Aging prediction method and device, electronic equipment, storage medium and computer program product

Abstract

The invention relates to an aging prediction method and device, electronic equipment, a storage medium and a computer program product, relates to the technical field of information, and can effectively consider aging prediction cost and prediction accuracy. The method comprises the steps of determining a target business object; obtaining factor value prediction models corresponding to a plurality of aging influence factors of the target business object; the factor value prediction model is used for predicting future values of the aging influence factors; the factor value prediction model is obtained through training according to historical values of a plurality of historical business objects of the same type as the target business object under the aging influence factors, and the historical values are determined according to historical business information of the historical business objects; according to a plurality of factor value prediction models, obtaining a future value of the target business object under the target aging influence factor; the target aging influence factor is at least one of a plurality of aging influence factors; and determining an aging prediction result of the target business object according to the future value under each target aging influence factor.

Description

Technical Field

[0001] The present disclosure relates to the field of information technology, and in particular, to a method, apparatus, electronic device, storage medium, and computer program product for timeliness prediction. Background Art

[0002] With the development of computer technology, various service providers can provide various service objects to users. In order to enable users to know in advance the completion time of a certain service object, timeliness prediction of the service object can be performed.

[0003] In related technologies, timeliness prediction services additionally provided by service providers can be obtained; alternatively, timeliness standards for each link in the service processing process can also be obtained, and subsequently, users can perform timeliness prediction according to the timeliness standards. Taking logistics timeliness prediction as an example, logistics timeliness can be predicted based on the line distance between the shipping address and the receiving address and the logistics timeliness standards provided by logistics enterprises.

[0004] However, the former requires high prediction costs, and although the latter reduces the prediction costs, the timeliness prediction accuracy also significantly decreases. It can be seen that it is difficult for related technologies to balance prediction costs and prediction accuracy when predicting the timeliness of service objects. Summary of the Invention

[0005] The present disclosure provides a method, apparatus, electronic device, storage medium, and computer program product for timeliness prediction, so as to at least solve the problem in related technologies that it is difficult to balance prediction costs and prediction accuracy when predicting the timeliness of service objects. The technical solutions of the present disclosure are as follows:

[0006] According to a first aspect of an embodiment of the present disclosure, a method for timeliness prediction is provided, including:

[0007] Determine a target service object for which business timeliness prediction is to be performed;

[0008] Obtain a factor value prediction model corresponding to each of a plurality of timeliness influencing factors of the target service object; the factor value prediction model is used to predict future values of the timeliness influencing factors; the factor value prediction model is trained based on historical values of a plurality of historical service objects of the same type as the target service object under the timeliness influencing factors, and the historical values are determined according to historical service information of the historical service objects;

[0009] According to the plurality of factor value prediction models, obtain the future values of the target service object under target timeliness influencing factors; the target timeliness influencing factors are at least one of the plurality of timeliness influencing factors;

[0010] Determine a timeliness prediction result of the target service object according to the future values under each of the target timeliness influencing factors.

[0011] In an exemplary embodiment, before obtaining the factor value prediction models corresponding to the respective aging influencing factors of the target business object, it further includes:

[0012] Obtain the historical business information of each of the multiple historical business objects;

[0013] According to each of the historical business information, determine the business object transfer data representing the transfer situation of each of the historical business objects between transfer nodes; the transfer nodes include at least one of the following: the geographical location where the historical business object is located, and the status information of the historical business object;

[0014] According to the analysis results of each of the business object transfer data, obtain the historical values corresponding to the respective aging influencing factors;

[0015] According to the historical values corresponding to each of the aging influencing factors, train the factor value prediction models corresponding to each of the aging influencing factors.

[0016] In an exemplary embodiment, after training the factor value prediction models corresponding to each of the aging influencing factors according to the historical values corresponding to each of the aging influencing factors, it further includes:

[0017] Determine the update period set in advance for the factor value prediction model;

[0018] Within each update period, according to the historical business information newly added in the update period, determine the new business object transfer data of each of the historical business objects;

[0019] Update the trained factor value prediction model according to the new business object transfer data.

[0020] In an exemplary embodiment, the obtaining of the future value of the target business object under the target aging influencing factor according to the multiple factor value prediction models includes:

[0021] Determine the aging prediction scenario for the target business object to perform aging prediction, and obtain the aging prediction quality information of each of the multiple aging influencing factor combination strategies associated with the aging prediction scenario; each of the aging influencing factor combination strategies includes at least one of the aging influencing factors, and the aging influencing factors included in different aging influencing factor combination strategies are at least partially different;

[0022] From the multiple aging influencing factor combination strategies, determine the target aging influencing factor combination strategy whose aging prediction quality information meets the prediction quality condition, and obtain the target aging influencing factor;

[0023] Determine a target factor value prediction model according to the factor value prediction model corresponding to the target timeliness influencing factor;

[0024] Obtain the future value of the target service object under the target timeliness influencing factor according to the target factor value prediction model.

[0025] In an exemplary embodiment, the target service object includes a target logistics object for which logistics timeliness prediction is to be performed; the factor value prediction model corresponding to the target timeliness influencing factor includes a plurality of logistics trajectory prediction models corresponding to logistics trajectory factors, and different logistics trajectory prediction models predict different levels of logistics node levels, and the logistics node level is the node level of the logistics node in the logistics trajectory prediction result;

[0026] The determining a target factor value prediction model according to the factor value prediction model corresponding to the target timeliness influencing factor includes:

[0027] According to the respective call priorities of a plurality of predetermined logistics trajectory prediction models, sequentially determine the currently used logistics trajectory prediction model from the plurality of logistics trajectory prediction models; the call priority decreases as the logistics node level increases, or increases as the prediction accuracy of the logistics trajectory prediction model increases;

[0028] Determine the target factor value prediction model according to the currently used logistics trajectory prediction model.

[0029] In an exemplary embodiment, the obtaining the future value of the target logistics object under the target timeliness influencing factor according to the target factor value prediction model includes:

[0030] Predict the logistics trajectory of the target logistics object according to the currently used logistics trajectory prediction model;

[0031] In the case where the current logistics trajectory prediction result does not meet the prediction quality condition, return to execute the step of sequentially determining the currently used logistics trajectory prediction model from the plurality of logistics trajectory prediction models according to the respective call priorities of the plurality of predetermined logistics trajectory prediction models until the current logistics trajectory prediction result meets the prediction quality condition or is output by the logistics trajectory prediction model with the lowest priority;

[0032] Determine the future value of the target logistics object under the logistics trajectory factor according to the current logistics trajectory prediction result.

[0033] According to a second aspect of the embodiments of the present disclosure, there is provided a timeliness prediction device, including:

[0034] A target business object determination unit, configured to determine a target business object for which business timeliness prediction is to be performed;

[0035] A model acquisition unit, configured to acquire a factor value prediction model corresponding to each of multiple timeliness influencing factors of the target business object; the factor value prediction model is used to predict future values of the timeliness influencing factors; the factor value prediction model is trained based on historical values of multiple historical business objects of the same type as the target business object under the timeliness influencing factors, and the historical values are determined according to the historical business information of the historical business objects;

[0036] An influencing factor information prediction unit, configured to acquire the future values of the target business object under a target timeliness influencing factor according to multiple factor value prediction models; the target timeliness influencing factor is at least one of the multiple timeliness influencing factors;

[0037] A timeliness prediction unit, configured to determine a timeliness prediction result of the target business object according to the future values under each target timeliness influencing factor.

[0038] According to a third aspect of the embodiments of the present disclosure, there is provided an electronic device, including:

[0039] A processor;

[0040] A memory for storing instructions executable by the processor;

[0041] Wherein, the processor is configured to execute the instructions to implement the timeliness prediction method as described in any one of the above.

[0042] According to a fourth aspect of the embodiments of the present disclosure, there is provided a computer-readable storage medium, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enabling the electronic device to execute the timeliness prediction method as described in any one of the above.

[0043] According to a fifth aspect of the embodiments of the present disclosure, there is provided a computer program product, the computer program product includes instructions, when the instructions are executed by a processor of an electronic device, enabling the electronic device to execute the timeliness prediction method as described in any one of the above.

[0044] The technical solutions provided by the embodiments of the present disclosure at least bring the following beneficial effects:

[0045] On the one hand, by using various factor value prediction models to predict the future values of the target time limit impact factors, it is possible to obtain the time limit prediction results by combining the future values of the target time limit impact factors in the absence of professional information provided by the business side, reducing the information dependence on external logistics enterprises and effectively reducing the time limit prediction cost. On the other hand, the factor value prediction model is trained based on the historical values of the time limit impact factors determined from the actually occurred historical business information, enabling the factor value prediction model to learn the variation law of the time limit impact factors and improving the accuracy of the future values output by the factor value prediction model and the accuracy of the time limit prediction results. Thus, it is possible to effectively balance the time limit prediction cost and the prediction accuracy.

[0046] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and do not limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings herein are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present disclosure and, together with the specification, are used to explain the principles of the present disclosure and do not constitute an improper limitation to the present disclosure.

[0048] Figure 1 is a flowchart of a time limit prediction method shown according to an exemplary embodiment.

[0049] Figure 2 is a flowchart of steps for training a factor value prediction model shown according to an exemplary embodiment.

[0050] Figure 3 is a schematic diagram of a process for training a factor value prediction model shown according to an exemplary embodiment.

[0051] Figure 4 is a flowchart of another time limit prediction method shown according to an exemplary embodiment.

[0052] Figure 5 is an application schematic diagram of a time limit prediction method shown according to an exemplary embodiment.

[0053] Figure 6 is a block diagram of a time limit prediction device shown according to an exemplary embodiment.

[0054] Figure 7 is a block diagram of an electronic device shown according to an exemplary embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] In order to enable those of ordinary skill in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0056] It should be noted that the terms "first", "second", etc. in the description, claims and above-mentioned drawings of the present disclosure are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present disclosure described herein can be implemented in an order different from those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0057] It should also 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 for display, data for analysis, etc.) involved in the present disclosure are all information and data authorized by the user or fully authorized by all parties.

[0058] In one embodiment, as Figure 1 shown, a timeliness prediction method is provided. In this embodiment, taking the application of this method to a server as an example, it can be understood that this method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0059] In step S101, determine the target service object for which service timeliness prediction is to be performed.

[0060] In specific implementation, the service object for which timeliness prediction is to be performed can be determined. For the convenience of distinction, this service object is referred to as the target service object. Among them, the service object can be any service that requires investment of time resources, such as logistics transportation service, broadband installation service, etc.

[0061] In some exemplary embodiments, taking service timeliness prediction as logistics timeliness prediction and the target service object as a logistics object as an example, the logistics object can be a resource object provided by an enterprise to an individual user through logistics, or a resource object transferred between an enterprise / resource network point and other enterprise / resource network points through logistics. For example, goods are transferred and traded between enterprise A and enterprise B through logistics, an enterprise allocates the goods it owns among its multiple stores, or a warehousing network point that provides warehousing services provides the goods it stores to an enterprise or an individual user.

[0062] In some exemplary embodiments, a certain business platform may perform business timeliness prediction on target business objects provided by other business platforms. For the sake of distinction, the business platform to perform business timeliness prediction is referred to as the first business platform, and the business platform that provides the target business objects is referred to as the second business platform. Among them, the second business platform can obtain the business information of the target business object in multiple links, and can restrict the access rights of the first business platform to the business information.

[0063] In step S102, obtain the factor value prediction models respectively corresponding to multiple timeliness influencing factors of the target business object; the factor value prediction model is used to predict the future values of the timeliness influencing factors; the factor value prediction model is trained according to the historical values of multiple historical business objects of the same type as the target business object under the timeliness influencing factors, and the historical values are determined according to the historical business information that can be queried for the historical business objects.

[0064] On the other hand, it is also possible to obtain the factor value prediction models respectively corresponding to multiple timeliness influencing factors.

[0065] Among them, the timeliness influencing factor refers to a factor that will affect the business timeliness of the business object. For business objects of different types or in different fields, there may be the same or different timeliness influencing factors.

[0066] In some embodiments, taking equipment installation as an example, its timeliness influencing factors may include factors that affect the equipment purchase completion time, and may also include factors that affect the efficiency of dispatching equipment installation personnel. In some other exemplary embodiments, taking the timeliness influencing factors of logistics objects as an example, it may include timeliness influencing factors related to logistics links, and may also include timeliness influencing factors related to non-logistics links. Among them, the timeliness influencing factors related to logistics refer to the factors that affect the logistics timeliness during the actual transfer process of the logistics object. For example, after the logistics object is shipped, one or more of the transportation route of the logistics object (such as the cities or locations passed through), the logistics network points involved in the transportation process (such as the operation sites set up by logistics enterprises), and the operation capabilities of the logistics network points (such as the site waves and site timeliness of the logistics network points); the timeliness influencing factors related to non-logistics links can be understood as the factors that affect the logistics timeliness before the logistics object starts to transfer. For example, the shipping time of the logistics object at the resource provider, the time when the logistics enterprise staff pick up the logistics object, and the operation capabilities of the warehousing network points providing warehousing services, etc.

[0067] When performing time efficiency prediction, the relevant technology can provide time efficiency prediction services by the business party, for example, the logistics company can directly provide the logistics time efficiency of the corresponding logistics object; it can also be based on the delivery information, combined with the network relationship, scheduling plan and other professional information provided by the logistics company, to determine the corresponding logistics route planning, and jointly predict the estimated delivery time of the logistics object on the route. However, although this method has high accuracy, the prediction cost is high, and a high cost needs to be paid to obtain the time efficiency prediction service or related information provided by the company. In other methods, time efficiency prediction can be performed according to the preset time efficiency standards of each link in the business processing process, but this method reduces the prediction cost while also reducing the quality of the logistics time efficiency prediction results.

[0068] In this regard, in the embodiment of the present application, the corresponding factor value prediction models can be trained in advance for multiple time-limited influencing factors of the business object. The factor value prediction model corresponding to each time-limited influencing factor can predict the future value of the time-limited influencing factor, wherein the value corresponding to the time-limited influencing factor can include the value result in the form of a numerical value or text under the time-limited influencing factor. For example, for the time-limited influencing factor "the delivery time of the logistics object at the resource provider", for merchant A, the corresponding value of the time-limited influencing factor is 12 hours, and for merchant B, the corresponding value of the time-limited influencing factor is 15 hours.

[0069] For the sake of distinction, in this application, the values ​​of the timeliness influencing factors determined based on historical data are referred to as historical values ​​of the timeliness influencing factors, and the values ​​of the timeliness influencing factors determined by the factor value prediction model are referred to as future values ​​of the timeliness influencing factors; wherein, the historical values ​​of the timeliness influencing factors can be historical data of the actual occurrence process, or can be the results obtained by statistical analysis of the historical data that actually occurred.

[0070] Specifically, multiple historical business objects of the same type as the target business object can be determined, wherein a historical business object refers to a business object that already has corresponding business information, and the business information may include a variety of information involved in the business object during the business processing. For ease of distinction, the business information of the historical business object is referred to as historical logistics information, wherein the historical business information may be all the information in the business process, such as all the logistics information corresponding to the historical logistics object that has been delivered to the designated location, or it may be part of the information in the business processing process, such as the historical logistics object that has not been delivered to the designated location, but the logistics object has generated and recorded relevant logistics information during the circulation process. In some embodiments, the historical business information may be accessible historical business information provided by the second business platform to the first business platform, that is, for historical business objects, the second business platform may selectively provide part or all of the historical business information to the first business platform.

[0071] Furthermore, historical values under multiple timeliness impact factors can be obtained based on historical business information, and then model training can be performed using the historical values under multiple timeliness impact factors to obtain a factor value prediction model corresponding to each timeliness impact factor. It should be emphasized that the historical logistics information used in each embodiment of this application is information and data that have been authorized by users or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0072] In step S103, according to multiple factor value prediction models, obtain the future value of the target business object under the target timeliness impact factor; the target timeliness impact factor is at least one of the multiple timeliness impact factors.

[0073] In practical applications, the trained factor value prediction model can learn the value change rule of the timeliness impact factor based on historical business information. Correspondingly, the future value of the timeliness impact factor can be predicted through the factor value prediction model. It can be understood that by training the factor value prediction model based on the historical value of the timeliness impact factor, the timeliness prediction problem can be solved at a lower cost in the absence of real business information provided by the second business platform, providing the timeliness prediction ability for relevant parties. For example, in the absence of the real route planning ability of logistics enterprises, the logistics timeliness can be predicted at a lower cost; at the same time, for some regions, such as remote regions with a low level of logistics industry development, even professional logistics enterprises have relatively low timeliness prediction quality for these regions. If a logistics enterprise is requested to provide timeliness services for these regions, additional resources need to be provided to obtain customized timeliness prediction services. However, through the factor value prediction model of this embodiment, various accurate and reliable high-quality future values can be provided through the factor value prediction model without the support of logistics enterprises, which helps to obtain the timeliness prediction ability for various regions at the lowest cost.

[0074] Furthermore, the future value of the target logistics object under the target timeliness impact factor can be predicted according to multiple factor value prediction models to determine the future values of each target timeliness impact factor related to the target business object. For example, for the target logistics object sent by a merchant to a user, the future delivery timeliness of the merchant, the logistics enterprise used, the route of the target logistics object from the starting point of transportation to the ending point of transportation, etc. can be predicted according to multiple factor value prediction models.

[0075] In step S104, determine the timeliness prediction result of the target business object according to the future values under each target timeliness impact factor.

[0076] After obtaining the future values under various target timeliness influencing factors, the timeliness calculation can be comprehensively performed based on the future values of various target timeliness influencing factors to obtain the timeliness prediction result of the target business object. Since the future values of the target timeliness influencing factors can include detailed content related to business timeliness, such as the delivery time of merchants, the pickup time of express delivery personnel, network capacity, shifts, and other information, in some alternative embodiments, the timeliness prediction result of the target business object can be calculated by combining the business timeliness prediction algorithm in related technologies and the future values of various target timeliness influencing factors.

[0077] In the above timeliness prediction method, the target business object to be predicted for business timeliness can be determined; the respective factor value prediction models corresponding to multiple timeliness influencing factors of the target business object are obtained, where the factor value prediction model is used to predict the future values of the timeliness influencing factors, and the factor value prediction model is trained based on the historical values of multiple historical business objects of the same type as the target business object under the timeliness influencing factors, and the historical values are determined according to the historical business information of the historical business objects; then, according to the multiple factor value prediction models, the future values of the target business object under the target timeliness influencing factors are obtained, and the target timeliness influencing factors are at least one of the multiple timeliness influencing factors; furthermore, according to the future values under each target timeliness influencing factor, the timeliness prediction result of the target business object is determined. In this embodiment, on the one hand, by using various factor value prediction models to predict the future values of the target timeliness influencing factors, it is possible to obtain the timeliness prediction result by combining the future values of the target timeliness influencing factors in the case of lacking professional information provided by the business party, reducing the information dependence on external logistics enterprises and effectively reducing the timeliness prediction cost. On the other hand, the factor value prediction model is trained based on the historical values of the timeliness influencing factors determined according to the actually occurred historical business information, enabling the factor value prediction model to learn the change rules of the timeliness influencing factors and improving the accuracy of the future values output by the factor value prediction model and the accuracy of the timeliness prediction result. Thus, it is possible to effectively balance the timeliness prediction cost and prediction accuracy.

[0078] In an exemplary embodiment, as Figure 2 shown, before step S102, the method can also obtain various factor value prediction models through the following steps:

[0079] In step S201, the respective historical business information of multiple historical business objects is obtained.

[0080] In this step, multiple historical business objects can be determined, and then, with the access permission to business information, the business information of each of the multiple historical business objects is obtained to obtain multiple historical business information.

[0081] Specifically, for e-commerce, one or more second business platforms may provide logistics objects. The first business platform, with the authorization of relevant parties, may obtain the historical logistics information of historical logistics objects based on the logistics documents or logistics tracking information uploaded by the second business platform for historical logistics objects. Exemplarily, the historical logistics information may be the logistics details of historical logistics objects.

[0082] In step S202, according to each piece of historical business information, determine the business object transfer data that characterizes the transfer of each historical business object between transfer nodes; the transfer nodes include at least one of the following: the geographical location where the historical business object is located, and the status information of the historical business object.

[0083] Among them, the historical business information may record each transfer node that the historical business object has passed through. The transfer node can be understood as a key position or state point where the business object is located during the business processing. In this embodiment, the transfer node may include the geographical location where the historical business object is located. For example, in the logistics business, the logistics object will pass through different locations during transportation, including one or more of the origin warehouse, transfer stations, destination distribution centers, etc. Thus, the transfer node can be obtained based on the geographical location of the historical business object, so as to understand the transportation track of the logistics object. In addition, the transfer node may also include the status information of the historical business object. For example, the status of the business object may switch among states such as ordered, shipped, in delivery, etc. Furthermore, according to the historical business information corresponding to each historical business object, the business object transfer data that can characterize the transfer situation of each historical business object can be obtained. The business object transfer data can reflect how the historical business object moves and changes between each transfer node.

[0084] For example, for a historical logistics object that has completed at least one logistics activity in the past, the real logistics object transfer data can be learned based on its historical logistics information. The logistics object transfer data may include at least one piece of information such as the logistics enterprise that delivers the historical logistics object, the shipping address, the receiving address, the real pickup time of the historical logistics object, the real signing time of the historical logistics object, the network path, the transfer time, etc.

[0085] In step S203, according to the analysis results of each business object transfer data, obtain the historical values corresponding to each of the multiple timeliness influencing factors.

[0086] It can be understood that the business object transfer data will include the specific content of at least one timeliness influencing factor related to business timeliness, such as the delivery time, the time when the courier picks up the logistics object, etc., which affect the logistics timeliness information. In this regard, in this step, after obtaining the business object transfer data of each business object, the business object transfer data of each business object can be analyzed and split to obtain the historical values corresponding to each timeliness influencing factor. When analyzing, the timeliness influencing factor can be used as a single dimension for analysis and splitting. For example, according to all the business object transfer data, the respective historical values of the timeliness influencing factor A are determined; of course, the historical values of each timeliness influencing factor under the sub-categories can also be obtained. Taking the logistics object as an example, according to all the logistics object transfer data, the historical values corresponding to the merchant delivery time under different categories (such as fresh products and daily necessities) can be determined. Another example is that according to all the logistics object transfer data, the pick-up time of the logistics object of different logistics enterprises can be determined.

[0087] In step S204, according to the historical values corresponding to each timeliness influencing factor, a factor value prediction model corresponding to each timeliness influencing factor is trained.

[0088] After obtaining the historical values corresponding to each timeliness influencing factor, a factor value prediction model corresponding to each timeliness influencing factor can be trained according to each historical value. In some exemplary embodiments, the historical values under a certain timeliness influencing factor can be used alone to train the factor value prediction model corresponding to the timeliness influencing factor, or the historical values under multiple timeliness influencing factors can be used to train the factor value prediction model corresponding to a timeliness influencing factor. For example, for the timeliness influencing factors A and B with an associated relationship, the historical values under the timeliness influencing factor A and the historical values under the timeliness influencing factor B can be used to train the factor value prediction model corresponding to the timeliness influencing factor A.

[0089] In some exemplary embodiments, a standard model can be established, and then, using the historical values corresponding to each timeliness influencing factor, the standard model is connected in series respectively to obtain the factor value prediction model corresponding to each timeliness influencing factor. Among them, the standard model can be various machine learning models in the related art, such as any one of a neural network model, a linear regression model, a logistic regression model, a decision tree model, a random forest model, a support vector machine model, and a time series prediction model.

[0090] For example, as Figure 3As shown in the figure, taking the two factor value prediction models of the delivery time prediction model and the actual logistics track feedback model as examples, the data source is obtained first. For the delivery time prediction model, its data source is the delivery time data of express delivery outlets obtained according to historical logistics information. For the actual logistics track feedback model, its data source is the logistics track (such as urban routes or logistics outlet tracks) obtained according to historical logistics information. Then, the data source is classified and labeled, and data cleaning is performed to obtain the processed original data. The original data is divided into a training data set and a test data set. After that, machine learning can be performed using the training data set to enable the factor value prediction model to learn the Gaussian distribution of the original data. Then, simulation verification is performed by the simulation system. Among them, simulation verification means inputting the classified and labeled test data set into the trained factor value prediction model, and comparing and verifying the model output result with the label corresponding to the test data. After the verification passes, a business model applicable to a specific business scenario can be obtained, such as a delivery time prediction model for predicting delivery time and an actual logistics track feedback model for predicting the actual logistics track.

[0091] In this embodiment, on the one hand, by learning the business object transfer data that has occurred in the past to obtain the factor value prediction model, the problem of time prediction can be solved at low cost without relying on the enterprise to provide real business information. On the other hand, by obtaining the historical values corresponding to each time impact factor according to the analysis results of the business object transfer data of each business object and training the factor value prediction models corresponding to each time impact factor, it helps to perform time prediction calculations in multiple dimensions subsequently and increase business coverage.

[0092] In an exemplary embodiment, after step S204, the following steps may further be included:

[0093] Determine the update period set in advance for the factor value prediction model; within each update period, determine the new business object transfer data of each historical business object according to the historical business information newly added in the update period; update the trained factor value prediction model according to the new business object transfer data.

[0094] In specific implementation, the business may be abnormal due to some abnormal or extreme situations. For example, in the logistics business, when extreme weather or special festivals come, there may be an obvious phenomenon in the logistics status. If the previous business object transfer data is continued to be used to predict the future values of the time impact factors, such as predicting the logistics time after extreme weather according to the logistics object transfer data before extreme weather, or predicting the logistics time after the end of a special festival (such as a promotion festival for e-commerce) according to the logistics object transfer data of the special festival, inaccurate predictions may occur.

[0095] In this regard, in this embodiment, an update period set in advance for the factor value prediction model can be determined. The update period can be set according to the actual situation, so that the factor value prediction model can be updated regularly according to the update period. Among them, the update periods of the multiple factor value prediction models can be the same or different. For example, the update period can be determined according to the time interval at which various influencing factor information changes regularly. The shorter the time interval, the shorter the update period. For example, the update period of the factor value prediction model corresponding to the timeliness influencing factor A is 7 days, while the update period of the factor value prediction model corresponding to the timeliness influencing factor B is 1 month.

[0096] Furthermore, within each update period, the server can obtain the newly added historical business information during the update period and obtain the new business object transfer data of each historical business object based on the newly added historical business information. Subsequently, steps S202 to S204 can be executed according to the new business object transfer data, thereby realizing the model update of the factor value prediction model. Through the above processing, the factor value prediction model can timely learn the newly added historical values of the timeliness influencing factors. For example, after a logistics enterprise adds a network line, the new logistics object transfer data will also be discovered and learned by the model and put into actual use for timeliness prediction in combination with the new network line. Another example is that when extreme weather, natural disasters, festival activities, or the start or end of line control occur, the affected line range can be timely reflected by the model, and the corresponding expression content can be controlled through model update and quality inspection to minimize the impact on users to the greatest extent.

[0097] In some examples, the process of each model update can be as Figure 3 shown, that is, according to the new business object transfer data, steps such as data acquisition, preprocessing, model training, and verification are performed again. In addition, the coverage rate can be used to compare model versions, and relevant warning capabilities can be provided for the trained factor value prediction model. For example, when it is found that the prediction deviation reaches a preset threshold, the staff is prompted to manually trigger the iterative update of the model. Thus, through algorithm rolling, simulation inspection, quality calculation, etc., the service scope of timeliness prediction can be dynamically affected, avoiding the negative impact on users caused by inaccurate timeliness prediction due to some extreme situations.

[0098] In this embodiment, the model iteration of each factor value prediction model can be performed periodically, so that the factor value prediction model can learn the latest business object transfer data in a timely manner and learn the changing rules of new timeliness impact information, and can continuously perform statistical learning with real business data, maintaining that the factor value prediction model can reflect the real situation in a timely manner, and improving the accuracy and reliability of each factor value prediction model; at the same time, by regularly updating the factor value prediction model, the entire link can be continuously iterated autonomously, and the most real line conditions can be reflected in a timely manner, so that the overall solution provided by this application can run stably and self-sufficiently for a long time, and sudden natural disasters, business activities, management and control, etc. can be reflected in the system in a timely manner during the use process, effectively improving the stability and reliability of the timeliness prediction method of this application.

[0099] In an exemplary embodiment, after step S203, the method may further include the following steps:

[0100] Determine the detection indicators for detecting the timeliness prediction quality of the factor value prediction model, and obtain the index information of each factor value prediction model under the detection indicators; according to the index information of each factor value prediction model, determine the model evaluation information of each factor value prediction model.

[0101] In practical applications, the detection indicators can be set in advance. The factor value prediction model outputs the future values under the corresponding timeliness impact factors. It can be understood that the future values output by the factor value prediction model affect the accuracy of the logistics timeliness prediction result. In this regard, the detection indicators for detecting the timeliness prediction quality of the factor value prediction model can be set in this embodiment. Furthermore, for each factor value prediction model, the index information of the information detection model under the detection indicators can be calculated according to the future values output by the factor value prediction model. In some examples, the future values output by different factor value prediction models are information in different dimensions, and the index information under the same detection indicators can be obtained by calculating and transforming the future values.

[0102] Furthermore, the model evaluation information of each factor value prediction model can be determined according to the index information of each factor value prediction model. Among them, the model evaluation information can be information characterizing the high or low timeliness prediction quality of the factor value prediction model. In some embodiments, the index information can be directly used as the model evaluation information, or the index information of each factor value prediction model can be compared, and the model evaluation information can be obtained according to the comparison result.

[0103] It can be understood that since each model evaluation information can be index information obtained under the same detection index, by obtaining the index information of each factor value prediction model under the detection index, for various types of factor value prediction models, comparable model evaluation information can ultimately be obtained, which helps to better apply each factor value prediction model to various business scenarios according to the model evaluation information. For example, in step S103, a target factor value prediction model whose model evaluation information meets the conditions can be determined from multiple factor value prediction models, and according to the target factor value prediction model, the future value of the target logistics object under the target timeliness impact factor can be obtained. Also, for example, after model training or model update, the model evaluation information can be obtained to perform a quality check on the trained factor value prediction model, improving the reliability of each factor value prediction model.

[0104] In an exemplary embodiment, determining a detection index for detecting the timeliness prediction quality of a factor value prediction model may include the following steps:

[0105] Obtain a detection index for detecting the timeliness prediction quality of a factor value prediction model according to the timeliness indicator and / or the timeliness accuracy indicator.

[0106] In a specific implementation, at least one of the timeliness indicator and the timeliness accuracy indicator can be set. Among them, the timeliness indicator can indicate whether the predicted delivery time obtained according to the future value output by a certain factor value prediction model is timely. For example, for the future value output by a certain factor value prediction model, whether the predicted delivery time obtained according to the future value is before the actual delivery time. If so, it can be determined that the timeliness indicator is met; if not, it is determined that the timeliness indicator is not met.

[0107] The timeliness accuracy indicator can indicate whether the predicted delivery time obtained according to the future value output by a certain factor value prediction model is accurate. For example, for the future value output by a certain factor value prediction model, whether the predicted delivery time obtained according to the future value is on the same day as the actual delivery time. If so, it can be determined that the timeliness accuracy indicator is met; if not, it is determined that the timeliness accuracy indicator is not met.

[0108] In this embodiment, based on the above indicators, the factor value prediction model can be conveniently iterated multiple times and continuously fitted in layers. While effectively detecting the impact of the factor value prediction model on the timeliness prediction result, it helps to obtain a factor value prediction model that meets the expected timeliness prediction quality and apply it to the business.

[0109] In an exemplary embodiment, in step S103, obtaining the future value of the target business object under the target timeliness impact factor according to multiple factor value prediction models may include the following steps:

[0110] Determine the aging prediction scenario for predicting the aging of the target business object, and obtain the aging prediction quality information of each of the multiple aging influence factor combination strategies associated with the aging prediction scenario; each aging influence factor combination strategy includes at least one aging influence factor, and the aging influence factors included in different aging influence factor combination strategies are at least partially different; from the multiple aging influence factor combination strategies, determine the target aging influence factor combination strategy whose aging prediction quality information meets the prediction quality condition, and obtain the target aging influence factor; according to the factor value prediction model corresponding to the target aging influence factor, determine the target factor value prediction model; according to the target factor value prediction model, obtain the future value of the target business object under the target aging influence factor.

[0111] In practical applications, the multiple factor value prediction models obtained through pre-training can be independent of each other, that is, each factor value prediction model can be used alone and the future value under the corresponding aging influence factor can be predicted through a separate factor value prediction model. At the same time, there can be multiple aging prediction scenarios for the logistics aging prediction of the target logistics object. Among them, the aging prediction scenario can be understood as the timing (or scenario) for performing the aging prediction. According to the change of the user of the aging prediction result or the different time of the prediction, multiple aging prediction scenarios can be determined.

[0112] For example, for the aging prediction of logistics objects, taking the aging prediction of logistics by individual users and enterprises as an example, the aging prediction when an individual user purchases a commodity and the aging prediction when an enterprise schedules the commodity to different stores are different aging prediction scenarios, and the aging influence factors involved are different. Another example is that the aging prediction of logistics that the platform can provide for an individual user at the time point after the individual user places an order and before the merchant ships the goods, and the aging prediction of logistics after the merchant ships the goods and before the user signs for it are also different aging prediction scenarios. The former needs to predict the merchant's shipping time, while the latter does not.

[0113] In this regard, in this embodiment, multiple aging influence factor combination strategies can be set in advance for each aging prediction scenario. In each aging influence factor combination strategy, at least one aging influence factor that affects the prediction quality of the aging prediction result in the aging prediction scenario can be included, and the aging influence factors included in different aging influence factor combination strategies are at least partially different. In some embodiments, the aging influence factors in the aging influence factor combination strategy can be key aging influence factors, and the degree of influence of the key aging influence factors on the prediction quality is greater than a preset threshold. For example, when considering and not considering the key aging influence factors, there are obvious differences in the aging prediction quality information of the aging prediction results (such as the difference in accuracy rate and / or timeliness rate before and after is greater than the threshold).

[0114] After determining the time limit prediction scenario for predicting the logistics time limit of the target business object, the time limit prediction quality information of each combination strategy of time limit influencing factors can be obtained. Among them, the time limit prediction quality information can be information reflecting the quality of the business time limit prediction result. For example, it can be the accuracy of the time limit prediction result. In some embodiments, as the factor value prediction model under the time limit influencing factors is updated, the time limit prediction quality information of the combination strategy of time limit influencing factors will also change accordingly. In this regard, after the factor value prediction model corresponding to the time limit influencing factors in the combination strategy of time limit influencing factors is updated, the time limit prediction quality information of the combination strategy of time limit influencing factors can be updated accordingly.

[0115] Furthermore, from multiple combination strategies of time limit influencing factors, the target combination strategy of time limit influencing factors whose time limit prediction quality information meets the prediction quality conditions can be determined to obtain the target time limit influencing factors. Then, according to the target factor value prediction model, the future value of the target business object under the target time limit influencing factors corresponding to the target factor value prediction model can be obtained.

[0116] In this embodiment, the factor value prediction models of multiple time limit influencing factors can be used separately. By presetting multiple combination strategies of time limit influencing factors for the time limit prediction scenario and obtaining the appropriate combination strategy of time limit influencing factors according to the time limit prediction quality information of the combination strategy of time limit influencing factors, and predicting according to the factor value prediction model of the target time limit influencing factors in the strategy, it is possible to flexibly combine each factor value prediction model according to various time limit prediction scenarios, avoid mechanically predicting the business time limit of various time limit prediction scenarios in the same way, effectively improve the accuracy of the business time limit prediction results under different time limit prediction scenarios, and at the same time save the redundant calculation overhead in the business time limit prediction process.

[0117] In an exemplary embodiment, the target business object includes the target logistics object to be predicted for the logistics time limit; the factor value prediction model corresponding to the target time limit influencing factors includes multiple logistics trajectory prediction models for predicting the logistics trajectory. Specifically, the logistics trajectory can be understood as a sequence composed of multiple logistics nodes in turn. Among them, each logistics node can be a geographical location or area that actually exists. Exemplarily, the multiple logistics nodes in the logistics trajectory can include at least one type of logistics network point and geographical area. Among them, the geographical area can be various administrative divisions (such as cities, administrative regions, or streets, etc., one or more), or various areas delimited by staff according to business needs. The logistics trajectory can be composed of the same or different types of logistics nodes. In this regard, for different granularities of logistics node division, in this embodiment, various factor value prediction models for predicting the logistics trajectory, that is, various logistics trajectory prediction models, can be trained respectively.

[0118] For example, the routing factor value prediction model may include a route model and a network point model. The route model predicts the logistics track in terms of the "city" dimension. For example, for a target logistics object transported from address 1 to address 2, the cities A, B, and C it passes through can be predicted through the route model. When training the route model, route statistics can be performed on logistics enterprises, shipping addresses, receiving addresses, and their corresponding passing cities, and information such as actual timeliness, fitted timeliness, and exception statistics can be determined. The network point model predicts the logistics track in terms of the "logistics network point" dimension. It can be understood that multiple logistics network points can be included in the same city, so the granularity of predicting the logistics track in terms of the "logistics network point" dimension is greater than that in terms of the "city" dimension. The network point model can clean out the network point transportation relationship and related data indicators, including information such as timeliness, wave number, and route relationship, according to the network point path generated by the actual logistics route. In addition, a logistics track prediction model can also be obtained by combining logistics nodes of different dimensions. Such a logistics track prediction model is also called a combined model. For example, a statistical learning and training of the logistics track prediction model can be performed based on the path of "address + network point". Relevant data such as the number of recommendations and timeliness will also be precipitated during the training process.

[0119] Correspondingly, according to the factor value prediction model corresponding to the target timeliness influencing factor, determining the target factor value prediction model may include the following steps:

[0120] According to the respective call priorities of a plurality of pre-determined logistics track prediction models, sequentially determine the currently used logistics track prediction model from the plurality of logistics track prediction models; the call priority decreases as the logistics node level rises, or increases as the prediction accuracy of the logistics track prediction model rises; according to the currently used logistics track prediction model, determine the target factor value prediction model.

[0121] In specific implementation, a plurality of logistics track prediction models can be obtained. Different logistics track prediction models predict different logistics node levels, where the logistics node level is the node level of the logistics nodes in the logistics track prediction result.

[0122] For a plurality of logistics track prediction models, respective call priorities corresponding to the plurality of logistics track prediction models can be preset. For example, it can be determined according to the prediction accuracy of the logistics track prediction model. As the prediction accuracy of the logistics track prediction model increases, the call priority of the model also increases accordingly. In some embodiments, the call priority can be set according to the model evaluation information of the plurality of logistics track prediction models. For example, the call priority is determined according to the accuracy rate (i.e., the proportion of the number of predictions that meet the accuracy index in the total number of predictions) and the timeliness rate (i.e., the proportion of the number of predictions that meet the timeliness index in the total number of predictions) of the model.

[0123] Of course, the call priority can also be determined according to the logistics node levels involved in each of the multiple logistics trajectory prediction models. As the logistics node level involved in the logistics trajectory prediction model increases, the granularity of the logistics nodes decreases, and their call priorities also decrease accordingly. For example, the granularity of predicting the logistics trajectory in the "logistics network point" dimension is greater than that of predicting the logistics trajectory in the "city" dimension, and the logistics trajectory prediction model based on the "logistics network point" dimension can be preferentially used.

[0124] Furthermore, the consumption routing factor value prediction model can be carried out according to the call priority. Specifically, according to the call priorities of each of the multiple logistics trajectory prediction models, the currently used logistics trajectory prediction model can be determined from the multiple logistics trajectory prediction models in sequence, and then the currently used logistics trajectory prediction model is determined as the target factor value prediction model.

[0125] It can be understood that if there are multiple similar factor value prediction models with the same or similar functions among the multiple factor value prediction models, the call priorities of each of the multiple similar factor value prediction models can also be set, and then different model data can be consumed according to the priorities to obtain the optimal result for use.

[0126] In this embodiment, multiple logistics trajectory prediction models for predicting logistics estimates can be pre-trained, which helps to provide multiple optional logistics trajectory prediction methods in various complex environments, improve the flexibility of timeliness prediction. At the same time, by sequentially calling the logistics trajectory prediction models according to the call priority, the speed of obtaining reliable logistics trajectory prediction results can be accelerated, and the processing time can be saved.

[0127] In an exemplary embodiment, obtaining the future value of the target logistics object under the target timeliness influencing factors according to the target factor value prediction model may include the following:

[0128] Predict the logistics trajectory of the target logistics object according to the currently used logistics trajectory prediction model; in the case where the current logistics trajectory prediction result does not meet the prediction quality condition, return to execute the step of determining the currently used logistics trajectory prediction model from the multiple logistics trajectory prediction models in sequence according to the call priorities determined in advance for each of the multiple logistics trajectory prediction models, until the current logistics trajectory prediction result meets the prediction quality condition or is output by the logistics trajectory prediction model with the lowest priority; determine the future value of the target logistics object under the logistics trajectory factor according to the current logistics trajectory prediction result.

[0129] In a specific implementation, after determining the currently used logistics trajectory prediction model, the logistics trajectory of the target logistics object can be predicted according to the currently used logistics trajectory prediction model to obtain a logistics trajectory prediction result. Among them, the logistics trajectory prediction result can include a predicted logistics trajectory and an unpredictable logistics trajectory.

[0130] After obtaining the logistics trajectory prediction result, it can be determined whether the logistics trajectory prediction result meets the prediction quality condition. In some embodiments, if the logistics trajectory prediction result is an unpredictable logistics trajectory, it can be determined that the prediction quality condition is not met. In other embodiments, if the logistics trajectory prediction result is a predicted logistics trajectory, it can be determined whether the predicted logistics trajectory meets the expectation. For example, if the predicted logistics trajectory matches the actual logistics trajectory of the target logistics object, the prediction quality condition can be met; if the predicted logistics trajectory does not match the actual logistics trajectory of the target logistics object, it is determined that the prediction quality condition is not met.

[0131] When it is determined that the current logistics trajectory prediction result does not meet the prediction quality condition, the server will return to the step of determining the currently used logistics trajectory prediction model from multiple logistics trajectory prediction models in turn according to the respective call priorities of the pre-determined multiple logistics trajectory prediction models, and continue to use the logistics trajectory prediction model with the next priority to re-predict the logistics trajectory until the current logistics trajectory prediction result obtained by the logistics trajectory prediction model meets the prediction quality condition or is output by the logistics trajectory prediction model with the lowest priority. Then, according to the current logistics trajectory prediction result, the future value of the target logistics object under the target timeliness influencing factors can be determined. For example, the current logistics trajectory prediction result can be used as the future value under the logistics trajectory influencing factors.

[0132] Specifically, for example, first, the terminal delivery network point can be predicted through the logistics company and the receiving address; then, the network point model can perform network point routing prediction through the current network point (such as the pick-up network point) and the terminal delivery network point; if the prediction fails or the quality is poor, the line model can use the current network point and the receiving address to perform another line prediction; if it still does not meet the expectation, it will be downgraded to the line prediction of the logistics company, the shipping address, and the receiving address; if it still does not meet the expectation, the line timeliness will be predicted through other aggregated data. The line or network point relationship is predicted through the above various routing factor value prediction models to obtain the timeliness fitting result between two points (between the shipping address and the receiving address). After obtaining the timeliness fitting result, it can be checked whether the timeliness fitting result and the business time fall within the latest operation time of the terminal delivery network point. If so, the date where the latest operation time is located can be used as the estimated delivery date of the target logistics object; if not, the next earliest operation time of the terminal delivery network point is taken to obtain the estimated delivery date.

[0133] In this embodiment, on the one hand, by adopting multiple logistics trajectory prediction models and attempting to make predictions in sequence according to their respective priorities, the prediction results can be continuously adjusted and optimized when the prediction quality conditions are not met, which helps to avoid the biases or deficiencies that may exist in a single model, thereby improving the accuracy and reliability of the prediction. On the other hand, due to the complex and ever-changing logistics scenarios, factors such as different logistics objects and time limit requirements may affect the selection of logistics trajectories. By switching and selecting multiple logistics trajectory prediction models, it is possible to more flexibly adapt to different logistics scenarios and requirements, ensuring that the prediction results are more in line with the actual situation.

[0134] To enable those skilled in the art to better understand the above steps, taking logistics objects as an example, the embodiments of the present disclosure will be exemplarily described through an example below. However, it should be understood that the embodiments of the present disclosure are not limited thereto.

[0135] As Figure 4 shown, this embodiment may include steps S401 to S408.

[0136] S401. According to the historical logistics information corresponding to each historical logistics object, determine the logistics object transfer data representing the transfer situation of each historical logistics object between transfer nodes.

[0137] In specific implementation, as Figure 5 shown, various data of the historical logistics object can be obtained to obtain aggregated data, such as the transaction information of the platform, the logistics trajectory of the historical logistics object, the logistics electronic waybill, and other subscription information. It should be emphasized that all kinds of aggregated data are obtained and used under the full authorization of all parties. Then, the logistics object transfer data can be obtained according to the aggregated data.

[0138] S402. According to the analysis results of the logistics object transfer data, obtain the historical values corresponding to each of the multiple time limit influencing factors.

[0139] S403. According to the historical values corresponding to each time limit influencing factor, train the factor value prediction model corresponding to each time limit influencing factor.

[0140] In some examples, as Figure 5 shown, the factor value prediction models corresponding to each time limit influencing factor such as the prediction path, station time limit, and station wave can be obtained, and then one or more factor value prediction models can be used for time limit prediction. In addition, one or more factor value prediction models can be used to recommend paths and recommended time limits to relevant users. Among them, the recommended path and the recommended time limit in the background can be understood as comparing the logistics time limits of different paths through the factor value prediction model to obtain the optimal path and the optimal time limit.

[0141] S404, determine the detection indicators for predicting the quality of the time limit prediction of the factor value prediction model, and obtain the index information of each factor value prediction model under the detection indicators.

[0142] S405, determine the model evaluation information of each factor value prediction model according to the index information of each factor value prediction model.

[0143] In some examples, each factor value prediction model can be quality inspected based on the real data of the logistics object and the predicted data output by the factor value prediction model, and a unified quality model can be constructed, such as a quality model including detection indicators such as the number of samples, timeliness rate, accuracy rate, and coverage rate. Then, obtain the index information of each factor value prediction model under the detection indicators. Then, as Figure 5 shown, the model evaluation information including recommended configurations of each factor value prediction model can be obtained according to the index information, such as the regional restrictions, route matching conditions, time limit configuration information, etc. of each factor value prediction model.

[0144] S406, determine the target logistics object to be predicted for logistics time limit, and obtain the factor value prediction models corresponding to each of the multiple time limit influencing factors.

[0145] S407, determine the target factor value prediction model according to the model evaluation information of each of the multiple factor value prediction models, and obtain the future value of the target logistics object under the target time limit influencing factor corresponding to the target factor value prediction model.

[0146] S408, determine the logistics time limit prediction result of the target logistics object according to the future value under the target time limit influencing factor.

[0147] In some embodiments, the logistics time limit prediction result can be displayed on the trajectory details page. The logistics time limit prediction result can include the expected delivery time of the target logistics object. After obtaining the expected delivery time, it can be checked whether the expected delivery time of the current waybill time limit of the target logistics object has exceeded the actual time. If so, the calculation will be re-triggered and the expression mode of the logistics time limit prediction result will be adjusted. If the number of changes in the time limit prediction result of the current waybill exceeds the number threshold, the expression copy in the trajectory details page can be optimized. In addition, it can also be checked whether the current logistics information has been stagnant for a long time. If so, the expression copy in the trajectory details page can also be optimized according to the exception registration.

[0148] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0149] It can be understood that the same / similar parts among the various embodiments of the above methods in this specification can be referred to each other. Each embodiment focuses on the differences from other embodiments. For the relevant parts, refer to the descriptions of other method embodiments.

[0150] Based on the same inventive concept, the embodiments of the present disclosure also provide a timeliness prediction device for implementing the timeliness prediction method involved above.

[0151] Figure 6 It is a block diagram of a timeliness prediction device shown according to an exemplary embodiment. Referring to Figure 6 , the device includes:

[0152] A target service object determination unit 601, configured to determine a target service object for which service timeliness prediction is to be performed;

[0153] A model acquisition unit 602, configured to acquire a factor value prediction model corresponding to each of a plurality of timeliness impact factors of the target service object; the factor value prediction model is used to predict future values of the timeliness impact factors; the factor value prediction model is trained according to historical values of a plurality of historical service objects of the same type as the target service object under the timeliness impact factors, and the historical values are determined according to the historical service information of the historical service objects;

[0154] An impact factor information prediction unit 603, configured to acquire the future values of the target service object under target timeliness impact factors according to a plurality of the factor value prediction models; the target timeliness impact factors are at least one of the plurality of timeliness impact factors;

[0155] A timeliness prediction unit 604, configured to determine a timeliness prediction result of the target service object according to the future values under each of the target timeliness impact factors.

[0156] In an exemplary embodiment, the apparatus further includes a model training module, and the model training module is configured to perform:

[0157] Obtain the historical service information of each of the multiple historical service objects;

[0158] According to each piece of the historical service information, determine service object transfer data representing the transfer situation of each historical service object among transfer nodes; the transfer nodes include at least one of the following: the geographical location where the historical service object is located, and the status information of the historical service object;

[0159] According to the analysis results of each piece of the service object transfer data, obtain the historical values corresponding to each of the multiple timeliness impact factors;

[0160] According to the historical values corresponding to each timeliness impact factor, train the factor value prediction model corresponding to each timeliness impact factor.

[0161] In an exemplary embodiment, the model training module is specifically configured to perform:

[0162] Determine an update period set in advance for the factor value prediction model;

[0163] Within each update period, according to the historical service information newly added in the update period, determine the new service object transfer data of each historical service object;

[0164] Update the trained factor value prediction model according to the new service object transfer data.

[0165] In an exemplary embodiment, the impact factor information prediction unit 603 is configured to perform:

[0166] Determine a timeliness prediction scenario for the target service object to perform timeliness prediction, and obtain the timeliness prediction quality information of each of the multiple timeliness impact factor combination strategies associated with the timeliness prediction scenario; each timeliness impact factor combination strategy includes at least one of the timeliness impact factors, and the timeliness impact factors included in different timeliness impact factor combination strategies are at least partially different;

[0167] From the multiple timeliness impact factor combination strategies, determine a target timeliness impact factor combination strategy whose timeliness prediction quality information meets the prediction quality condition, and obtain the target timeliness impact factor;

[0168] According to the factor value prediction model corresponding to the target timeliness impact factor, determine a target factor value prediction model;

[0169] Based on the target factor value prediction model, obtain the future value of the target service object under the influence of the target time effect factor.

[0170] In an exemplary embodiment, the target service object includes a target logistics object for which logistics time effect prediction is to be performed; the factor value prediction model corresponding to the target time effect factor includes a plurality of logistics track prediction models corresponding to logistics track factors, and different logistics track prediction models predict different levels of logistics node levels, where the logistics node level is the node level of the logistics node in the logistics track prediction result;

[0171] The influence factor information prediction unit 603 is configured to perform:

[0172] According to the respective call priorities of the plurality of pre-determined logistics track prediction models, sequentially determine the currently used logistics track prediction model from the plurality of logistics track prediction models; the call priority decreases as the logistics node level rises, or increases as the prediction accuracy of the logistics track prediction model rises;

[0173] According to the currently used logistics track prediction model, determine the target factor value prediction model.

[0174] In an exemplary embodiment, the influence factor information prediction unit 603 is configured to perform:

[0175] According to the currently used logistics track prediction model, predict the logistics track of the target logistics object;

[0176] In the case where the current logistics track prediction result does not meet the prediction quality condition, return to execute the step of sequentially determining the currently used logistics track prediction model from the plurality of pre-determined logistics track prediction models according to their respective call priorities until the current logistics track prediction result meets the prediction quality condition or is output by the logistics track prediction model with the lowest priority;

[0177] According to the current logistics track prediction result, determine the future value of the target logistics object under the logistics track factor.

[0178] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment related to the method, and will not be elaborated here.

[0179] Each module in the above aging prediction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so as to facilitate the processor to call and execute the operations corresponding to each of the above modules.

[0180] Figure 7 FIG. 700 is a block diagram of an electronic device for implementing an aging prediction method according to an exemplary embodiment. For example, the electronic device 700 can be a server. Referring to Figure 7 FIG. 700, the electronic device 700 includes a processing component 720, which further includes one or more processors, and memory resources represented by a memory 722 for storing instructions executable by the processing component 720, such as application programs. The application programs stored in the memory 722 can include one or more modules each corresponding to a set of instructions. In addition, the processing component 720 is configured to execute instructions to perform the above method.

[0181] The electronic device 700 may further include: a power component 724 configured to perform power management of the electronic device 700, a wired or wireless network interface 726 configured to connect the electronic device 700 to a network, and an input / output (I / O) interface 728. The electronic device 700 can operate based on an operating system stored in the memory 722, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, or the like.

[0182] In an exemplary embodiment, there is also provided a computer-readable storage medium including instructions, such as the memory 722 including instructions, and the above instructions can be executed by the processor of the electronic device 700 to complete the above method. The storage medium can be a computer-readable storage medium. For example, the computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0183] In an exemplary embodiment, there is also provided a computer program product, which includes instructions, and the above instructions can be executed by the processor of the electronic device 700 to complete the above method.

[0184] It should be noted that the above device, electronic device, computer-readable storage medium, computer program product, etc. may further include other implementation manners according to the description of the method embodiments. The specific implementation manners can refer to the description of the relevant method embodiments and will not be elaborated here one by one.

[0185] Other embodiments of the present disclosure will be readily apparent to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. The present disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include known common general knowledge or conventional technical means in the technical field not disclosed by the present disclosure. The specification and examples are only to be considered as exemplary, and the true scope and spirit of the present disclosure are pointed out by the claims.

[0186] It should be understood that the present disclosure is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is only limited by the appended claims.

Claims

1. An aging prediction method, characterized in that, Including: Determine a target business object for which business timeliness prediction is to be performed; Obtain a factor value prediction model corresponding to each of multiple timeliness influencing factors of the target business object; the factor value prediction model is used to predict the future value of the timeliness influencing factor; the factor value prediction model is trained based on the historical values of multiple historical business objects of the same type as the target business object under the timeliness influencing factor, and the historical values are determined according to the historical business information of the historical business objects; According to multiple factor value prediction models, obtain the future value of the target business object under a target timeliness influencing factor; the target timeliness influencing factor is at least one of the multiple timeliness influencing factors; Determine the timeliness prediction result of the target business object according to the future values under each target timeliness influencing factor.

2. The method according to claim 1, characterized in that Before obtaining the factor value prediction model corresponding to each of multiple timeliness influencing factors of the target business object, it further includes: Obtain the historical business information of multiple historical business objects; According to each piece of historical business information, determine business object transfer data representing the transfer situation of each historical business object between transfer nodes; the transfer nodes include at least one of the following: the geographical location where the historical business object is located, and the status information of the historical business object; According to the analysis results of each piece of business object transfer data, obtain the historical value corresponding to each of the multiple timeliness influencing factors; Train the factor value prediction model corresponding to each timeliness influencing factor according to the historical value corresponding to each timeliness influencing factor.

3. The method according to claim 2, wherein After training the factor value prediction model corresponding to each timeliness influencing factor according to the historical value corresponding to each timeliness influencing factor, it further includes: Determine an update period set in advance for the factor value prediction model; Within each update period, according to the historical business information newly added in the update period, determine the new business object transfer data of each historical business object; Update the trained factor value prediction model according to the new business object transfer data.

4. The method according to claim 1, wherein The step of obtaining the future value of the target business object under a target timeliness influencing factor according to multiple factor value prediction models includes: Determine a timeliness prediction scenario for which the target business object performs timeliness prediction, and obtain the timeliness prediction quality information of each of multiple timeliness influencing factor combination strategies associated with the timeliness prediction scenario; each timeliness influencing factor combination strategy includes at least one of the timeliness influencing factors, and at least some of the timeliness influencing factors included in different timeliness influencing factor combination strategies are different; From the multiple timeliness influencing factor combination strategies, determine a target timeliness influencing factor combination strategy whose timeliness prediction quality information meets the prediction quality condition, and obtain the target timeliness influencing factor; Determine a target factor value prediction model according to the factor value prediction model corresponding to the target timeliness influencing factor; Obtain the future value of the target business object under the target timeliness influencing factor according to the target factor value prediction model.

5. The method according to claim 4, characterized in that The target business object includes a target logistics object for which logistics timeliness prediction is to be performed; the factor value prediction model corresponding to the target timeliness influencing factor includes a plurality of logistics trajectory prediction models corresponding to logistics trajectory factors, and different logistics trajectory prediction models predict different logistics node levels, where the logistics node level is the node level of the logistics node in the logistics trajectory prediction result; Determining the target factor value prediction model according to the factor value prediction model corresponding to the target timeliness influencing factor includes: Determining the currently used logistics trajectory prediction model from the plurality of logistics trajectory prediction models in sequence according to the respective call priorities of the plurality of pre-determined logistics trajectory prediction models; the call priority decreases as the logistics node level increases, or increases as the prediction accuracy of the logistics trajectory prediction model increases; Determining the target factor value prediction model according to the currently used logistics trajectory prediction model.

6. The method according to claim 5, characterized in that, Obtaining the future value of the target logistics object under the target timeliness influencing factor according to the target factor value prediction model includes: Predicting the logistics trajectory of the target logistics object according to the currently used logistics trajectory prediction model; In the case where the current logistics trajectory prediction result does not meet the prediction quality condition, returning to execute the step of determining the currently used logistics trajectory prediction model from the plurality of logistics trajectory prediction models in sequence according to the respective call priorities of the plurality of pre-determined logistics trajectory prediction models, until the current logistics trajectory prediction result meets the prediction quality condition or is output by the logistics trajectory prediction model with the lowest priority; Determining the future value of the target logistics object under the logistics trajectory factor according to the current logistics trajectory prediction result.

7. An aging prediction device, characterized in that, Including: A target business object determination unit configured to execute determining a target business object for which business timeliness prediction is to be performed; A model acquisition unit configured to execute acquiring factor value prediction models corresponding to respective multiple timeliness influencing factors of the target business object; the factor value prediction model is used to predict the future value of the timeliness influencing factor; the factor value prediction model is trained according to the historical values of multiple historical business objects of the same type as the target business object under the timeliness influencing factor, and the historical value is determined according to the historical business information of the historical business object; An influencing factor information prediction unit configured to execute obtaining the future value of the target business object under a target timeliness influencing factor according to the multiple factor value prediction models; the target timeliness influencing factor is at least one of the multiple timeliness influencing factors; A timeliness prediction unit configured to execute determining a timeliness prediction result of the target business object according to the future values under respective target timeliness influencing factors.

8. An electronic device, characterized in that, Including: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to execute the instructions to implement the timeliness prediction method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the aging prediction method according to any one of claims 1 to 6.

10. A computer program product, comprising instructions therein, characterized in that, When the instructions are executed by a processor of an electronic device, the electronic device is enabled to execute the aging prediction method according to any one of claims 1 to 6.

Images