CN116258224A - Express delivery time prediction method, device, computer equipment and storage medium - Google Patents

Abstract

The application provides a method, device, computer equipment and storage medium for predicting the timeliness of a shipment. The method includes: obtaining the shipment attribute information of the target shipment; performing feature engineering processing on the shipment attribute information to obtain the shipment characteristics of the target shipment; inputting the shipment characteristics into The trained timeliness prediction model outputs the predicted probability value of the target shipment under each preset timeliness type; the trained timeliness prediction model is composed of a fully connected layer, a dropout layer, a fusion layer, and a normalized layer; analysis Each prediction probability value is used to determine the logistics timeliness of the target express. Adopting this method can improve the prediction accuracy rate of express delivery timeliness.

Description

Express delivery time prediction method, device, computer equipment and storage medium

technical field

The embodiments of the present application relate to the technical field of artificial intelligence, and in particular, to a method, device, computer equipment, and storage medium for express delivery timeliness prediction.

Background technique

In the field of logistics, the timeliness of express delivery commitments is usually roughly estimated based on business experience, collectively referred to as express delivery timeliness prediction methods, including but not limited to the following methods: (1) 85% of the express delivery of a certain product in a representative network in the city The timeliness that can be achieved is the promised timeliness of the flow of the product at all outlets in this city; (2) Divide the express delivery life cycle into six points and five stages: receiving end-outlet-transit station-transit station-outlet-delivering end , each section calculates 85% of the achievable time limit according to the daily shift planning of the outlet/transit station, and then accumulates it based on the historical 14-day data.

However, the disadvantage of the first method is that a fixed commitment time limit is given to the outlets in the same city without distinction, which will produce a certain deviation, and the timeliness of the representative outlets selected according to the number of pieces is not necessarily representative; the second method The disadvantage is that the historical 14-day data can only cover 70% of the flow direction, resulting in missing data during forecasting, and the accumulation of errors caused by independent calculations for each segment. The 5 tables need to be refreshed twice a week, which not only consumes a lot of machine resources, but also has some problems. The risk of re-running subsequent links if the link scheduling fails.

Therefore, the traditional method of forecasting the timeliness of express shipments has the problem of low prediction accuracy.

Contents of the invention

The purpose of the present application is to provide a method, device, computer equipment and storage medium for predicting the timeliness of express shipments, so as to improve the prediction accuracy of the timeliness of express shipments.

In the first aspect, the present application provides a method for predicting the timeliness of express shipments, including:

Obtain the shipment attribute information of the target shipment;

Perform feature engineering processing on the attribute information of the shipment to obtain the shipment characteristics of the target shipment;

Input the features of the express shipment into the trained timeliness prediction model, and output the predicted probability value of the target shipment under each preset timeliness type; where the trained timeliness prediction model is composed of a fully connected layer, a dropout layer, a fusion layer and a normalized layered

Analyze each predicted probability value to determine the logistics timeliness of the target express.

In some embodiments of the present application, the express shipment features are input into the trained aging prediction model, and the predicted probability values of the target shipments under each preset aging type are output, including: combining the express shipment features to obtain the combined express shipment features ; Input the express features into the trained aging prediction model, and output the first correlation information inside the features; and input the combined express features into the trained aging prediction model, and output the second correlation information between the features; analyze the first The first association information and the second association information are used to obtain the predicted probability values of the target express shipment under each preset aging type.

In some embodiments of the present application, the express feature is input into the trained time-efficiency prediction model, and the first associated information inside the feature is output, including: inputting the express feature into the trained time-efficiency prediction model, through the fully connected layer and the dropout layer Classify the express features to obtain the initial feature vector and the differentiation coefficient of the initial feature vector; fuse the initial feature vector and the differentiation coefficient through the fusion layer to obtain the target feature vector with the differentiation coefficient; through the normalization layer The target feature vector is normalized to obtain the first associated information inside the feature.

In some embodiments of the present application, analyzing the first association information and the second association information to obtain the predicted probability value of the target express shipment under each preset aging type includes: obtaining the first characteristic matrix corresponding to the first association information, and the second The second feature matrix corresponding to the two related information; matrix multiplication of the first feature matrix and the second feature matrix to obtain the comprehensive feature; through the target fully connected layer carrying the sigmoid function, the feature classification of the comprehensive feature is obtained to obtain the target express Predicted probability values under each preset aging type.

In some embodiments of the present application, before inputting the features of the shipment into the trained aging prediction model, it also includes: constructing an initial aging prediction model; performing preliminary training on the initial aging prediction model based on a preset initial learning rate, Until the number of training times reaches the preset first threshold, the time-effect prediction model after the preliminary training is obtained; based on the preset peak learning rate, the time-effect prediction model after the preliminary training is trained horizontally until the number of training times reaches the preset second threshold, and the time-effect prediction model is obtained The time-effect prediction model after level training; according to the trigonometric function curve of the peak learning rate, the time-effect prediction model after level training is debugged and trained to obtain the trained time-effect prediction model.

In some embodiments of the present application, feature engineering processing is performed on the attribute information of the express to obtain the express features of the target express, including: extracting address information, time information and product information included in the attribute information of the express; performing feature engineering processing on the address information to obtain The outlet information and distance information are used as the express features; and the time information is processed by feature engineering, and the festival information and delivery time are obtained as the express features; and the product information is processed by feature engineering, and the product type information is obtained as the express features.

In some embodiments of the present application, analyzing each predicted probability value to determine the logistics timeliness of the target express delivery includes: screening out the maximum value of each predicted probability value to obtain the target predicted probability value; determining the timeliness type corresponding to the target predicted probability value , as the logistics timeliness of the target shipment.

In the second aspect, the present application provides a device for predicting the aging of express shipments, including:

An information acquisition module, configured to acquire the attribute information of the target express;

The feature engineering module is used to perform feature engineering processing on the attribute information of the express to obtain the express features of the target express;

The model analysis module is used to input the characteristics of the express shipment into the trained aging prediction model, and output the predicted probability value of the target shipment under each preset aging type; wherein, the trained aging prediction model is composed of a fully connected layer, a dropout layer, Composed of fusion layer and normalization layer;

The timeliness determining module is used for analyzing each predicted probability value to determine the logistics timeliness of the target express.

In a third aspect, the present application also provides a computer device, including:

one or more processors;

memory; and one or more application programs, wherein one or more application programs are stored in the memory and configured to be executed by the processor to implement the method for predicting the delivery time in the first aspect above.

In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored, and the computer program is loaded by a processor to execute the steps in the method for express delivery timeliness prediction.

In a fifth aspect, embodiments of the present application provide a computer program product or computer program, where the computer program product or computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the method provided by the first aspect above.

In the aforementioned method, device, computer equipment, and storage medium for predicting the timeliness of a shipment, the server obtains the shipment characteristics of the target shipment by obtaining the shipment attribute information of the target shipment, and performs feature engineering processing on the shipment attribute information, and then inputs the shipment characteristics into the trained The timeliness prediction model can output the predicted probability value of the target express under each preset timeliness type, and analyze each predicted probability value to determine the logistics timeliness of the target express. Therefore, the timeliness prediction model proposed in this application is used to predict the timeliness of the target express shipment, without paying attention to the time-consuming impact of the express delivery link, ensuring the stability of the prediction results, and finally improving the prediction accuracy of the expressage timeliness.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings that need to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, other drawings can also be obtained based on these drawings without any creative effort.

Fig. 1 is the schematic diagram of the scenario of the method for predicting the express delivery timeliness provided in the embodiment of the present application;

Fig. 2 is a schematic flow chart of the method for predicting the express delivery timeliness provided in the embodiment of the present application;

Fig. 3 is the framework application flow diagram of the aging prediction model provided in the embodiment of the present application;

Fig. 4 is a schematic structural diagram of an express delivery time prediction device provided in an embodiment of the present application;

Fig. 5 is a schematic structural diagram of a computer device provided in an embodiment of the present application.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the application with reference to the drawings in the embodiments of the application. Apparently, the described embodiments are only some of the embodiments of the application, not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts belong to the scope of protection of this application.

In the description of the present application, the terms "first" and "second" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of said features. In the description of the present application, "plurality" means two or more, unless otherwise specifically defined.

In the description of this application, the term "for example" is used to mean "serving as an example, illustration or illustration". Any embodiment described in this application as "such as" is not necessarily to be construed as preferred or advantageous over other embodiments. The following description is given to enable any person skilled in the art to make and use the invention. In the following description, details are set forth for purposes of explanation. It should be understood that one of ordinary skill in the art would recognize that the present invention may be practiced without the use of these specific details. In other instances, well-known structures and procedures are not described in detail to avoid obscuring the description of the present invention with unnecessary detail. Thus, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed in this application.

In the embodiment of the present application, the express delivery aging prediction method provided in the embodiment of the present application can be applied to the express delivery aging prediction system as shown in FIG. 1 . Wherein, the delivery time prediction system includes a terminal 102 and a server 104 . The terminal 102 may be a device including both receiving and transmitting hardware, that is, a device having receiving and transmitting hardware capable of performing bidirectional communication over a bidirectional communication link. Such devices may include cellular or other communication devices having a single line display or a multi-line display or a cellular or other communication device without a multi-line display. Specifically, the terminal 102 may be a desktop terminal or a mobile terminal, and specifically, the terminal 102 may also be one of a mobile phone, a tablet computer, and a notebook computer. The server 104 can be an independent server, or a server network or server cluster composed of servers, including but not limited to computers, network hosts, single network servers, multiple network server sets, or cloud servers composed of multiple servers. Wherein, the cloud server is composed of a large number of computers or network servers based on cloud computing (Cloud Computing). In addition, a communication connection is established between the terminal 102 and the server 104 through a network, and the network may specifically be any one of a wide area network, a local area network, and a metropolitan area network.

Those skilled in the art can understand that the application environment shown in Figure 1 is only an application scenario applicable to the solution of this application, and does not constitute a limitation on the application scenario of the solution of this application. more or less equipment than shown. For example, only one server is shown in FIG. 1 . It can be understood that the delivery time prediction system may also include one or more other servers, which are not specifically limited here. In addition, as shown in FIG. 1 , the express mail aging prediction system may further include a memory for storing data, such as storing express mail attribute information of a target express mail.

It should be noted that the schematic diagram of the scene of the delivery timeliness prediction system shown in FIG. Constituting a limitation on the technical solutions provided by the embodiments of the present invention, those of ordinary skill in the art know that with the evolution of the express delivery timeliness prediction system and the emergence of new business scenarios, the technical solutions provided by the embodiments of the present invention are equally applicable to similar technical problems Be applicable.

Referring to Fig. 2, the embodiment of the present application provides a method for predicting the timeliness of express mail. This embodiment mainly uses the method to be applied to the server 104 in the above Fig. 1 as an example. The method includes steps S201 to S204, as follows:

S201. Obtain shipment attribute information of a target shipment.

Wherein, the target express shipment may be the goods to be consigned selected for timeliness prediction, and the goods to be consigned are specifically "express shipments" with faster transportation speed and higher freight costs. It should be noted that the embodiment of this application proposes the use of target express shipments for logistics timeliness prediction. The original intention is not to distinguish the types of goods in the logistics field, that is, to not distinguish between "express shipments" and "slow shipments", and to use "express shipments" to refer to logistics The goods to be consigned in the field, but if the types of goods such as "express" and "slow" are distinguished in other embodiments, the corresponding type of goods information can be used for logistics timeliness prediction.

Among them, the express attribute information may include address information, time information, and product information, and the address information may include sending address information, receiving address information, etc.; time information may include sending time, time-consuming history of the flow, etc.; product information may include product type etc.

In the specific implementation, in order to improve the prediction accuracy of the delivery timeliness, the server 104 may first obtain the delivery attribute information of the target delivery to be predicted, and the methods for obtaining the delivery attribute information include but are not limited to the following methods: 1. In the network structure, the server 104 receives the express mail attribute information from the terminal 102 or other devices with network connections; 2. In the pre-established blockchain network, the server 104 can obtain the express mail attribute information synchronously from other terminal nodes or server nodes information, the blockchain network can be a public chain, a private chain, etc.; 3. In the preset tree structure, the server 104 can request the express mail attribute information from the upper-level server, or poll the lower-level server to obtain the express mail attribute information .

In addition, the server 104 can also obtain the address information, time information and product information of each shipment, and then integrate and obtain the shipment attribute information of the target shipment. At this time, the address information, time information and product information of each express shipment can be obtained by the same device, or can be obtained by multiple devices respectively.

S202. Perform feature engineering processing on the attribute information of the express to obtain the express features of the target express.

Among them, feature engineering refers to the process of transforming the original data into the training data of the model. Its purpose is to obtain better training data characteristics, so that the machine learning model can approach this upper limit. Feature engineering can improve the performance of the model, and sometimes even achieve good results on simple models. Feature engineering plays a very important role in machine learning, and generally includes three parts: "feature construction, feature extraction, and feature selection".

In a specific implementation, the server 104 performs feature engineering processing operations on the attribute information of the express, which can actually be regarded as feature design. The feature design in the embodiment of the present application will highlight time information and static information, and remove the date difference between the training set and the test set. The impact on the model improves the generalization ability of the model.

In one embodiment, this step includes: extracting the address information, time information and product information included in the attribute information of the express; performing feature engineering processing on the address information to obtain network point information and distance information as express features; and performing feature engineering on the time information Process to get holiday information and delivery time as express features; and perform feature engineering processing on product information to get product type information as express features.

In a specific implementation, the server 104 can perform feature engineering processing on the address information to obtain information about outlets, outlet levels, upper-level outlets, outlet types, districts, regions, city representative outlets, administrative district representative outlets, etc. of the receiving and sending ends, and Distance information such as city line distance, city distance, longitude and latitude, and distance type can all be used as express features. The features obtained from the analysis of the embodiment of the present application can include 11 numerical features and 27 category features, a total of 38 express features.

For example, the respective outlets of the receiving and sending ends can be a certain school, building, community office, etc.; the outlets can be divided into first, second, and third grades according to business rules. , the number of staff, etc.; the types of outlets can include hospitals, schools, etc.; districts can include urban districts, such as Nanshan District, Shenzhen, Futian District, etc.; large districts can include Central China, South China, West China, North China, etc.; It is the straight-line distance on the map between the sending and receiving ends, and the distance types include the same city, within the province, outside the province, and the economic circle.

Further, the server 104 can perform feature engineering processing on the time information to obtain information such as whether it is a holiday, the city journey time, and the 75th, 80th, and 85th quantile of the historical time of the flow, and then "yyyy-MM-dd hh:mm The shipping time of ":ss" is converted into the float value of "hh*3600+mm*60+ss", which can be used as the express feature. Among them, "yyyy-MM-dd hh:mm:ss" means converting the time to a 12-hour clock, for example, "2018-06-27 03:24:21".

Furthermore, the server 104 can perform feature engineering processing on product information to obtain product type information, such as standard express, express and other product types, as express features. It should be noted that, in this embodiment, all express features are "static features" describing the place of origin and destination, and the information of the place of origin and destination are generally complete, and other information about distance and Historical time-consuming features can be directly obtained by correlating existing known information, so the data coverage can reach 100%, and there is no problem of missing data, that is, the accuracy of express delivery time prediction will not be reduced.

S203. Input the features of the express shipment into the trained time-effect prediction model, and output the predicted probability values of the target shipment under each preset time-effect type; wherein, the trained time-effect prediction model is composed of a fully connected layer, a dropout layer, a fusion layer and a regression layer. It is composed of one chemical layer.

Wherein, the preset aging type may be an aging type composed of a time range, for example, the aging type includes "2d12", "2d18", "2d22", etc., and "2d12" means the time range of "12 o'clock from today to tomorrow", "2d18" means the time range of "18:00 today to tomorrow", "2d22" means the time range of "22:00 today to tomorrow", and extrapolate accordingly. It can be understood that the specific time range can be set according to actual business requirements, and is not specifically limited in this application.

In a specific implementation, after the server 104 analyzes and acquires the characteristics of the express shipment, it can call the trained timeliness prediction model, and then input the expressage features into the trained timeliness prediction model for feature analysis, so that the trained timeliness prediction model outputs the target shipment Predicted probability values under each preset aging type. In addition, before the server 104 invokes the trained timeliness prediction model, model training can be performed on the timeliness prediction model, and this application proposes to automatically train the timeliness prediction model using the learning rate of "straight up -> maintain -> curve down", To maximize model performance. In this way, the model predicts the target only based on the information at the two ends of the delivery, and blurs the various intermediate transit links, and infers the "overall time consumption" of the express life cycle in an end-to-end manner, that is, there is no need to calculate and superimpose each period of time consumption in stages, It avoids the defect of multi-link scheduling interdependence, and improves the prediction accuracy of express delivery timeliness. The model training steps and model analysis steps involved in this embodiment will be described in detail below.

In one embodiment, this step includes: performing feature merging on the express features to obtain the combined express features; inputting the express features to the trained aging prediction model, and outputting the first associated information inside the features; and combining the combined The features of the shipment are input to the trained aging prediction model, and the second correlation information between the features is output; the first correlation information and the second correlation information are analyzed to obtain the predicted probability value of the target shipment under each preset aging type.

In the specific implementation, please refer to Fig. 3, which is a flowchart of the application of the framework of the time-sensitive prediction model provided by the embodiment of the present application. The normalization layer is composed, which can not only realize the abstraction of express features, but also automatically learn the difference coefficients of different dimensions of express features.

Specifically, after the server 104 inputs 38 shipment features into the trained time-efficiency prediction model, each shipment feature will sequentially pass through the fully connected layer, dropout layer, fusion layer, and normalization layer to obtain the internal difference information and correlation of the features. Information, as the first associated information; at the same time, the aging prediction model can not only be applied to the 38 independent express features, but also can be applied to the total express features "after the merger". That is to say, the server 104 can combine the features of the express shipments to obtain the combined express features as the total express features, and then input the total express features into the trained time-efficiency prediction model for analysis, and also go through the fully connected layer, dropout Layer, fusion layer and normalization layer, the difference information and related information between features can be obtained as the second related information. Finally, by analyzing the first associated information and the second associated information, the predicted probability values of the target express shipment under each preset aging type can be obtained.

In one embodiment, the express feature is input into the trained time-efficiency prediction model, and the first associated information inside the feature is output, including: inputting the express feature into the trained time-efficiency prediction model, and the express is processed through the fully connected layer and the dropout layer. Features are classified to obtain the initial feature vector and the differentiation coefficient of the initial feature vector; the initial feature vector and the differentiation coefficient are fused through the fusion layer to obtain the target feature vector with the differentiation coefficient; the target feature vector is obtained through the normalization layer The feature vector is normalized to obtain the first associated information inside the feature.

In a specific implementation, the server 104 can input the express features into the trained time-effect prediction model, so that each express feature can pass through the dense fully connected layer with the ReLu activation function, the dense fully connected layer without the activation function, the dropout layer, and the 16-dimensional The dense fully connected layer obtains a 16-dimensional feature vector as the initial feature vector. At this time, if the initial feature vector passes through the fully connected layer with sigmoid function, the differentiation coefficient of each initial feature vector will be calculated.

Further, after using the initial feature vector calculated by "dense-16" and the differentiation coefficient calculated by "Dense(Sigmoid)-16", after dot multiplication, the target feature with the differentiation coefficient is obtained Vector, use the layer normalization technique to calculate the mean and variance of the elements of the same dimension of the 38 target feature vectors, and perform normalization processing, you can get the distribution with the mean value of "0" and the variance of "1", and learn the feature The internal "difference" and "related" information is used as the first associated information.

In one embodiment, analyzing the first association information and the second association information to obtain the predicted probability value of the target shipment under each preset aging type includes: obtaining the first feature matrix corresponding to the first association information, and the second association information The second feature matrix corresponding to the information; matrix multiplication of the first feature matrix and the second feature matrix to obtain the comprehensive feature; through the target fully connected layer carrying the sigmoid function, the feature classification of the comprehensive feature is carried out to obtain the target express in each The predicted probability value under the default aging type.

In the specific implementation, the first associated information is expressed as the internal abstract information of 38 express features, and the actual output through the aging prediction model is the "16*38" first feature matrix; at the same time, the second associated information is expressed as 38 express features The external abstract information of , the actual output through the time-effect prediction model is the first feature matrix of "16*1". After that, the server 104 can transpose the first feature matrix so that the dimension of the matrix is "38*16", and then perform matrix multiplication with the second feature matrix through "matrix multiplication", so that the feature of each express can be realized. cross fusion. Finally, the comprehensive features obtained by multiplying the matrices pass through a fully connected layer with a sigmoid function, and the timeliness prediction model will output the predicted probability values of the target shipment under each preset timeliness type.

It can be understood that the activation function set in the embodiment of the present application is specifically used to add nonlinear factors, improve the expressive ability of the neural network to the model, and solve problems that cannot be solved by the linear model. Using the ReLu function can improve the calculation speed and Convergence speed, using the Sigmoid function can transform the continuous real value of the input into an output between 0 and 1.

In one embodiment, before this step, it also includes: constructing an initial time-effect prediction model; based on a preset initial learning rate, performing preliminary training on the initial time-effect prediction model until the number of training times reaches a preset first threshold, obtaining The time-effect prediction model after the preliminary training; based on the preset peak learning rate, the time-effect prediction model after the preliminary training is trained horizontally until the number of training times reaches the preset second threshold, and the time-effect prediction model after the level training is obtained; learn according to the peak value The trigonometric function curve of the rate is used to debug and train the time-effect prediction model after level training to obtain the trained time-effect prediction model.

In the specific implementation, the learning rate is usually kept constant in different iterations of model training. For example, the learning rate is often set to 0.1, 0.01, 0.005, etc. Among them, a larger learning rate will make the model easily fall into the local optimal solution, and the loss will fluctuate greatly; while a smaller learning rate will make the model converge too slowly, and eventually it will not be able to converge completely, and the global optimal solution will still not be found. Therefore, this application proposes a learning rate that changes automatically with the number of iterations, that is, the learning rate is automatically adjusted according to the rule of "straight up -> maintain -> curve down", which can avoid large fluctuations and speed up when the loss is stable Learning rate, reduce the learning rate when the subsequent model is close to the global optimal solution, and steadily approach the optimal solution, so as to avoid excessive learning rate and cross the optimal solution, resulting in poor loss.

Specifically, the initial learning rate can be set to "0.001", after two rounds of linear increase to the peak learning rate "0.004", after maintaining two rounds of level training, the learning rate will be gradually reduced according to the law of the trigonometric function curve, and the speed of decline can be slow first and then Fast, you can get the trained time-effect prediction model. In this way, dynamic adaptive learning rates can be configured for different stages of model optimization, realizing the steady convergence of the model to the global optimal solution, and avoiding the situation in which the model does not converge due to the fixed learning rate during the training process in traditional solutions.

S204, analyzing each predicted probability value to determine the logistics timeliness of the target express.

Among them, the logistics timeliness can include the delivery time of the logistics promise, which can also be called the "estimated delivery time". For example, the logistics timeliness of the target shipment "A" is "today to 12 o'clock tomorrow", and the logistics timeliness of the target shipment "B" is "Today to tomorrow at 18 o'clock".

In a specific implementation, after the server 104 analyzes and obtains the predicted probability values of the target express shipment under each preset timeliness type based on the above steps, it can further analyze each predicted probability value to determine the logistics timeliness of the target express piece. The analysis method can be to compare The magnitude of each predicted probability value.

In one embodiment, this step includes: screening out the maximum value of each predicted probability value to obtain the target predicted probability value; determining the timeliness type corresponding to the target predicted probability value as the logistics timeliness of the target express.

For specific implementation, please refer to Figure 3. If there is a 13-dimensional fully connected layer with a sigmoid function in the time-effect prediction model, there are 13 output prediction probability values, and the maximum value is selected from these 13 prediction probability values. As the logistics timeliness of the target shipment. However, it is understandable that if there is an N-dimensional fully connected layer with a sigmoid function in the time-effect prediction model, N prediction probability values will be output, and N≥2.

In the method for predicting the timeliness of a shipment in the above embodiment, the server obtains the shipment attribute information of the target shipment and performs feature engineering processing on the shipment attribute information to obtain the shipment characteristics of the target shipment, and then inputs the shipment characteristics into the trained timeliness prediction model , the predicted probability value of the target express shipment under each preset aging type can be outputted, and each predicted probability value can be analyzed to determine the logistics aging effect of the target express shipment. Therefore, using the timeliness prediction model proposed in this application to predict the timeliness of the target shipment can greatly increase the data coverage, avoid the interdependence of multiple links, and finally improve the prediction accuracy of the timeliness of the express piece under the state of the best model performance .

It should be understood that although the various steps in the flow chart of FIG. 2 are displayed sequentially according to the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in FIG. 2 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but may be executed at different times. The execution of these sub-steps or stages The order is not necessarily performed sequentially, but may be performed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

In order to better implement the express delivery aging prediction method provided in the embodiment of the present application, on the basis of the express delivery aging prediction method proposed in the embodiment of the present application, an express delivery aging prediction device is also provided in the embodiment of the application, as shown in Figure 4 As shown, the delivery aging prediction device 400 includes:

An information acquisition module 410, configured to acquire the attribute information of the target express;

The feature engineering module 420 is used to perform feature engineering processing on the attribute information of the express to obtain the express features of the target express;

The model analysis module 430 is used to input the characteristics of the express shipment into the trained aging prediction model, and output the predicted probability value of the target shipment under each preset aging type; wherein, the trained aging prediction model is composed of a fully connected layer and a dropout layer. , fusion layer and normalization layer;

The timeliness determination module 440 is configured to analyze each predicted probability value to determine the logistics timeliness of the target express.

In one embodiment, the model analysis module 430 is also used to combine the features of the express shipment to obtain the combined express features; input the express shipment features to the trained aging prediction model, and output the first associated information inside the feature; and The combined express features are input to the trained aging prediction model, and the second correlation information between the features is output; the first correlation information and the second correlation information are analyzed to obtain the predicted probability value of the target shipment under each preset aging type.

In one embodiment, the model analysis module 430 is also used to input the features of the shipment into the trained time-effect prediction model, and classify the features of the shipment through the fully connected layer and the dropout layer to obtain the difference between the initial feature vector and the initial feature vector Coefficient; through the fusion layer, the initial feature vector and the differentiation coefficient are fused to obtain the target feature vector carrying the differentiation coefficient; through the normalization layer, the target feature vector is normalized to obtain the first associated information inside the feature .

In one embodiment, the model analysis module 430 is further configured to obtain a first feature matrix corresponding to the first correlation information, and a second feature matrix corresponding to the second correlation information; perform matrix comparison of the first feature matrix and the second feature matrix Multiplying to get the comprehensive feature; through the target fully connected layer with the sigmoid function, the comprehensive feature is classified, and the predicted probability value of the target express shipment under each preset aging type is obtained.

In one embodiment, the express delivery aging prediction device 400 also includes a model training module for constructing an initial aging prediction model; based on a preset initial learning rate, the initial aging prediction model is initially trained until the number of training times reaches the preset The first threshold is to obtain the time-effect prediction model after preliminary training; based on the preset peak learning rate, the time-effect prediction model after preliminary training is trained horizontally until the number of training times reaches the preset second threshold, and the time-effect prediction after level training is obtained Model; according to the trigonometric function curve of the peak learning rate, the time-effect prediction model after level training is debugged and trained to obtain the trained time-effect prediction model.

In one embodiment, the feature engineering module 420 is also used to extract address information, time information and product information included in the attribute information of the express shipment; carry out feature engineering processing on the address information, and obtain network point information and distance information as express mail features; and analyze the time information Perform feature engineering processing to obtain holiday information and delivery time as express features; and perform feature engineering processing on product information to obtain product type information as express features.

In one embodiment, the timeliness determination module 440 is also used to filter out the maximum value among the predicted probability values to obtain the target predicted probability value; determine the timeliness type corresponding to the target predicted probability value as the logistics timeliness of the target express.

In the above-mentioned embodiment, using the timeliness prediction model proposed by this application to predict the timeliness of the target shipment can greatly increase the data coverage, avoid the interdependence of multiple links, and finally improve the timeliness prediction of the express piece under the state of the best model performance Accuracy.

It should be noted that, for the specific limitations of the express mail aging prediction device, please refer to the above-mentioned limitation on the express mail aging prediction method, which will not be repeated here. Each module in the above-mentioned delivery aging prediction device can be fully or partially realized by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the electronic device in the form of hardware, and can also be stored in the memory of the electronic device in the form of software, so that the processor can invoke and execute the corresponding operations of the above-mentioned modules.

In some embodiments of the present application, the delivery time prediction apparatus 400 may be implemented in the form of a computer program, and the computer program may be run on a computer device as shown in FIG. 5 . Can be stored in the memory of computer equipment and form each program module of this express delivery aging prediction device 400, such as, the information acquisition module 410 shown in Figure 4, feature engineering module 420, model analysis module 430 and aging determination module 440; The computer program enables the processor to execute the steps in the method for predicting the delivery time of each embodiment of the application described in this specification. For example, the computer equipment shown in FIG. 5 may execute step S201 through the information acquisition module 410 in the delivery time prediction apparatus 400 shown in FIG. 4 . The computer device can execute step S202 through the feature engineering module 420 . The computer device can execute step S203 through the model analysis module 430 . The computer device may execute step S204 through the aging determination module 440 . Wherein, the computer device includes a processor, a memory and a network interface connected through a system bus. Wherein, the processor of the computer device is used to provide calculation and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used to communicate with external computer devices via a network connection. When the computer program is executed by a processor, a method for predicting the time limit of express shipments is realized.

Those skilled in the art can understand that the structure shown in Figure 5 is only a block diagram of a part of the structure related to the solution of this application, and does not constitute a limitation to the computer equipment on which the solution of this application is applied. The specific computer equipment can be More or fewer components than shown in the figures may be included, or some components may be combined, or have a different arrangement of components.

In some embodiments of the present application, a computer device is provided, including one or more processors; memory; and one or more application programs, wherein one or more application programs are stored in the memory and configured as The processor executes the steps of the method for predicting the timeliness of express shipments above. Here, the steps in the express delivery aging prediction method may be the steps in the express delivery aging prediction methods in the above-mentioned various embodiments.

In some embodiments of the present application, a computer-readable storage medium is provided, which stores a computer program, and the computer program is loaded by a processor, so that the processor executes the steps of the above-mentioned method for express delivery timeliness prediction. Here, the steps in the express delivery aging prediction method may be the steps in the express delivery aging prediction methods in the above-mentioned various embodiments.

Those of ordinary skill in the art can understand that the realization of all or part of the processes in the methods of the above embodiments can be completed by instructing related hardware through computer programs, and the computer programs can be stored in non-volatile computer-readable storage In the medium, when the computer program is executed, it may include the processes of the embodiments of the above-mentioned methods. Any reference to memory, storage, database or other media used in the various embodiments provided in this application may include at least one of non-volatile and volatile memory. The non-volatile memory may include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory or optical memory, and the like. Volatile memory may include random access memory (Random Access Memory, RAM) or external cache memory. As an illustration and not a limitation, the RAM can be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (Dynamic Random Access Memory, DRAM).

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

A method, device, computer equipment, and storage medium for predicting the timeliness of express shipments provided by the embodiments of the present application have been described above in detail. In this paper, specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only It is used to help understand the method and its core idea of the present invention; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation and scope of application. In summary, this specification The content should not be construed as a limitation of the invention.

Claims (10)

1. A method for express delivery timeliness prediction, comprising: Obtain the shipment attribute information of the target shipment; performing feature engineering processing on the attribute information of the express to obtain the express characteristics of the target express; Input the features of the express shipment into the trained aging prediction model, and output the predicted probability value of the target shipment under each preset aging type; wherein, the trained aging prediction model is composed of a fully connected layer, a dropout layer, Composed of fusion layer and normalization layer; Each of the predicted probability values is analyzed to determine the logistics timeliness of the target express. 2. The method according to claim 1, wherein the input of the features of the express shipment into the trained aging prediction model, and outputting the predicted probability values of the target express shipment under each preset aging type, include: Combining the features of the express mail to obtain the merged express mail features; Inputting the features of the shipment into the trained aging prediction model, and outputting the first associated information inside the features; and Inputting the combined express features into the trained aging prediction model, and outputting the second correlation information between the features; The first association information and the second association information are analyzed to obtain the predicted probability values of the target express shipment under each preset aging type. 3. The method according to claim 2, wherein the input of the express features into the trained aging prediction model, and the output of the first associated information inside the features include: The express feature is input to the trained time-effect prediction model, and the express feature is classified through the fully connected layer and the dropout layer to obtain an initial feature vector and a differentiation coefficient of the initial feature vector; Fusing the initial feature vector and the differentiation coefficient through the fusion layer to obtain a target feature vector carrying the differentiation coefficient; The normalization process is performed on the target feature vector through the normalization layer to obtain the first associated information inside the feature. 4. The method according to claim 2, wherein the analysis of the first associated information and the second associated information to obtain the predicted probability value of the target express shipment under each preset aging type includes: : Obtaining a first feature matrix corresponding to the first associated information, and a second feature matrix corresponding to the second associated information; performing matrix multiplication of the first feature matrix and the second feature matrix to obtain comprehensive features; Through the target fully-connected layer carrying the sigmoid function, the feature classification is performed on the comprehensive feature, and the predicted probability value of the target express shipment under each preset aging type is obtained. 5. The method according to any one of claims 1 to 4, further comprising: Build an initial time-efficiency prediction model; Preliminary training is performed on the initial time-effect prediction model based on a preset initial learning rate until the number of training times reaches a preset first threshold to obtain a time-effect prediction model after preliminary training; Based on the preset peak learning rate, perform horizontal training on the time-effect prediction model after the preliminary training, until the number of training times reaches the preset second threshold, and obtain the time-effect prediction model after the level training; According to the trigonometric function curve of the peak learning rate, the time-effect prediction model after the level training is debugged and trained to obtain the trained time-effect prediction model. 6. The method according to claim 1, wherein said carrying out feature engineering processing to said express piece attribute information to obtain the express piece feature of said target express piece includes: Extracting the address information, time information and product information included in the attribute information of the express shipment; performing feature engineering processing on the address information to obtain outlet information and distance information as the express features; and Perform feature engineering processing on the time information to obtain holiday information and delivery time as the express features; and Perform feature engineering processing on the product information to obtain product type information as the express feature. 7. The method according to claim 1, wherein the analysis of each of the predicted probability values to determine the logistics timeliness of the target express mail comprises: Screening out the maximum value of each of the predicted probability values to obtain the target predicted probability value; The timeliness type corresponding to the target predicted probability value is determined as the logistics timeliness of the target express. 8. A device for predicting the aging of express mail, characterized in that it comprises: An information acquisition module, configured to acquire the attribute information of the target express; A feature engineering module, configured to perform feature engineering processing on the attribute information of the shipment to obtain the shipment features of the target shipment; The model analysis module is used to input the characteristics of the express shipment into the trained aging prediction model, and output the predicted probability value of the target express shipment under each preset aging type; wherein, the trained aging prediction model is composed of all Consists of connection layer, dropout layer, fusion layer and normalization layer; The timeliness determination module is used to analyze each of the predicted probability values to determine the logistics timeliness of the target express. 9. A computer device, characterized in that the computer device comprises: one or more processors; memory; and one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the processor to implement any one of claims 1 to 7 Express delivery time prediction method. 10. A computer-readable storage medium, characterized in that a computer program is stored thereon, and the computer program is loaded by a processor to execute the method for express delivery aging prediction according to any one of claims 1 to 7. step.

Images