatively correlated with the fourth object number.

[0185] In some embodiments, the negative statistical value includes a first negative statistical value, and the classification accuracy includes a first classification accuracy; the accuracy determination module is also used to: count the number of objects that meet the fifth category conditions among multiple objects to obtain the fifth object number; the fifth category conditions include: the actual category of the object in the first historical period is the first preset category, the first predicted category of the object is the second preset category, the actual category of the object in the second historical period is the first preset category, and the second predicted category of the object is the second preset category; statistics are performed based on the first object number, the third object number and the fifth object number to obtain a first negative statistical value; the first negative statistical value is positively correlated with the fifth object number; based on the positive statistical value and the first negative statistical value, the first classification accuracy of the object classification model in the current period is determined.

[0186] In some embodiments, the accuracy determination module is further used to: count the number of objects among multiple objects that meet the fourth category conditions to obtain the fourth object number; the fourth category conditions include: the actual category of the object in the current period is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; statistics are performed based on the first object number, the third object number, the fourth object number and the fifth object number to obtain a first negative statistical value; the first negative statistical value is negatively correlated with the fourth object number.

[0187] In some embodiments, the accuracy determination module is further used to: count the number of objects among multiple objects that meet the sixth category condition to obtain the sixth object number; the sixth category condition includes: the actual category of the object in the current period is the second preset category, the current predicted category of the object is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; based on the first object number, the third object number, the fourth object number, the fifth object number and the sixth object number, statistics are performed to obtain a first negative statistical value; the first negative statistical value is negatively correlated with the sixth object number.

[0188] In some embodiments, the negative statistical value includes a second negative statistical value, and the classification accuracy includes a second classification accuracy; the accuracy determination module is also used to: perform statistics based on the first object number, the second object number and the third object number to obtain a second negative statistical value; the second negative statistical value is positively correlated with the second object number; based on the positive statistical value and the negative statistical value, determining the classification accuracy of the object classification model in the current cycle includes: determining the second classification accuracy of the object classification model in the current cycle based on the positive statistical value and the second negative statistical value.

[0189] In some embodiments, the device also includes: a first model determination module, which is used to determine the object classification model with the classification accuracy to be analyzed as the object classification model to be trained when the classification accuracy is less than the accuracy threshold; a second model determination module, which is used to train the object classification model to be trained to obtain a new object classification model with the classification accuracy to be analyzed, and return to the step of inputting the historical object features of each object into the object classification model for classification until the classification accuracy reaches the accuracy threshold; a third model determination module, which is used to determine the object classification model with the classification accuracy to be analyzed when the classification accuracy reaches the accuracy threshold as the trained object classification model.

[0190] In some embodiments, the second model determination module is also used to: input the object features of the training object in the first historical period into the object classification model to be trained for classification, and obtain the predicted category of the training object in the current period; based on the difference between the predicted category of the training object in the current period and the actual category of the training object in the current period, adjust the parameters of the object classification model to be trained to obtain a new object classification model with classification accuracy to be analyzed.

[0191] In some embodiments, the current cycle is the current promotion cycle for the target service, and the classification category of the trained object classification model is either retention or loss; the device is also used to: obtain object features of multiple candidate objects in the current promotion cycle; input the object features of each candidate object in the current promotion cycle into the trained object classification model for classification, and obtain the predicted category of each candidate object in the next promotion cycle of the target service; select the target object with the predicted category of retention from each candidate object; and push promotional content related to the target service to the target object in the next promotion cycle.

[0192] Each module in the object classification processing device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0193] In some embodiments, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 9 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store data involved in the object classification processing method. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an object classification processing method is implemented.

[0194] In some embodiments, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 10 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, an object classification processing method is implemented. The display unit of the computer device is used to form a visually visible image, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse, etc.

[0195] Those skilled in the art will understand that Figure 9 and Figure 10The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0196] In some embodiments, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned object classification processing method when executing the computer program.

[0197] In some embodiments, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned object classification processing method are implemented.

[0198] In some embodiments, a computer program product is provided, comprising a computer program, which implements the steps in the above-mentioned object classification processing method when executed by a processor.

[0199] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, storage, and display, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, in this application, the owner's log data is desensitized and used with the owner's consent.

[0200] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0201] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0202] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. An object classification processing method, characterized in that: The method comprises: Determine the classification accuracy of the object classification model to be analyzed; Obtaining historical object features of a plurality of objects in at least three target historical periods; the at least three target historical periods are at least three consecutive historical periods selected from the period preceding the current period; Inputting historical object features of each object in each target historical period into the object classification model for classification, thereby predicting a predicted category of each object in the next period of the target historical period, and obtaining a predicted category of each object in the current period and a predicted category of each object in at least two adjacent historical periods; the at least two adjacent historical periods being target historical periods that are closer to the current period among the at least three target historical periods; the at least two adjacent historical periods comprising: a first historical period that is prior to and adjacent to the current period, and a second historical period that is adjacent to and prior to the first historical period; A positive statistical value is obtained based on the statistics of the first object number and the second object number; the first object number is the number of objects among the multiple objects whose actual category in the current period is the first preset category and whose current predicted category is the first preset category; the second object number is the number of objects whose actual category in the second historical period is the first preset category and whose second predicted category is the second preset category, and whose actual category in the first historical period is the first preset category and whose first predicted category is the first preset category; the current predicted category of each of the objects is the predicted category of each of the objects in the current period; the first predicted category of each of the objects is the predicted category of each of the objects in the first historical period; and the second predicted category of each of the objects is the predicted category of each of the objects in the second historical period; A negative statistical value is obtained by performing statistics based on the first number of objects and the third number of objects; the third number of objects is the number of objects among the plurality of objects whose actual category in the second historical period is the first preset category and whose second predicted category is the second preset category, and whose actual category in the first historical period is the second preset category and whose first predicted category is the first preset category; The classification accuracy of the object classification model in the current cycle is determined according to the ratio of the positive statistical value to the negative statistical value.

2. The method according to claim 1, characterized in that The method further comprises: Counting the number of objects that meet a first category condition among the multiple objects to obtain a first number of objects; the first category condition includes: the actual category of the object in the current period is a first preset category, and the current predicted category of the object is the first preset category; Counting the number of objects among the multiple objects that meet a second category condition to obtain a second object number; the second category condition includes: the actual category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the first preset category.

3. The method according to claim 2, characterized in that The method further comprises: Counting the number of objects that meet a third category condition among the multiple objects to obtain a third number of objects; the third category condition includes: the actual category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the actual category of the object in the first historical period is the second preset category, and the first predicted category of the object is the first preset category; The positive statistical value is positively correlated with the number of the first objects and the number of the second objects; The negative statistical value is positively correlated with the number of the first objects and the number of the third objects; The classification accuracy is positively correlated with the positive statistical value, and the classification accuracy is negatively correlated with the negative statistical value.

4. The method according to claim 3, characterized in that The obtaining of a positive statistical value based on the number of first objects and the number of second objects includes: Counting the number of objects that meet a fourth category condition among the multiple objects to obtain a fourth number of objects, wherein the fourth category condition includes: the actual category of the object in the current period is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; Statistics are performed based on the first object number, the second object number, and the fourth object number to obtain a positive statistical value; the positive statistical value is negatively correlated with the fourth object number.

5. The method according to claim 3, characterized in that The negative statistic includes a first negative statistic, and the classification accuracy includes a first classification accuracy; and the negative statistic obtained by performing statistics based on the first number of objects and the third number of objects includes: Counting the number of objects among the multiple objects that meet a fifth category condition to obtain a fifth number of objects; the fifth category condition comprising: the actual category of the object in the first historical period is the first preset category, the first predicted category of the object is the second preset category, the actual category of the object in the second historical period is the first preset category, and the second predicted category of the object is the second preset category; Performing statistics based on the first number of objects, the third number of objects, and the fifth number of objects to obtain a first negative statistical value; the first negative statistical value is positively correlated with the fifth number of objects; Determining the classification accuracy of the object classification model in the current cycle according to the ratio of the positive statistical value to the negative statistical value includes: A first classification accuracy of the object classification model in the current cycle is determined according to a ratio of the positive statistic to the first negative statistic.

6. The method according to claim 5, characterized in that The obtaining of a first negative statistical value by performing statistics based on the first number of objects, the third number of objects, and the fifth number of objects includes: Counting the number of objects that meet a fourth category condition among the multiple objects to obtain a fourth number of objects, wherein the fourth category condition includes: the actual category of the object in the current period is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; Statistics are performed based on the first object number, the third object number, the fourth object number, and the fifth object number to obtain a first negative statistical value; the first negative statistical value is negatively correlated with the fourth object number.

7. The method according to claim 6, characterized in that The obtaining of a first negative statistical value by performing statistics based on the first number of objects, the third number of objects, the fourth number of objects, and the fifth number of objects includes: Counting the number of objects that meet a sixth category condition among the multiple objects to obtain a sixth number of objects, wherein the sixth category condition includes: the actual category of the object in the current period is the second preset category, the current predicted category of the object is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; Statistics are performed based on the first object number, the third object number, the fourth object number, the fifth object number, and the sixth object number to obtain a first negative statistical value; the first negative statistical value is negatively correlated with the sixth object number.

8. The method according to claim 3, characterized in that The negative statistic includes a second negative statistic, and the classification accuracy includes a second classification accuracy; and the negative statistic obtained by performing statistics based on the first number of objects and the third number of objects includes: Performing statistics based on the first number of objects, the second number of objects, and the third number of objects to obtain a second negative statistical value; the second negative statistical value is positively correlated with the second number of objects; Determining the classification accuracy of the object classification model in the current cycle according to the ratio of the positive statistical value to the negative statistical value includes: A second classification accuracy of the object classification model in the current cycle is determined according to a ratio of the positive statistical value to the second negative statistical value.

9. The method according to claim 1, characterized in that The method further comprises: In a case where the classification accuracy is less than the accuracy threshold, determining the object classification model with the classification accuracy to be analyzed as the object classification model to be trained; Training the object classification model to be trained to obtain a new object classification model whose classification accuracy is to be analyzed, and returning to the step of inputting the historical object features of each object into the object classification model for classification until the classification accuracy reaches an accuracy threshold; The object classification model whose classification accuracy reaches the accuracy threshold is determined as the trained object classification model.

10. The method according to claim 9, characterized in that The training of the object classification model to be trained to obtain a new object classification model whose classification accuracy is to be analyzed comprises: Inputting the object features of the training object in the first historical period into the object classification model to be trained for classification, and obtaining a predicted category of the training object in the current period; Based on the difference between the predicted category of the training object in the current cycle and the actual category of the training object in the current cycle, the parameters of the object classification model to be trained are adjusted to obtain a new object classification model whose classification accuracy is to be analyzed.

11. The method according to claim 9, characterized in that The current cycle is a current promotion cycle for the target service, and the classification category of the trained object classification model is either retention or churn; the method further includes: Obtain object features of multiple candidate objects in the current promotion cycle; Inputting the object features of each candidate object in the current promotion cycle into the trained object classification model for classification, and obtaining a predicted category of each candidate object in the next promotion cycle of the target service; Selecting a target object with a predicted category of retention from each of the candidate objects; In the next promotion cycle, promotional content related to the target service is pushed to the target object.

12. An object classification processing device, characterized in that: The device comprises: A model determination module, used to determine the object classification model to be analyzed for classification accuracy; A feature acquisition module is used to acquire historical object features of multiple objects in at least three target historical periods; the at least three target historical periods are at least three consecutive historical periods selected from the period before the current period; a category obtaining module for inputting historical object features of each object in each target historical period into the object classification model for classification, so as to predict a predicted category of each object in the next period of the target historical period, and obtain the predicted category of each object in the current period and the predicted category of each object in at least two adjacent historical periods; the at least two adjacent historical periods being target historical periods that are closer to the current period among the at least three target historical periods; the at least two adjacent historical periods comprising: a first historical period that is prior to and adjacent to the current period, and a second historical period that is adjacent to and prior to the first historical period; a statistical module configured to perform statistics based on a first number of objects and a second number of objects to obtain a positive statistical value; the first number of objects being the number of objects whose actual category in a current period is the first preset category and whose current predicted category is the first preset category; the second number of objects being the number of objects whose actual category in a second historical period is the first preset category, whose second predicted category is the second preset category, whose actual category in the first historical period is the first preset category, and whose first predicted category is the first preset category; and performing statistics based on the first number of objects and a third number of objects to obtain a negative statistical value; the third number of objects being the number of objects whose actual category in the second historical period is the first preset category, whose second predicted category is the second preset category, whose actual category in the first historical period is the second preset category, and whose first predicted category is the first preset category; the current predicted category of each object being the predicted category of each object in the current period; the first predicted category of each object being the predicted category of each object in the first historical period; and the second predicted category of each object being the predicted category of each object in the second historical period. The accuracy determination module is used to determine the classification accuracy of the object classification model in the current cycle according to the ratio of the positive statistical value to the negative statistical value.

13. The object classification processing device according to claim 12, characterized in that: The statistics module is also used to: Counting the number of objects that meet a first category condition among the multiple objects to obtain a first number of objects; the first category condition includes: the actual category of the object in the current period is a first preset category, and the current predicted category of the object is the first preset category; Counting the number of objects among the multiple objects that meet a second category condition to obtain a second object number; the second category condition includes: the actual category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the first preset category.

14. The object classification processing device according to claim 13, wherein: The statistics module is also used to: Counting the number of objects that meet a third category condition among the multiple objects to obtain a third number of objects; the third category condition includes: the actual category of the object in the second historical period is the first preset category, the second predicted category of the object is the second preset category, the actual category of the object in the first historical period is the second preset category, and the first predicted category of the object is the first preset category; The positive statistical value is positively correlated with the number of the first objects and the number of the second objects; The negative statistical value is positively correlated with the number of the first objects and the number of the third objects; The classification accuracy is positively correlated with the positive statistical value, and the classification accuracy is negatively correlated with the negative statistical value.

15. The object classification processing device according to claim 14, characterized in that: The statistics module is also used to: Counting the number of objects that meet a fourth category condition among the multiple objects to obtain a fourth number of objects, wherein the fourth category condition includes: the actual category of the object in the current period is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; Statistics are performed based on the first object number, the second object number, and the fourth object number to obtain a positive statistical value; the positive statistical value is negatively correlated with the fourth object number.

16. The object classification processing device according to claim 14, characterized in that: The negative statistic includes a first negative statistic, and the classification accuracy includes a first classification accuracy; and the statistical module is further configured to: Counting the number of objects among the multiple objects that meet a fifth category condition to obtain a fifth number of objects; the fifth category condition comprising: the actual category of the object in the first historical period is the first preset category, the first predicted category of the object is the second preset category, the actual category of the object in the second historical period is the first preset category, and the second predicted category of the object is the second preset category; Performing statistics based on the first number of objects, the third number of objects, and the fifth number of objects to obtain a first negative statistical value; the first negative statistical value is positively correlated with the fifth number of objects; The accuracy determination module is further configured to determine a first classification accuracy of the object classification model in the current cycle according to a ratio of the positive statistic to the first negative statistic.

17. The object classification processing device according to claim 16, characterized in that: The statistics module is also used to: Counting the number of objects that meet a fourth category condition among the multiple objects to obtain a fourth number of objects, wherein the fourth category condition includes: the actual category of the object in the current period is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; Statistics are performed based on the first object number, the third object number, the fourth object number, and the fifth object number to obtain a first negative statistical value; the first negative statistical value is negatively correlated with the fourth object number.

18. The object classification processing device according to claim 17, characterized in that: The statistics module is also used to: Counting the number of objects that meet a sixth category condition among the multiple objects to obtain a sixth number of objects, wherein the sixth category condition includes: the actual category of the object in the current period is the second preset category, the current predicted category of the object is the first preset category, the actual category of the object in the first historical period is the first preset category, and the first predicted category of the object is the second preset category; Statistics are performed based on the first object number, the third object number, the fourth object number, the fifth object number, and the sixth object number to obtain a first negative statistical value; the first negative statistical value is negatively correlated with the sixth object number.

19. The object classification processing device according to claim 14, characterized in that: The negative statistic includes a second negative statistic, and the classification accuracy includes a second classification accuracy; and the statistical module is further configured to: Performing statistics based on the first number of objects, the second number of objects, and the third number of objects to obtain a second negative statistical value; the second negative statistical value is positively correlated with the second number of objects; The accuracy determination module is further configured to determine a second classification accuracy of the object classification model in the current cycle according to a ratio of the positive statistical value to the second negative statistical value.

20. The object classification processing device according to claim 12, wherein: The device further comprises: The model training module is used to determine the object classification model with the classification accuracy to be analyzed as the object classification model to be trained when the classification accuracy is less than the accuracy threshold; train the object classification model to be trained to obtain a new object classification model with the classification accuracy to be analyzed, and return to the step of inputting the historical object features of each of the objects into the object classification model for classification until the classification accuracy reaches the accuracy threshold; and determine the object classification model with the classification accuracy to be analyzed when the classification accuracy reaches the accuracy threshold as the trained object classification model.

21. The object classification processing device according to claim 20, characterized in that: The model training module is also used to: Inputting the object features of the training object in the first historical period into the object classification model to be trained for classification, and obtaining a predicted category of the training object in the current period; Based on the difference between the predicted category of the training object in the current cycle and the actual category of the training object in the current cycle, the parameters of the object classification model to be trained are adjusted to obtain a new object classification model whose classification accuracy is to be analyzed.

22. The object classification processing device according to claim 20, characterized in that: The current cycle is a current promotion cycle for the target service, and the classification category of the trained object classification model is either retention or churn; the device further includes: A promotion module is used to obtain object features of multiple candidate objects in the current promotion cycle; input the object features of each candidate object in the current promotion cycle into the trained object classification model for classification, and obtain the predicted category of each candidate object in the next promotion cycle of the target service; select a target object with a predicted category of retention from each candidate object; and push promotional content related to the target service to the target object in the next promotion cycle.

23. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 11 is implemented.

24. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 11 is implemented.

25. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the method according to any one of claims 1 to 11 is implemented.

Images