Packaging recommendation method, device, computer equipment and storage medium

Abstract

The present application provides a packaging recommendation method, device, computer equipment and storage medium, the method comprising: obtaining the object circulation data of the target object, the object circulation data including the user identification of the consignee to whom the target object belongs; determining the target used packaging data in the pre-stored historical used packaging data according to the user identification; if the target used packaging data does not meet the preset packaging recommendation triggering condition, classifying and predicting the preferred packaging data in the target used packaging data based on the trained packaging classifier to obtain the recommended label of the preferred packaging data; determining the recommended packaging of the target object according to the recommended label. The adoption of this method can not only save labor costs, but also improve the efficiency and accuracy of packaging recommendation.

Description

Technical Field

[0001] The present application relates to the field of logistics packaging, and specifically to a packaging recommendation method, device, computer equipment and storage medium. Background Art

[0002] Logistics packaging is a necessary material that combines the characteristics of the packaged goods (including physical and chemical properties), volume, quantity, and circulation environment, and uses specific materials to control the temperature and humidity requirements of the goods during transportation. With the rapid development of society, the demand for logistics packaging continues to expand. How to quickly and effectively select logistics packaging is one of the goals of the current logistics industry to improve its efficiency.

[0003] However, due to the wide variety of existing packaging materials and filling items, and the fact that the industry has not yet formulated unified specifications and standards, couriers are required to select packaging materials based on their personal judgment and historical experience during daily delivery operations, which results in high labor costs.

[0004] Therefore, the existing technology has the problem of low efficiency in selecting logistics packaging. Summary of the invention

[0005] Based on this, it is necessary to provide a packaging recommendation method, device, computer equipment and storage medium that can improve the efficiency of packaging selection in response to the above technical problems.

[0006] In a first aspect, the present application provides a packaging recommendation method, the packaging recommendation method comprising:

[0007] Acquire the object circulation data of the target object, wherein the object circulation data includes the user identification of the sender or recipient of the target object;

[0008] According to the user identifier, determining target used packaging data in pre-stored historical used packaging data;

[0009] If the target used packaging data does not meet the preset packaging recommendation triggering condition, classifying and predicting the preferred packaging data in the target used packaging data based on the trained packaging classifier to obtain a recommended label for the preferred packaging data;

[0010] Determine the recommended packaging of the target item according to the recommended label.

[0011] In a second aspect, the present application provides a packaging recommendation device, the packaging recommendation device comprising:

[0012] A data acquisition module, used to acquire the object circulation data of the target object, wherein the object circulation data includes the user identification of the sender or receiver of the target object;

[0013] A data determination module, configured to determine target used packaging data from pre-stored historical used packaging data according to the user identifier;

[0014] a label acquisition module, configured to classify and predict the preferred packaging data in the target used packaging data based on a trained packaging classifier to obtain a recommended label for the preferred packaging data if the target used packaging data does not meet a preset packaging recommendation triggering condition;

[0015] The packaging determination module is used to determine the recommended packaging of the target item according to the recommended tag.

[0016] In a third aspect, the present application further provides a server, the server comprising:

[0017] one or more processors;

[0018] Memory; and

[0019] 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 the packaging recommendation method.

[0020] In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is loaded by a processor to execute the steps in the packaging recommendation method.

[0021] The above-mentioned packaging recommendation method, device, computer equipment and storage medium obtain the user ID of the consignor or recipient of the target item by obtaining the item circulation data of the target item, so as to filter out the target used packaging data from the historical used packaging data according to the user ID, and then, when the target used packaging data does not meet the packaging recommendation triggering condition, it can classify and predict the preferred packaging data in the target used packaging data based on the trained packaging classifier, and finally determine the recommended packaging of the target item according to the recommended label of each preferred packaging data. The method is adopted to determine the recommended available packaging based on the associated application of historical data, which is more accurate than the traditional manual judgment method. At the same time, when the historical data analysis alone cannot meet the packaging recommendation needs, it is proposed to use the packaging classifier to realize the classification prediction of packaging, which can not only save labor costs, but also improve the efficiency and accuracy of packaging recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 It is a schematic diagram of a scenario of a packaging recommendation method in an embodiment of the present application;

[0024] Figure 2 It is a schematic diagram of the process of the recommended packaging method in the embodiment of the present application;

[0025] Figure 3 It is a flowchart of the steps of obtaining historical used packaging data in an embodiment of the present application;

[0026] Figure 4 This is a specific flow chart of the classification result acquisition step in the embodiment of the present application;

[0027] Figure 5 It is a specific flow chart of the recommended packaging method in the embodiment of the present application;

[0028] Figure 6 It is a structural schematic diagram of a packaging recommendation device in an embodiment of the present application;

[0029] Figure 7 It is a schematic diagram of the structure of a computer device in an embodiment of the present application. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0031] In the description of the present application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present application, "plurality" means two or more, unless otherwise clearly and specifically defined.

[0032] In this application, the word "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in this application.

[0033] It should be noted that since the method of the embodiment of the present application is executed in a computer device, the processing objects of each computer device exist in the form of data or information. For example, time is actually time information. It can be understood that if size, quantity, position, etc. are mentioned in subsequent embodiments, they are all corresponding data for processing by the computer device. The details will not be repeated here.

[0034] The embodiments of the present application provide a packaging recommendation method, apparatus, computer device, and storage medium, which are described in detail below.

[0035] See also Figure 1 , Figure 1 A schematic diagram of a scenario of a packaging recommendation method provided in an embodiment of the present application, which can be applied to a packaging recommendation system, wherein the packaging recommendation system includes a terminal 100 and a server 200. The terminal 100 can be a device that includes both receiving and transmitting hardware, that is, a device that has receiving and transmitting hardware that can perform two-way communication on a two-way communication link. Such a device may include: a cellular or other communication device having a single-line display or a multi-line display or a cellular or other communication device without a multi-line display. The terminal 100 can specifically be a desktop terminal or a mobile terminal, and the terminal 100 can specifically be one of a mobile phone, a tablet computer, a laptop computer, etc. The server 200 can be an independent server, or a server network or a server cluster composed of servers, which includes but is not limited to a computer, a network host, a single network server, a plurality of network server sets, or a cloud server composed of a plurality of servers. Among them, the cloud server is composed of a large number of computers or network servers based on cloud computing (Cloud Computing).

[0036] Those skilled in the art will understand that Figure 1 The application environment shown in the figure is only one application scenario of the present application solution and does not constitute a limitation on the application scenario of the present application solution. Other application environments may also include Figure 1More or less computer equipment as shown in Figure 1 Only one server 200 is shown in the figure. It is understandable that the packaging recommendation system may also include one or more other servers, which are not specifically limited here. Figure 1 As shown, the packaging recommendation system may also include a memory 300 for storing data, such as logistics data, for example, various data of the logistics platform, such as logistics transportation information of logistics outlets such as transfer yards, specifically, express information, delivery vehicle information and logistics outlet information.

[0037] It should be noted that Figure 1 The scenario diagram of the packaging recommendation system shown is only an example. The packaging recommendation system and scenario described in the embodiment of the present invention are intended to more clearly illustrate the technical solution of the embodiment of the present invention, and do not constitute a limitation on the technical solution provided by the embodiment of the present invention. Those skilled in the art can appreciate that with the evolution of the packaging recommendation system and the emergence of new business scenarios, the technical solution provided by the embodiment of the present invention is also applicable to similar technical problems.

[0038] It should also be noted that in the logistics industry and supply chain management process, there is a gradual consensus on the impact that the packaging link may have on the logistics process, especially the transportation and storage links. People have begun to consider packaging design in combination with logistics needs, processing and manufacturing, marketing, and product design requirements in order to achieve the protection function of the goods and improve logistics efficiency. However, with the increasing number of packaging material types and designs, how to select suitable and available packaging materials has become a new problem. In order to solve this problem, this application provides a packaging recommendation method to solve the problem that the existing packaging material categories and total number are huge and the company's delivery personnel cannot quickly and effectively select suitable packaging, thereby improving their logistics packaging operation efficiency.

[0039] like Figure 2 As shown, in one embodiment, a packaging recommendation method is provided. This embodiment mainly applies this method to the above Figure 1 The server 200 in FIG. 1 is used as an example. Figure 2 The packaging recommendation method specifically includes steps S201 to S204, which are as follows:

[0040] S201, obtaining the object circulation data of the target object, wherein the object circulation data includes the user identification of the sender or recipient of the target object.

[0041] The target item refers to the goods currently to be packaged and transported.

[0042] Among them, the user identification includes the user identification of the sender to whom the target item belongs, and the user identification of the delivery employee to whom the target item belongs; the user identification of the sender is such as: ID number or mobile phone number "123", and the user identification of the delivery user is such as: work number "abc".

[0043] Specifically, before recommending suitable packaging (packaging materials) for the target item in the actual application scenario, the server 200 first needs to obtain the item circulation data of the target item, and the item circulation data can be uploaded by the terminal 100 held by the company's delivery staff (scanning or inputting). The item circulation data should at least include the unique identity of the sender to whom the target item belongs, the unique identity of the delivery employee to whom it belongs, the text information of the category to which it belongs, and the weight, etc., which are used as the analysis basis for subsequent packaging recommendations. It can be understood that in other embodiments, the item circulation data can actually cover more data content based on specific business needs. For example, the volume, size, material, purpose, color, light-proof property, etc. of the target item. Therefore, this application does not specifically limit the content of the item circulation data. Although only a few of the data analyses are described in detail later, the interchangeability between the data is not excluded, that is, the basis for subsequent analysis is determined by the actual application needs.

[0044] S202: Determine target used packaging data in pre-stored historical used packaging data according to the user identifier.

[0045] Among them, the historical used packaging data refers to the packaging materials that have been historically stored and used, and their related data, such as the number of times the packaging material has been used historically, and the usage ratio of the packaging material in all historically used packaging materials.

[0046] The target used packaging data refers to the historically used packaging data corresponding to the user ID of the target item, that is, the historically used packaging materials corresponding to the user ID and their associated data.

[0047] Specifically, before the server 200 determines the target used packaging data of the target item, it first needs to obtain all stored historical used packaging data from the memory 300. However, since the historical used packaging data actually comes from the historical waybill data, each historical waybill data is not only marked with a different waybill number, but also records the detailed content of the waybill, that is, it includes the historical used packaging data corresponding to a single user ID. Therefore, in order to obtain the target used packaging data corresponding to the target item, the user ID of the consignor to whom the target item belongs must be matched one by one with the user ID corresponding to the historical used packaging data to obtain the historical used packaging data that matches the user ID as the target used packaging data.

[0048] For example, there are three historical waybill data currently obtained, which record the historical used packaging data corresponding to different user IDs: user ID A-packaging material ID a, user ID B-packaging material ID b, user ID C-packaging material ID c. If the user ID of the target item is A, then the target used packaging data is packaging material ID c. In addition, other statistical data associated with the packaging material ID c may also be included, such as the historical number of times user C has used packaging material c is 5, the number of times accounts for 10%, etc.

[0049] S203: If the target used packaging data does not meet the preset packaging recommendation triggering condition, classify and predict the preferred packaging data in the target used packaging data based on the trained packaging classifier to obtain a recommended label for the preferred packaging data.

[0050] Among them, the packaging classifier can be used to classify packaging materials. The packaging classifier in this embodiment is set to a deep neural network classifier (Deep Neural Networks, DNN). It can be understood that in other embodiments, the packaging classifier can also be set to at least one of a convolutional neural network classifier (Convolutional Neural Networks, CNN), a K-nearest neighbor (KNN, k-Nearest Neighbor) classifier, an adboost classifier, an xgboost (Extreme Gradient Boosting) classifier, a logistic regression classifier, and a support vector machine classifier according to actual application requirements.

[0051] Among them, the preferred packaging data can be packaging data of preferred packaging materials screened out from the target used packaging data based on a preset packaging material preference algorithm, that is, including the packaging material identification of the preferred packaging material, the number of times the packaging material is used, the proportion of the number of times the packaging material is used, etc.

[0052] The recommendation label may refer to two types of labels, for example, a recommendation label of "1" indicates recommendation, and a recommendation label of "0" indicates non-recommendation.

[0053] Specifically, after the server 200 obtains the target used packaging data corresponding to the target item, it can further determine the preset packaging recommendation trigger condition, and determine whether the currently obtained target used packaging data meets the packaging recommendation requirements based on the packaging recommendation trigger condition. If the target used packaging data does not meet the packaging recommendation trigger condition, it is necessary to use the trained packaging classifier to further analyze the target used packaging data, that is, to obtain the recommended labels of each preferred packaging (identification) by making classification predictions on the preferred packaging data screened out from the target used packaging data, so as to finally determine the recommended packaging of the target item using the recommended labels obtained by machine classification.

[0054] More specifically, determining whether the target used packaging data meets the preset packaging recommendation trigger conditions is actually to determine whether the statistical information associated with the packaging material identification in the target used packaging data meets the conditions, that is, the role of the number of times the packaging material is used and the proportion of the number of times the packaging material is used mentioned above, which can be used to determine whether the target used packaging data meets the preset packaging recommendation trigger conditions.

[0055] For example, user C, who is currently to send a target item, has used packaging material c 5 times in history and the proportion of the times is 10%, while the preset packaging recommendation trigger conditions include a critical value of 10 for the times and a critical value of 18% for the proportion. The number of times and the proportion of packaging material c used by user C in history have not reached their respective critical values, that is, the preset packaging recommendation trigger conditions are not met, and the trained packaging classifier needs to be used to complete the subsequent classification prediction operation.

[0056] S204: Determine a recommended package for the target item according to the recommended tag.

[0057] Among them, recommended packaging refers to suitable and available packaging materials.

[0058] Specifically, after the server 200 determines that the target used packaging data does not meet the preset packaging recommendation triggering condition and obtains the recommended label of the preferred packaging data in the target used packaging data, it can filter out the recommended packaging material identification according to each recommended label as the recommended packaging of the target item. It is understandable that after the server 200 analyzes and obtains the recommended packaging of the target item, it can feed back the recommended packaging to the terminal 100, so that the enterprise delivery personnel can quickly locate the recommended packaging of the current target item to be packaged, thereby improving their work efficiency.

[0059] For example, the recommended label of packaging material identification a is "1", the recommended label of packaging material identification b is "0", and the recommended label of packaging material identification c is "1". Label "1" indicates that it can be recommended for application, and label "0" indicates that it cannot be recommended for application. Then the recommended packaging of the target item is packaging material a and packaging material c. Packaging materials a and c can be fed back by server 200 to terminal 100 for display.

[0060] In this embodiment, the user ID of the consignor to whom the target item belongs is obtained by obtaining the item circulation data of the target item, so that the target used packaging data can be screened out from the historical used packaging data according to the user ID, and then, when the target used packaging data does not meet the packaging recommendation triggering condition, the preferred packaging data in the target used packaging data can be classified and predicted based on the trained packaging classifier, and finally the recommended packaging of the target item is determined based on the recommended labels of each preferred packaging data. The method adopts the method to determine the recommended available packaging based on the associated application of historical data, which is more accurate than the traditional manual judgment method. At the same time, when the historical data analysis alone cannot meet the packaging recommendation requirements, it is proposed to use the packaging classifier to realize the classification prediction of packaging, which can not only save labor costs, but also improve the efficiency and accuracy of packaging recommendation.

[0061] In one embodiment, before step S202, the following steps are specifically included:

[0062] S301, obtaining historical waybill data, wherein the historical waybill data includes a sender ID, a receiver ID, and a packaging material ID;

[0063] S302, according to the sender identifier, the packaging material identifiers are aggregated to obtain a first packaging material identifier associated with the sender identifier, and the number of times and the proportion of packaging usage corresponding to the first packaging material identifier are obtained by counting, and the results are combined as first historical used packaging data;

[0064] S303, according to the delivery personnel identification, the packaging material identification is aggregated to obtain a second packaging material identification associated with the delivery personnel identification, and the number of times and the proportion of packaging usage corresponding to the second packaging material identification are obtained by counting, and the results are combined as second historical used packaging data;

[0065] S304: Determine the first historical used packaging data and the second historical used packaging data as the historical used packaging data.

[0066] Among them, the waybill refers to a "one-time" written contract between the carrier and the shipper regarding the transportation of goods. It is the contract certificate for the transportation of goods and the transportation agent, and it is also the certificate for the transportation operator to accept the goods and be responsible for the safekeeping and delivery during the transportation period. The waybill information should list the name, packaging, various fees and amounts of the consigned goods, the departure and arrival stations, the consignor and consignee, the carriage and arrival dates, and other matters related to the transportation of goods. Historical waybill information refers to the waybill information recorded in history.

[0067] The sender ID may be a globally unique mark indicating the identity of the sender, such as an ID number, a phone number, etc. It is understandable that in other embodiments, the ID may be a special character consisting of numbers, uppercase and lowercase letters, underscores, etc. according to business requirements.

[0068] The delivery person identification may be a globally unique mark indicating the identity characteristics of the delivery person (delivery staff) of the enterprise, such as an ID number, a telephone number, a work number, etc. It is understandable that in other embodiments, the identification may also be composed of numbers, uppercase and lowercase letters, underscores, and other special characters according to business requirements.

[0069] Among them, the packaging material identification can be a globally unique mark that indicates the identity characteristics of the packaging material, which is an identification composed of numbers, uppercase and lowercase letters, underscores and other special characters according to business requirements.

[0070] The number of times a packaging material is used refers to the number of times a certain packaging material is used, for example, 3, 5, etc.

[0071] Among them, the packaging usage ratio refers to the percentage of the number of times a certain packaging material is used to the number of times all packaging materials are used, for example, 20%, 50%, etc.

[0072] Specifically, before the server 200 uses the historical used packaging data to obtain the target used packaging data corresponding to the user ID to which the target item belongs, it is necessary to first collect the historical used packaging data. Since the historical used packaging data is derived from the historical waybill data, it is necessary to classify and summarize the sender ID, the receiver ID and the packaging material ID included in the historical waybill data after obtaining the historical waybill data, that is, using the personnel ID as the classification basis, respectively obtain the packaging material ID corresponding to each sender ID and the packaging material ID corresponding to each receiver ID, obtain the first packaging material ID associated with the sender ID, and the second packaging material ID associated with the receiver ID, and then statistically summarize the first packaging material ID associated with each sender ID, and obtain the number of packaging uses and the packaging use ratio corresponding to the same first packaging material ID, then a single sender ID-first packaging material ID-number of packaging uses-packaging use ratio can be used as the first historical used packaging data. Similarly, the operation of obtaining the second historical used packaging data for the receiver ID is consistent.

[0073] For example, Figure 3As shown in the figure, the historical waybill data contains the purchase point data of the entire logistics life cycle. Its main fields include "p-ID" (sender ID), "e-ID" (receiver ID), and "package-ID" (packaging material ID). According to the data summarized by the field "p-ID", the first historical used packaging data corresponding to each sender ID can be gradually filtered. In addition, the second historical used packaging data can be obtained in the same way as Figure 3 The same as shown.

[0074] In this embodiment, by summarizing and organizing the historical waybill data in advance before analyzing and obtaining the target used packaging data, the historical used packaging data required for subsequent use can be obtained and stored for backup, which can not only save labor costs but also improve the efficiency and accuracy of packaging recommendations.

[0075] In one embodiment, the user identifier includes a first user identifier and a second user identifier, the historically used packaging data includes a first historically used packaging data and a second historically used packaging data, and step S202 specifically includes the following steps:

[0076] S401, determining at least one sender identifier in the first historical used packaging data, and determining at least one receiver identifier in the second historical used packaging data;

[0077] S402, acquiring first historical used packaging data corresponding to the target sender identifier that matches the first user identifier, to obtain first target used packaging data;

[0078] S403, acquiring second historical used packaging data corresponding to the target delivery personnel identifier that matches the second user identifier, to obtain second target used packaging data;

[0079] S404: Determine the first target used packaging data and the second target used packaging data as the target used packaging data.

[0080] The first user identification refers to the sender identification, and the second user identification refers to the receiver identification. The concepts of the sender identification and the receiver identification have been described in detail in the above embodiments and will not be repeated here.

[0081] Specifically, in the above embodiment, it has been explained in detail how to obtain the historical used packaging data before obtaining the target used packaging data, that is, the historical used packaging data actually includes the first historical used packaging data corresponding to the sender's ID, and the second historical used packaging data corresponding to the receiver's ID. Therefore, before determining the target used packaging data corresponding to the current user ID, it is necessary to first determine the personnel IDs contained in each of the two historical used packaging data, and then respectively match the first user ID with the sender's ID in the first historical used packaging data, and match the second user ID with the receiver's ID in the second historical used packaging data, so as to obtain the matching target used packaging data.

[0082] For example, the first history has been represented by the package data as a set P = {(p, T p )|p∈N *}, the second history has been expressed as a set of packaged data E = {(e, T e )|p∈R},T p =(p,t,count,percent),T e =(e, t, count, percent), where “p” represents the sender ID, “e” represents the delivery person ID, “t” represents the packaging material ID, “count” represents the number of times the packaging is used, and “percent” represents the percentage of packaging used.

[0083] For another example, for the first user ID p1 to which the target item belongs, the first target used packaging data can be obtained by filtering from the set P: For the second user ID e1 of the target item, the second target used packaging data can be obtained by filtering from the set E

[0084] In this embodiment, the target used packaging data of each sender and receiver of the target item is obtained by matching the user identifier, so that the two types of target used packaging data can be used to analyze and recommend packaging later, which can not only save labor costs but also improve the efficiency and accuracy of packaging recommendations.

[0085] In one embodiment, the target used packaging data includes first target used packaging data and second target used packaging data, and step S203 specifically includes the following steps:

[0086] S501, determining a preset package recommendation trigger condition, where the package recommendation trigger condition includes a first trigger condition, a second trigger condition, and a third trigger condition;

[0087] S502, if the first target used packaging data does not satisfy any one of the first trigger condition, the second trigger condition and the third trigger condition, respectively obtaining preferred packaging data from the first target used packaging data and the second target used packaging data;

[0088] S503: Based on the trained packaging classifier, classification prediction is performed on the preferred packaging data to obtain a recommended label for the preferred packaging data.

[0089] Among them, the first target used packaging data can be expressed as The second target used package data can be expressed as Combination Figure 3 As shown, Including the sender ID "p1", the packaging material ID "t", the number of times the packaging is used "count", and the proportion of packaging used "percent". Same reason.

[0090] Specifically, the packaging recommendation trigger conditions include a first trigger condition, a second trigger condition and a third trigger condition. The judgment order of the three conditions is interrelated, that is, the target used packaging data needs to be used in sequence with the first trigger condition, the second trigger condition and the third trigger condition for judgment. If all three conditions are met, the packaging materials stored in the target used packaging data can be used as the recommended packaging for the target item. If any of the three conditions is not met, the condition judgment is stopped, and the preferred packaging data are directly screened out for the first target used packaging data and the second target used packaging data currently obtained, and then the trained packaging classifier is used for classification prediction to obtain the recommended labels for each packaging material in the preferred packaging data.

[0091] More specifically, the data basis for executing the conditional judgment is the first target used packaging data, that is, Second target used packaging data As auxiliary information, it is convenient for the auxiliary package classifier to continue analyzing and determining the recommended package when no recommended package can be obtained by analyzing the used package data of the first target alone.

[0092] In this embodiment, by matching and analyzing the used packaging data of the first target with the preset first, second, and third trigger conditions one by one, when the match does not meet any condition, the recommended label of the preferred packaging data can be obtained based on the trained packaging classifier, which can not only save labor costs but also improve the efficiency and accuracy of packaging recommendations.

[0093] In one embodiment, step S502 specifically includes the following steps:

[0094] S601, determining a packaging material identifier in the first target used packaging data, wherein the packaging material identifiers respectively have corresponding packaging usage times and packaging usage ratios;

[0095] S602: if the number of times the package is used is less than a preset number threshold, determining that the first target used package data does not meet the first trigger condition;

[0096] S603: if the package usage ratio is less than a preset ratio threshold, determining that the first target used package data does not satisfy the second trigger condition;

[0097] S604: if the total number of packaging material identifiers that meet the first trigger condition and the second trigger condition is less than a preset quantity threshold, determine that the first target used packaging data does not meet the third trigger condition;

[0098] S605: If the first target used packaging data does not satisfy any one of the first trigger condition, the second trigger condition and the third trigger condition, respectively obtain the preferred packaging data from the first target used packaging data and the second target used packaging data.

[0099] The number threshold refers to the conditional trigger critical value at which the number of times the package is used meets the first trigger condition, for example, 3, 5, etc.

[0100] The percentage threshold refers to the critical value of the condition trigger at which the packaging usage percentage meets the second trigger condition, for example, 10%, 20%, etc.

[0101] Among them, the quantity threshold refers to the conditional trigger critical value at which the total number of packaging material identifications meets the third trigger condition, for example, 5, 10, etc. It should be noted that the setting of the quantity threshold is associated with the interface size of the terminal 100 that displays the recommended packaging. For example, when the terminal 100 is a mobile phone, the quantity threshold can be set to 5, indicating that a page of the mobile phone can only display 5 packaging materials at the same time; when the terminal 100 is a computer, the quantity threshold can be set to 10, indicating that a page of the computer can display 10 packaging materials at the same time. Of course, the terminal interface size mentioned here is relative, and the embodiment of the present application only illustrates this type of business demand, and it does not exclude that in other embodiments, the number of pages, the size of the icons of the packaging materials displayed on the page, etc. need to be considered.

[0102] Specifically, in the above embodiment, it has been explained that the packaging trigger condition is determined based on the first target used packaging data. Therefore, the server 200 needs to The sender identification "p1", packaging material identification "t", number of times the packaging is used "count", and the packaging usage ratio "percent" contained in the are determined separately according to the settings of the three trigger conditions.

[0103] More specifically, the first trigger condition is set as: The number of times the package is used "count" in is greater than or equal to the number threshold count_T, that is, if If count of a certain package t is less than count_T, it means that the historical usage times of the packaging material do not meet the limited recommendation conditions, and it cannot be included in the recommendation set as a recommended packaging; the second trigger condition is set as: The packaging usage ratio "percent" in is greater than or equal to the number threshold percentage_T, that is, if If the percentage of a certain package t in the example is less than percentage_T, it means that the historical usage ratio of the packaging material does not meet the limited recommendation conditions, and it cannot be included in the recommendation set as a recommended packaging. If both the “count” and the “percent” of a certain packaging material t in the recommendation set satisfy the corresponding triggering condition, the packaging material t can be included in the recommendation set to continue to perform the subsequent determination of the third triggering condition.

[0104] However, the third trigger condition is set as follows: the total quantity total of all packaging materials t included in the recommendation set must be greater than or equal to the quantity threshold total_T, that is, if the total quantity total of each type of packaging material t in the recommendation set is less than total_T, it means that the quantity of packaging materials in the recommendation set does not meet the preset required quantity of recommended packaging materials, and then it is necessary to consider the role of the second target used packaging data, that is, respectively obtain the preferred packaging data in the first target used packaging data and the second target used packaging data to continue analyzing and determining the recommended packaging of the target item.

[0105] For example, combined with Figure 3 , the current first target has used packaging data The first target used packaging data of the first user ID "AAA0001" contains three packaging materials "package-001A", "package-001B" and "package-001C". If the number threshold is 18, the proportion threshold is 30% (0.3), and the quantity threshold is 2, then the packaging materials that meet the first trigger condition are "package-001A" and "package-001B", and the packaging material that meets both the first and second trigger conditions is "package-001A". However, due to Only one packaging material can be recommended, and the current required recommendation quantity threshold is 2, so "package-001A" cannot be fed back to the terminal 100 as the recommended packaging of the target item. The server 200 also needs to combine the second target used packaging data For further analysis.

[0106] It should be noted that the determination order of the first and second trigger conditions can be replaced according to actual application requirements. This application only provides a detailed description of one of the determination orders, and does not impose specific restrictions on the actual determination order of the two.

[0107] In this embodiment, by analyzing the packaging material identification in the first target used packaging data, as well as the number of packaging uses and the packaging usage ratio of the packaging material identification, it is determined whether the first target used packaging data meets the three preset packaging recommendation trigger conditions, which can not only save labor costs but also improve the efficiency and accuracy of packaging recommendations.

[0108] In one embodiment, step S605 specifically includes the following steps:

[0109] S701, if the first target used packaging data does not satisfy any one of the first trigger condition, the second trigger condition, and the third trigger condition, then based on the number of times the packaging is used, respectively arranging the packaging material identifiers in the first target used packaging data and the second target used packaging data in descending order to obtain a first packaging material identifier sequence and a second packaging material identifier sequence;

[0110] S702, acquiring the first N packaging material identifications in the first packaging material identification sequence and the second packaging material identification sequence to obtain a first target packaging material identification and a second target packaging material identification; N≥1;

[0111] S703: Determine the first target packaging material identifier and the second target packaging material identifier as the preferred packaging data.

[0112] Specifically, if the first target has packaged data If the preset packaging recommendation triggering condition is not met, the server 200 may combine the second target used packaging data Further analysis of recommended packaging, since the second goal has used packaging data In the above embodiment, the acquisition method is explained, that is, the second user identification of the target item is collected and summarized. Therefore, under the premise that the used packaging data of the first target does not meet the packaging recommendation triggering condition, the used packaging data of the second target can be combined for auxiliary analysis, that is, according to the number of times the packaging is used "count", the used packaging data of the second target can be obtained respectively. and The top-N packaging data (packaging material identifier t) in and and and The N packaging material identifiers t and their corresponding "count" and "percent" can be used as the preferred packaging data.

[0113] In this embodiment, by limiting the sorting factors of the packaging material identification to screen out the preferred packaging data, it is not only possible to save labor costs, but also to improve the efficiency and accuracy of packaging recommendation.

[0114] In one embodiment, the item circulation data includes the consignment information and consignment weight data of the target item, and step S503 specifically includes the following steps:

[0115] S801, determining the trained packaging classifier, where the trained packaging classifier is obtained by training with pre-stored historical waybill data;

[0116] S802, inputting the consignment information, the consignment weight data, and the preferred packaging data into the trained packaging classifier;

[0117] S803, obtaining the classification result output by the trained packaging classifier to obtain a recommended label for the preferred packaging data; the recommended label is obtained by the trained packaging classifier according to the consignment information, the consignment weight data and the preferred packaging data.

[0118] Specifically, the server 200 obtains the first target used packaging data respectively Preferred packaging data And the second target has used packaging data Preferred packaging data Afterwards, the consignment information “c” and consignment weight data “w” contained in the object circulation data of the target object can be combined as the input parameters of the trained package classifier, that is, Input X σ In order to package the classifier for classification prediction, we can get and Recommended labeling for each packaging material included in the .

[0119] In this embodiment, by using a trained packaging classifier to analyze consignment information, consignment weight data and preferred packaging data, classification prediction of packaging materials can be achieved, and the recommended packaging of the target item can be determined based on the prediction results, which can not only save labor costs but also improve the efficiency and accuracy of packaging recommendations.

[0120] In one embodiment, before step S802, the following steps are specifically included:

[0121] S901, cleaning and simplifying the consignment information, and performing vector conversion on the consignment weight data and the preferred packaging data;

[0122] S902, performing word segmentation processing on the simplified consignment information to obtain first input data, and performing vector merging on the converted consignment weight data and the preferred packaging data to obtain a second input vector; the second input vector and the first input data are used to be input into the trained packaging classifier.

[0123] Specifically, the input parameter X of the input wrapper classifier σ Previously, the server 200 also needed to respond to the input parameter X σ Preprocessing, that is, The text data in the consignment information "c" can first be cleaned (special characters are cleaned, and Chinese and English characters A(a)-Z(z), and numbers 0-9 are retained), and then traditional Chinese characters are converted into simplified Chinese characters. The Viterbi algorithm and the Hidden Markov Model (HMM) model are used to perform Chinese word segmentation, and finally the character strings obtained by word segmentation are connected with spaces to obtain the first input data "input-1".

[0124] For example, "I love China" is split into "I", "love", and "China", and connected by a space into "I_love_China".

[0125] More specifically, for The numerical data in the consignment weight data "w" can be first binned by weight and then converted into a one-hot vector. Among them, data binning (also known as discrete binning or segmentation) is a data preprocessing technique used to reduce the impact of minor observation errors. It is a method of grouping multiple continuous values ​​into a smaller number of "bins". Preferred packaging data in and First, one-hot vectorization can be performed on each packaging material identifier t to obtain Then perform element-wise multiplication to obtain the vector The second input vector "input-2" is obtained by combining this vector with the one-hot vector converted from the consignment weight data "w".

[0126] In this embodiment, the input parameters of the packaging classifier are preprocessed to obtain a data structure that can be input into the packaging classifier, and then the packaging classifier is used for classification prediction based on feasible input parameters, which can not only save labor costs but also improve the efficiency and accuracy of packaging recommendations.

[0127] In one embodiment, the trained packaging classifier includes a data input module, a data segmentation module, a first convolution module, a vector concatenation module, a link processing module, a multi-label classification module and an output module. Step S803 specifically includes the following steps:

[0128] S1001, obtaining, through the input module, first input data corresponding to the consignment information, and a second input vector corresponding to the consignment weight data and the preferred packaging data;

[0129] S1002, performing serialization processing on the first input data through the data segmentation module, and performing sequence length control processing on the first input data after the serialization processing to obtain fixed sequence data;

[0130] S1003, performing vector conversion on the fixed sequence data through the first convolution module, and performing convolution processing on the converted fixed sequence data to obtain a first convolution vector;

[0131] S1004, performing a splicing process on the first convolution vector and the second input vector by the vector splicing module to obtain a splicing vector;

[0132] S1005, performing link processing on the concatenated vector through a full link layer or a convolution link layer in the link processing module to obtain a link vector;

[0133] S1006, performing matrix transformation processing on the link vector by using a target activation function preset in the multi-label classification module to obtain a recommended label for the preferred packaging data;

[0134] S1007: Output the recommended tag as the classification result through the output module.

[0135] Specifically, see Figure 4, the packaging classifier proposed in this application includes at least two data input modules (input), a data segmentation module (Tokenization), a first convolution module (1D Separable CNN), a vector splicing module (Concat), a link processing module (DNN / CNN Layer), a multi-label classification module (Multilabel-label binary layer module) and an output module (output). Among them, the input module (input) is used to receive the first input data "input-1" and the second input vector "input-2" mentioned in the above embodiment, and the data segmentation module (Tokenization), the first convolution module (1D Separable CNN), the vector splicing module (Concat), the link processing module (DNN / CNN Layer) and the multi-label classification module (Multilabel-label binary layer module) are used to perform the following operations:

[0136] (1) The data tokenization module can be used to first perform text serialization operations to convert "input-1" into a sparse vector That is, the original text sequence is represented by id. In the actual operation process, the sparse vector needs to be converted into a dense vector (Dense Vector). The sparse vector is used here mainly to save storage space and avoid memory overflow. Then for the vector Perform text serialization length control operation, that is, to ensure that the sequence data input to the convolutional layer is of fixed length, vector Compare with the preset sequence length L and perform the following operations:

[0137] a. right Truncate and keep a vector of length L make b. right Expand "0" to make c. return

[0138] (2) The first convolution module (1D Separable CNN) actually includes two submodules, one for vector conversion and the other for vector convolution. The word2vector vector conversion steps include:

[0139] a. For vector or The elements in are converted into one-hot to obtain a dense matrix D∈N L*w, where w is the length of the dictionary; b. Transform the matrix D as follows, where m is the length of the encoded dense vector, and the column vector in D' represents the encoded word vector; cD·M=D',M∈R w*m ; Where M is the matrix that controls the change of the vector, R is the natural number matrix, and N is the integer matrix.

[0140] In addition, in the vector convolution step, 1D Separable convolution is performed using the 1D DepthwiseSeparable CNN convolution method, which includes the following 8 sub-steps:

[0141] a.Depthwise Convolution-1: Perform a one-dimensional (1D) single-channel convolution transformation on the input D' matrix, where the convolution kernel is The convolution stride is 1, and the vector in b∈R,D' i is a sub-region of D';

[0142] b.Pointwise Convolution-1: The vector obtained in the previous step Perform convolution in the channel dimension to obtain the matrix Among them because is a single channel, k is the convolution kernel, for any

[0143] c.Depthwise Convolution-2: For D 3 For further abstraction, the process is similar to step a above, except that the number of channels is greater than 1. Execute step a for each channel to obtain the matrix D 4 ;

[0144] d.Dropout: For The elements in D are set to 0 with a preset probability p, and D 5 ,This step is mainly to avoid overfitting of the model during training and improve the model's generalization ability during prediction;

[0145] e.Pointwise Convolution-2: corresponding to D 5 Perform convolution of the channel dimension to obtain the matrix Because D 5 For multiple channels, for any

[0146] f.Max pooling: for D 6 Perform max pooling on the matrix of each channel to compress the size of the matrix and avoid overfitting of the model;

[0147] g. The above steps cf are a Separable CNN unit. The network structure in actual production can be composed of multiple such units stacked together;

[0148] h.1D GobalAveragePooling: for D 6 Merge channels and find the average value to obtain vector

[0149] (3) The vector concatenation module (Concat) can be used to concatenate the encoded text vectors Concatenate it with the input vector of "input-2" to produce a new vector

[0150] (4) Link processing module (DNN / CNN Layer) is an optional structure for packaging separators. It can be DNN (full link layer) or CNN (convolutional link layer). This solution uses a layer of DNN to link get

[0151] (5) The multi-label classification module (Multilabel-label binary layer module) can use the Sigmoid function as the activation function:

[0152]

[0153] in, Relative to the specific model data, the function g represents the matrix transformation before the last layer of the network structure.

[0154] In this embodiment, by using a packaging classifier to classify and predict the recommended packaging of the target item, not only can labor costs be saved, but also the efficiency and accuracy of packaging recommendation can be improved.

[0155] In one embodiment, step S801 specifically includes the following steps:

[0156] S1101, obtaining historical waybill data;

[0157] S1102, inputting the historical waybill data into a pre-established packaging classifier for model training, wherein the pre-established packaging classifier is preset with a target loss function;

[0158] S1103, obtaining a training result output by the pre-established packaging classifier according to the target loss function;

[0159] S1104: If the training result satisfies a preset training end condition, the model training of the packaging classifier is ended to obtain the trained packaging classifier.

[0160] Specifically, the commonly used loss function for binary classification problems is binary cross entropy, as follows:

[0161]

[0162] Among them, y i ∈{0,1} is the actual label, p(y i )∈[0,1], which is equivalent to f(x) in the multi-label classification module, but this application needs to solve the problem of extreme imbalance between positive and negative samples. Therefore, the commonly used loss function cannot meet the needs of current model training and prediction, so the target loss function used in this application is as follows:

[0163]

[0164]

[0165]

[0166]

[0167]

[0168] cost=(cost1+cost2) / 2

[0169] in, is a very small real number, such as 1e -16 , which is mainly used to ensure that the denominator is not 0. There are two main advantages of using the above loss function: (1) There is no need to tune the threshold. For different thresholds, the F1 index of the model remains unchanged; (2) For the case of extremely unbalanced positive and negative samples, the commonly used loss function tends to judge all samples as negative samples. The above function can avoid this situation because it directly tunes the f1-score.

[0170] In this embodiment, the packaging classifier is trained by acquiring historical waybill data to obtain a trained packaging classifier for subsequent prediction, which can not only save labor costs but also improve the efficiency and accuracy of packaging recommendations.

[0171] In one embodiment, the target used packaging data includes first target used packaging data and second target used packaging data, and the recommended tag includes a first recommended tag. Step S204 specifically includes the following steps:

[0172] S1201, screening out the preferred packaging data whose recommended tag is the first recommended tag as target preferred packaging data; the target preferred packaging data has a corresponding target packaging material identifier;

[0173] S1202, obtaining the total quantity of the target packaging material identifications;

[0174] S1203, if the total quantity is less than a preset quantity threshold, determining a recommended package for the target item according to the second target used package data;

[0175] S1204: If the total quantity of the target preferred packages is greater than or equal to the quantity threshold, determine that the target packaging material is identified as the recommended package for the target item.

[0176] Specifically, after the server 200 analyzes and obtains the recommended labels of each packaging material in the preferred packaging data, it can filter out the packaging material with the first recommended label according to the recommended labels corresponding to each packaging material, as the target preferred packaging material, and the packaging material identification of the target preferred packaging material and its associated data are added together as the target preferred packaging data. If the number of types of the target preferred packaging material is greater than or equal to the preset quantity threshold, it means that the target preferred packaging material obtained by the current analysis meets the actual business needs, and the target packaging material identification can be used as the marking character of the recommended packaging of the target item. However, if the number of types of the target preferred packaging material is less than the preset quantity threshold, it means that the target preferred packaging material obtained by the current analysis does not meet the actual business needs, and it is necessary to combine the second target used packaging data corresponding to the second user identification of the target item again to analyze the pre-recommended packaging of the quantity threshold to be verified.

[0177] In this embodiment, by determining whether the packaging materials after label screening meet the quantity threshold, it is determined whether the current total quantity of packaging meets the recommendation requirements, so that subsequent operations can be continued after obtaining different judgment results, which can not only save labor costs, but also improve the efficiency and accuracy of packaging recommendations.

[0178] In one embodiment, the packaging recommendation method further includes the following steps:

[0179] S1301: If the target used packaging data meets a preset packaging recommendation trigger condition, determine a packaging material identifier in the target used packaging data as a recommended packaging for the target item.

[0180] Specifically, if the server 200 analyzes and determines that the packaging material identification, packaging usage count and packaging usage ratio included in the target used packaging data all meet the sub-conditions included in the packaging recommendation trigger condition, the packaging material identification included in the target used packaging data can be used as the recommended packaging for the target item, and the recommended packaging can be fed back to the terminal 100, which will display the recommended packaging.

[0181] In this embodiment, by determining whether the target used packaging data meets the preset packaging recommendation triggering condition, the recommended packaging of the target item is determined when the condition is met, thereby improving the efficiency and accuracy of packaging recommendation.

[0182] In order to enable those skilled in the art to fully understand the packaging recommendation solution proposed in this application, this application also provides an application scenario, which applies the above-mentioned packaging recommendation method. Specifically, the application of the packaging recommendation method in this application scenario will be combined with Figure 5 The following instructions are given:

[0183] like Figure 5 As shown, when the server 200 obtains the circulation data of the target item, it will associate the first user identifier (customer ID) and the second user identifier (delivery person ID) included in the circulation data of the target item based on the business data of a historical time window, and then combine the real-time circulation data and the historical data to analyze and determine whether the current number of packages that can be used as recommended packages meets the preset recommended number. The preset recommendation has been explained in the above embodiment as being determined by the terminal 100 interface. If it meets the preset number, the recommended package of the target item (set R) is output to the terminal 100. 1 ), if not, the trained packaging classifier - deep neural network classifier is used to further analyze and obtain the recommended packaging to be verified (set R 2 ), if the R 2 The number of meets the recommended number, then the recommended package output to the terminal 100 is R 2 If not, continue to analyze and determine the new recommended packaging (set R 3 ), if the R 3 The number of meets the recommended number, then the recommended package output to the terminal 100 is R 3 .

[0184] In this embodiment, the recommended available packaging is determined based on the associated use of historical data, which is more accurate than the traditional manual judgment method. At the same time, when historical data analysis alone cannot meet the packaging recommendation needs, it is proposed to use a packaging classifier to realize packaging classification prediction, which can not only save labor costs, but also improve the efficiency and accuracy of packaging recommendations.

[0185] In order to better implement the packaging recommendation method in the embodiment of the present application, based on the packaging recommendation method, the embodiment of the present application also provides a packaging recommendation device, such as Figure 6 As shown, the packaging recommendation device 600 includes:

[0186] A data acquisition module 602 is used to acquire the object circulation data of the target object, wherein the object circulation data includes the user identification of the sender or recipient of the target object;

[0187] A data determination module 604 is used to determine target used packaging data in pre-stored historical used packaging data according to the user identifier;

[0188] The label acquisition module 606 is used to classify and predict the preferred packaging data in the target used packaging data based on the trained packaging classifier to obtain a recommended label for the preferred packaging data if the target used packaging data does not meet the preset packaging recommendation triggering condition;

[0189] The packaging determination module 608 is used to determine the recommended packaging of the target item according to the recommended tag.

[0190] In some embodiments of the present application, the packaging recommendation device 600 also includes a historical data acquisition module for acquiring historical waybill data, wherein the historical waybill data includes a sender identification, a delivery person identification, and a packaging material identification; according to the sender identification, the packaging material identification is aggregated to obtain a first packaging material identification associated with the sender identification, and the number of packaging uses and the packaging use ratio corresponding to the first packaging material identification are obtained by statistics, and the results are combined as first historical used packaging data; according to the delivery person identification, the packaging material identification is aggregated to obtain a second packaging material identification associated with the delivery person identification, and the number of packaging uses and the packaging use ratio corresponding to the second packaging material identification are obtained by statistics, and the results are combined as second historical used packaging data; the first historical used packaging data and the second historical used packaging data are determined as the historical used packaging data.

[0191] In some embodiments of the present application, the user identifier includes a first user identifier and a second user identifier, the historical used packaging data includes a first historical used packaging data and a second historical used packaging data, and the data determination module 604 is further used to determine the sender identifier of at least one of the first historical used packaging data, and to determine the delivery personnel identifier of at least one of the second historical used packaging data; to obtain the first historical used packaging data corresponding to the target sender identifier that matches the first user identifier, and to obtain the first target used packaging data; to obtain the second historical used packaging data corresponding to the target delivery personnel identifier that matches the second user identifier, and to obtain the second target used packaging data; and to determine the first target used packaging data and the second target used packaging data as the target used packaging data.

[0192] In some embodiments of the present application, the target used packaging data includes first target used packaging data and second target used packaging data, and the label acquisition module 606 is also used to determine a preset packaging recommendation trigger condition, and the packaging recommendation trigger condition includes a first trigger condition, a second trigger condition and a third trigger condition; if the first target used packaging data does not meet any one of the first trigger condition, the second trigger condition and the third trigger condition, then obtain the preferred packaging data in the first target used packaging data and the second target used packaging data respectively; based on the trained packaging classifier, classify and predict the preferred packaging data to obtain a recommended label for the preferred packaging data.

[0193] In some embodiments of the present application, the label acquisition module 606 is also used to determine the packaging material identification in the first target used packaging data, and the packaging material identification has a corresponding packaging use count and packaging use ratio respectively; if the packaging use count is less than a preset count threshold, it is determined that the first target used packaging data does not meet the first trigger condition; if the packaging use ratio is less than a preset ratio threshold, it is determined that the first target used packaging data does not meet the second trigger condition; if the total number of packaging material identifications that meet the first trigger condition and the second trigger condition is less than a preset number threshold, it is determined that the first target used packaging data does not meet the third trigger condition; if the first target used packaging data does not meet any one of the first trigger condition, the second trigger condition and the third trigger condition, then the preferred packaging data in the first target used packaging data and the second target used packaging data are respectively obtained.

[0194] In some embodiments of the present application, the label acquisition module 606 is also used to, if the first target used packaging data does not meet any one of the first trigger condition, the second trigger condition and the third trigger condition, then, based on the number of times the packaging is used, arrange the packaging material identifiers in the first target used packaging data and the second target used packaging data in descending order to obtain a first packaging material identifier sequence and a second packaging material identifier sequence; obtain the first N packaging material identifiers in the first packaging material identifier sequence and the second packaging material identifier sequence to obtain a first target packaging material identifier and a second target packaging material identifier; N≥1; determine the first target packaging material identifier and the second target packaging material identifier as the preferred packaging data.

[0195] In some embodiments of the present application, the label acquisition module 606 is also used to determine the trained packaging classifier, which is trained by pre-stored historical waybill data; input the consignment information, the consignment weight data, and the preferred packaging data into the trained packaging classifier; obtain the classification result output by the trained packaging classifier to obtain a recommended label for the preferred packaging data; the recommended label is obtained by the trained packaging classifier based on the consignment information, the consignment weight data and the preferred packaging data.

[0196] In some embodiments of the present application, the label acquisition module 606 is also used to clean and simplify the consignment information, and perform vector conversion on the consignment weight data and the preferred packaging data; perform word segmentation processing on the simplified consignment information to obtain first input data, and perform vector merging on the converted consignment weight data and the preferred packaging data to obtain a second input vector; the second input vector and the first input data are used to input into the trained packaging classifier.

[0197] In some embodiments of the present application, the label acquisition module 606 is also used to obtain, through the input module, the first input data corresponding to the consignment information, and the second input vector corresponding to the consignment weight data and the preferred packaging data; through the data segmentation module, the first input data is serialized, and the first input data after serialization is subjected to sequence length control processing to obtain fixed sequence data; through the first convolution module, the fixed sequence data is vectorized, and the converted fixed sequence data is subjected to convolution processing to obtain a first convolution vector; through the vector splicing module, the first convolution vector and the second input vector are spliced ​​to obtain a spliced ​​vector; through the full link layer or the convolution link layer in the link processing module, the spliced ​​vector is linked to obtain a link vector; through the target activation function preset in the multi-label classification module, the link vector is matrix transformed to obtain a recommended label for the preferred packaging data; through the output module, the recommended label is output as the classification result.

[0198] In some embodiments of the present application, the label acquisition module 606 is also used to obtain historical waybill data; input the historical waybill data into a pre-established packaging classifier for model training, wherein the pre-established packaging classifier is preset with a target loss function; obtain the training result output by the pre-established packaging classifier according to the target loss function; if the training result meets the preset training end condition, then end the model training of the packaging classifier to obtain the trained packaging classifier.

[0199] In some embodiments of the present application, the target used packaging data includes first target used packaging data and second target used packaging data, the recommended tag includes a first recommended tag, and the packaging determination module 608 is further used to filter out the preferred packaging data whose recommended tag is the first recommended tag as the target preferred packaging data; the target preferred packaging data has a corresponding target packaging material identification; the total number of the target packaging material identifications is obtained; if the total number is less than a preset quantity threshold, the recommended packaging of the target item is determined according to the second target used packaging data; if the total number of the target preferred packaging is greater than or equal to the quantity threshold, the target packaging material identification is determined as the recommended packaging of the target item.

[0200] In some embodiments of the present application, the packaging recommendation device 600 also includes a packaging recommendation module, which is used to determine the packaging material identifier in the target used packaging data as the recommended packaging for the target item if the target used packaging data meets a preset packaging recommendation trigger condition.

[0201] In the above embodiment, the user ID of the consignor to whom the target item belongs is obtained by obtaining the item circulation data of the target item, so that the target used packaging data can be screened out from the historical used packaging data according to the user ID, and then, when the target used packaging data does not meet the packaging recommendation triggering condition, the preferred packaging data in the target used packaging data can be classified and predicted based on the trained packaging classifier, and finally the recommended packaging of the target item is determined based on the recommended tags of each preferred packaging data. The present solution determines the recommended available packaging based on the associated application of historical data, which is more accurate than the traditional manual judgment method. At the same time, when the packaging recommendation requirements cannot be met by the historical data analysis alone, it is proposed to use the packaging classifier to realize the classification prediction of packaging, which can not only save labor costs, but also improve the efficiency and accuracy of packaging recommendation.

[0202] In some embodiments of the present application, the packaging recommendation device 600 can be implemented in the form of a computer program. The computer program can be used in Figure 7 The computer device shown in the figure is run. The memory of the computer device can store various program modules constituting the packaging recommendation device 600, for example, Figure 6 The data acquisition module 602, data determination module 604, label acquisition module 606 and packaging determination module 608 are shown. The computer program composed of various program modules enables the processor to execute the steps of the packaging recommendation method of each embodiment of the present application described in this specification.

[0203] For example, Figure 7 The computer device shown can be Figure 6 The data acquisition module 602 in the packaging recommendation device 600 shown executes step S201. The computer device can execute step S202 through the data determination module 604. The computer device can execute step S203 through the label acquisition module 606. The computer device can execute step S204 through the packaging determination module 608. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external computer device through a network connection. When the computer program is executed by the processor, a packaging recommendation method is implemented.

[0204] Those skilled in the art will understand that Figure 7The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0205] In some embodiments of the present application, a computer device is provided, comprising one or more processors; a 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 in the steps of the above-mentioned packaging recommendation method. The steps of the packaging recommendation method here may be the steps of the packaging recommendation method in the above-mentioned embodiments.

[0206] 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 packaging recommendation method. The steps of the packaging recommendation method here can be the steps of the packaging recommendation method in each of the above-mentioned embodiments.

[0207] A recommended packaging method provided in an embodiment of the present application is introduced in detail above. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. 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 method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A packaging recommendation method, characterized in that: The method comprises: Acquire the object circulation data of the target object, wherein the object circulation data includes the user identification of the sender or recipient of the target object; According to the user identifier, target used packaging data is determined in pre-stored historical used packaging data; the target used packaging data includes first target used packaging data and second target used packaging data; Determining a preset packaging recommendation triggering condition, wherein the packaging recommendation triggering condition includes a first triggering condition, a second triggering condition, and a third triggering condition; Determine a packaging material identifier in the first target used packaging data, wherein the packaging material identifiers respectively have corresponding packaging usage times and packaging usage ratios; If the number of times the package is used is less than a preset number threshold, it is determined that the first target used package data does not meet the first trigger condition; If the packaging usage ratio is less than a preset ratio threshold, it is determined that the first target used packaging data does not meet the second trigger condition; If the total number of packaging material identifiers that meet the first trigger condition and the second trigger condition is less than a preset quantity threshold, it is determined that the first target used packaging data does not meet the third trigger condition; If the first target used packaging data does not satisfy any one of the first trigger condition, the second trigger condition and the third trigger condition, then based on the number of times the packaging is used, the packaging material identifiers in the first target used packaging data and the second target used packaging data are respectively arranged in descending order to obtain a first packaging material identifier sequence and a second packaging material identifier sequence; Obtain the first N packaging material identifications in the first packaging material identification sequence and the second packaging material identification sequence to obtain a first target packaging material identification and a second target packaging material identification; N≥1; determining the first target packaging material identifier and the second target packaging material identifier as the preferred packaging data; Based on the trained packaging classifier, classify and predict the preferred packaging data to obtain a recommended label for the preferred packaging data; Determine the recommended packaging of the target item according to the recommended label.

2. The packaging recommendation method according to claim 1, characterized in that: Before determining the target used packaging data from the pre-stored historical used packaging data according to the user identifier, the method further includes: Obtaining historical waybill data, wherein the historical waybill data includes a sender ID, a receiver ID, and a packaging material ID; According to the sender identification, the packaging material identifications are aggregated to obtain a first packaging material identification associated with the sender identification, and the number of times the packaging is used and the packaging usage ratio corresponding to the first packaging material identification are obtained by counting, and the results are combined as the first historical used packaging data; According to the delivery personnel identification, the packaging material identification is aggregated to obtain a second packaging material identification associated with the delivery personnel identification, and the number of times the packaging is used and the packaging usage ratio corresponding to the second packaging material identification are obtained by statistics, and the results are combined as the second historical used packaging data; The first historical used package data and the second historical used package data are determined as the historical used package data.

3. The packaging recommendation method according to claim 1, characterized in that: The user identifier includes a first user identifier and a second user identifier, the historical used packaging data includes a first historical used packaging data and a second historical used packaging data, and determining target used packaging data from pre-stored historical used packaging data according to the user identifier includes: Determine at least one sender identification of the first historical used packaging data, and determine at least one delivery person identification of the second historical used packaging data; Acquire first historical used packaging data corresponding to the target sender identifier that matches the first user identifier, to obtain first target used packaging data; Acquire the second historical used packaging data corresponding to the target delivery personnel identifier that matches the second user identifier, and obtain the second target used packaging data; The first target used packaging data and the second target used packaging data are determined as the target used packaging data.

4. The packaging recommendation method according to claim 1, characterized in that: The article circulation data includes the consignment information and consignment weight data of the target article, and the classification prediction of the preferred packaging data based on the trained packaging classifier to obtain the recommended label of the preferred packaging data includes: Determining the trained packaging classifier, where the trained packaging classifier is obtained by training with pre-stored historical waybill data; Inputting the consignment information, the consignment weight data, and the preferred packaging data into the trained packaging classifier; The classification result output by the trained packaging classifier is obtained to obtain a recommended label for the preferred packaging data; the recommended label is obtained by the trained packaging classifier according to the consignment information, the consignment weight data and the preferred packaging data.

5. The packaging recommendation method according to claim 4, characterized in that: Before inputting the consignment information, the consignment weight data, and the preferred packaging data into the trained packaging classifier, the method further includes: Cleaning and simplifying the consignment information, and performing vector conversion on the consignment weight data and the preferred packaging data; The simplified consignment information is segmented to obtain first input data, and the converted consignment weight data and preferred packaging data are vectorized to obtain a second input vector; the second input vector and the first input data are used to input into the trained packaging classifier.

6. The packaging recommendation method according to claim 4, characterized in that: The trained packaging classifier includes a data input module, a data segmentation module, a first convolution module, a vector concatenation module, a link processing module, a multi-label classification module and an output module. The step of obtaining the classification result output by the trained packaging classifier to obtain the recommended label of the preferred packaging data includes: Acquire, through the input module, first input data corresponding to the consignment information, and second input vectors corresponding to the consignment weight data and the preferred packaging data; The first input data is serialized by the data segmentation module, and the first input data after serialization is subjected to sequence length control to obtain fixed sequence data; The fixed sequence data is vector-converted by the first convolution module, and the converted fixed sequence data is convolved to obtain a first convolution vector; The first convolution vector and the second input vector are concatenated by the vector concatenation module to obtain a concatenated vector; Performing link processing on the concatenated vector through a full link layer or a convolution link layer in the link processing module to obtain a link vector; Performing matrix transformation processing on the link vector through a target activation function preset in the multi-label classification module to obtain a recommended label for the preferred packaging data; The recommended label is outputted as the classification result through the output module.

7. The packaging recommendation method according to claim 4, characterized in that: The step of determining the trained packaging classifier comprises: Get historical waybill data; Inputting the historical waybill data into a pre-established packaging classifier for model training, wherein the pre-established packaging classifier is preset with a target loss function; Obtaining a training result output by the pre-established packaging classifier according to the target loss function; If the training result meets the preset training end condition, the model training of the packaging classifier is ended to obtain the trained packaging classifier.

8. The packaging recommendation method according to claim 1, characterized in that: The recommended tags include a first recommended tag, and determining the recommended packaging of the target item according to the recommended tags includes: Filter out the preferred packaging data whose recommended tag is the first recommended tag as the target preferred packaging data; the target preferred packaging data has a corresponding target packaging material identifier; Obtaining the total quantity of the target packaging material identification; If the total quantity is less than a preset quantity threshold, determining a recommended package of the target item according to the second target used package data; If the total quantity of the target preferred packages is greater than or equal to the quantity threshold, the target packaging material is identified as the recommended package for the target item.

9. The packaging recommendation method according to claim 1, characterized in that: The method further comprises: If the first target used packaging data meets a preset packaging recommendation trigger condition, a packaging material identifier in the target used packaging data is determined as a recommended packaging for the target item.

10. A packaging recommendation device, characterized in that: The device comprises: A data acquisition module, used to acquire the object circulation data of the target object, wherein the object circulation data includes the user identification of the sender or receiver of the target object; A data determination module, configured to determine target used packaging data from pre-stored historical used packaging data according to the user identifier; the target used packaging data includes first target used packaging data and second target used packaging data; The tag acquisition module is used to determine a preset packaging recommendation trigger condition, wherein the packaging recommendation trigger condition includes a first trigger condition, a second trigger condition, and a third trigger condition; determine a packaging material identifier in the packaging data used by the first target, wherein the packaging material identifier has a corresponding number of packaging uses and a packaging use ratio; if the number of packaging uses is less than a preset number threshold, it is determined that the packaging data used by the first target does not meet the first trigger condition; if the packaging use ratio is less than a preset ratio threshold, it is determined that the packaging data used by the first target does not meet the second trigger condition; if the total number of packaging material identifiers that meet the first trigger condition and the second trigger condition is less than a preset number threshold, it is determined that the packaging data used by the first target does not meet the third trigger condition; if the first If the target used packaging data does not meet any one of the first trigger condition, the second trigger condition and the third trigger condition, then based on the number of times the packaging is used, the packaging material identifiers in the first target used packaging data and the second target used packaging data are respectively arranged in descending order to obtain a first packaging material identifier sequence and a second packaging material identifier sequence; the first N packaging material identifiers in the first packaging material identifier sequence and the second packaging material identifier sequence are obtained to obtain a first target packaging material identifier and a second target packaging material identifier; N≥1; the first target packaging material identifier and the second target packaging material identifier are determined as the preferred packaging data; based on the trained packaging classifier, the preferred packaging data are classified and predicted to obtain a recommended label for the preferred packaging data; The packaging determination module is used to determine the recommended packaging of the target item according to the recommended tag.

11. 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 the packaging recommendation method according to any one of claims 1 to 9.

12. 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 steps in the packaging recommendation method according to any one of claims 1 to 9.

Images