Cargo source recommendation method and device, and electronic device

Abstract

The application provides a cargo source recommendation method and device and electronic equipment, and belongs to the technical field of logistics transportation. In the method, the real-time search request of a target driver is received, and a first preset number of first historical click behavior data corresponding to the target driver is obtained. Each first historical click behavior data and the real-time search request are subjected to generalization processing to calculate a total generalization score, and the first historical click behavior data is filtered according to a preset generalization score to obtain second historical click behavior data with higher relevance to the real-time search request and determine a second behavior sequence. Finally, the second behavior sequence is input into a target recommendation model to generate a recommendation result sorted according to the relevance. The method provided by the application can effectively utilize the second historical click behavior data with a longer coverage range and higher relevance to the real-time search request to model the long-term interest and periodicity of the driver under the condition that the online inference performance is limited, thereby effectively improving the accuracy of the cargo source recommendation result.

Description

Technical Field

[0001] This application relates to the field of logistics and transportation technology, and in particular to a method, apparatus and electronic equipment for recommending cargo sources. Background Technology

[0002] With the development of internet and mobile communication technologies, online freight platforms have gradually become an important channel for drivers and shippers to find freight and match transportation. Through these platforms, shippers can efficiently publish freight information, while drivers can conveniently obtain transportation tasks, thereby improving transportation efficiency and optimizing resource allocation. During this process, drivers' clicks, phone calls, and transactions on online freight platforms generate a large amount of analyzable behavioral data. How to effectively utilize this driver behavioral data to recommend freight and improve the accuracy of freight recommendations has become a key focus for the logistics and transportation industry.

[0003] In existing technologies, driver behavior sequences are used as input to model recommendations, and the model's output is used to recommend freight to drivers. When modeling based on driver behavior sequences, statistical Markov chain methods, Gated Recurrent Unit (GRU4Rec) models for session-based recommendation, and attention-based neural network models can be employed to capture drivers' interests and behavioral patterns. However, in online freight scenarios, drivers' freight-finding behavior generally exhibits periodic characteristics. Short-distance drivers typically have a freight-finding cycle of 1 to 2 days, while long-distance drivers may have a cycle as long as a week or even half a month. Limited by online inference performance and computational costs, existing online freight platforms typically only use the driver's most recent 20 clicks as model input. These 20 clicks only cover a short freight-finding cycle, which differs significantly from the driver's actual freight-finding cycle. Because the model's input sequence is short, it can only reflect short-term driver behavior characteristics, making it difficult to accurately model drivers' stable periodic patterns and long-term interests. This results in poor accuracy in freight recommendation results, failing to fully meet the actual needs of drivers.

[0004] Therefore, in freight recommendation methods, the poor accuracy of freight recommendation results is an urgent problem to be solved due to the high overhead of processing long behavioral sequences of drivers and the difficulty in modeling long-term interests. Summary of the Invention

[0005] This application provides a method, apparatus, and electronic device for recommending goods, in order to solve the problem of poor accuracy of goods recommendation results in existing goods recommendation methods.

[0006] Firstly, this application provides a method for recommending sources of goods, the method comprising:

[0007] Receive a real-time search request input by a target driver on an online freight platform; wherein the real-time search request includes a first departure point and a first destination selected by the target driver; or, the real-time search request includes the first departure point and the first destination, as well as at least one first vehicle length, at least one first vehicle type, and / or a target tonnage range selected by the target driver;

[0008] Obtain the first behavior sequence corresponding to the target driver. The first behavior sequence includes a first preset number of first historical click behavior data. Each first historical click behavior data includes the second departure point, second destination, second vehicle length, second vehicle type, and cargo tonnage corresponding to the first cargo information. The first cargo information is the cargo information clicked by the target driver within a first preset historical time period.

[0009] Based on the real-time search request, each of the first historical click behavior data is generalized to obtain the total generalization score corresponding to each of the first historical click behavior data.

[0010] Based on the preset generalization score and the total generalization score corresponding to each of the first historical click behavior data, the first preset number of first historical click behavior data are filtered to obtain at least one second historical click behavior data.

[0011] A second behavior sequence is determined based on the at least one second historical click behavior data, or the at least one second historical click behavior data and the first behavior sequence;

[0012] The second behavior sequence and the second cargo information, as well as the historical and real-time behavior statistics of the target driver, are processed using a target recommendation model to obtain the recommendation result corresponding to the real-time search request. The recommendation result includes multiple third cargo information, which are arranged in descending order of relevance between each third cargo information and the real-time search request. The second cargo information is the cargo information that has not yet been accepted on the online freight platform.

[0013] In one possible design, the generalization process performed on each of the first historical click behavior data based on the real-time search request to obtain a total generalization score corresponding to each of the first historical click behavior data includes:

[0014] For each of the first historical click behavior data, a first generalization score is determined based on the second origin and the first origin;

[0015] Determine the second generalization score based on the second destination and the first destination;

[0016] The third generalization score is determined based on the second vehicle length and the at least one first vehicle length;

[0017] A fourth generalization score is determined based on the second vehicle model and the at least one first vehicle model;

[0018] The fifth generalization score is determined based on the source weight and the target weight range.

[0019] The sum of the first generalization score, the second generalization score, the third generalization score, the fourth generalization score, and the fifth generalization score is determined as the total generalization score corresponding to the first historical click behavior data.

[0020] In one possible design, the method further includes:

[0021] If the first destination and the second destination are the same, the target generalization score is determined to be zero; wherein the first destination is the first departure point or the first destination, the second destination is the second departure point or the second destination, and the target generalization score is the first generalization score or the second generalization score; if the first destination is the first departure point, the second destination is the second departure point, and the target generalization score is the first generalization score; if the first destination is the first destination, the second destination is the second destination, and the target generalization score is the second generalization score.

[0022] When the first target location and the second target location are inconsistent, and the first target location is the detailed location selected by the target driver, the target generalization score is determined to be the ratio of the target empty driving distance to the segmented fixed generalization distance; wherein, the target empty driving distance is the straight-line distance between the second target location and the first target location, and the segmented fixed generalization distance is related to the target empty driving distance;

[0023] If the first destination and the second destination are different, and the first destination is a prefecture-level administrative region selected by the target driver, the target generalization score is determined to be the ratio of the target empty driving distance to the first distance; wherein, the first distance is the sum of a fixed distance and a preset proportion of the transportation distance, and the transportation distance is the distance between the second departure point and the second destination.

[0024] In one possible design, determining the third generalization score based on the second vehicle length and the at least one first vehicle length includes:

[0025] If the second vehicle length meets the first preset condition, the third generalization score is determined to be zero;

[0026] The first preset condition includes any one of the following:

[0027] The real-time search request includes the first vehicle length, and the second vehicle length exists among the at least one first vehicle length;

[0028] The real-time search request does not include the first vehicle length, and the second vehicle length is present in at least one of the behavioral vehicle lengths corresponding to the target driver; the behavioral vehicle length corresponding to the target driver is related to the third vehicle length corresponding to the third cargo information and the fourth vehicle length corresponding to the fourth cargo information; the third cargo information is the cargo information that the target driver accepted orders for within a second preset historical time period; the fourth cargo information is the cargo information that the target driver clicked on within the second preset historical time period.

[0029] If the real-time search request does not include the first vehicle length, and the second vehicle length is not among the at least one behavior vehicle length, the third generalization score is determined to be a first preset value;

[0030] If the real-time search request includes the first vehicle length, the second vehicle length is not present in the at least one first vehicle length, and the second vehicle length is present in the at least one behavioral vehicle length, then the third generalization score is determined to be the second preset value.

[0031] If the real-time search request includes the first vehicle length, and the second vehicle length does not exist among the at least one first vehicle length and the at least one behavior vehicle length, determine whether the second vehicle length is less than the target first vehicle length, wherein the target first vehicle length is the smallest first vehicle length among the at least one first vehicle length;

[0032] If the second vehicle length is less than the target first vehicle length, the third generalization score is determined to be a third preset value;

[0033] If the second vehicle length is greater than or equal to the target first vehicle length, the third generalization score is determined to be a fourth preset value; wherein the fourth preset value is greater than the third preset value.

[0034] In one possible design, determining the fourth generalization score based on the second vehicle model and the at least one first vehicle model includes:

[0035] If the second vehicle model meets the second preset condition, the fourth generalization score is determined to be zero;

[0036] The second preset condition includes any one of the following:

[0037] The real-time search request includes the first vehicle model, and the second vehicle model exists among the at least one first vehicle model;

[0038] The real-time search request includes the first vehicle model, the second vehicle model not existing in at least one of the first vehicle models, and the second vehicle model existing in at least one of the behavior vehicle models corresponding to the target driver;

[0039] The real-time search request does not include the first vehicle type, and the second vehicle type exists among the at least one behavioral vehicle type; wherein, the behavioral vehicle type corresponding to the target driver is related to the third vehicle type corresponding to the third cargo source information and the fourth vehicle type corresponding to the fourth cargo source information;

[0040] If the real-time search request includes the first vehicle model, and the second vehicle model is not present in either the at least one first vehicle model or the at least one behavioral vehicle model, then the fourth generalization score is determined to be the fifth preset value.

[0041] If the real-time search request does not include the first vehicle model, and the second vehicle model is not among the at least one behavioral vehicle model, the fourth generalization score is determined to be the sixth preset value.

[0042] In one possible design, determining the fifth generalization score based on the source tonnage and the target tonnage range includes:

[0043] If the real-time search request includes the target tonnage range and the tonnage of the source cargo is within the target tonnage range, the fifth generalization score is determined to be zero.

[0044] If the real-time search request does not include the target tonnage range, the fifth generalization score is determined to be the sum of a fixed value and a first ratio, where the first ratio is the ratio of the tonnage of the cargo source to the approved load capacity of the target driver's vehicle.

[0045] If the real-time search request includes the target tonnage range and the source tonnage is outside the target tonnage range, the fifth generalization score is determined to be the sum of the fixed value and the second ratio, where the second ratio is the ratio of the source tonnage to the first tonnage; if the approved load is greater than or equal to the second tonnage, the first tonnage is the approved load; if the approved load is less than the second tonnage, the first tonnage is the second tonnage; the second tonnage is the maximum tonnage corresponding to the target tonnage range.

[0046] In one possible design, the filtering of the first preset number of first historical click behavior data based on a preset generalization score and the total generalization score corresponding to each first historical click behavior data to obtain at least one second historical click behavior data includes:

[0047] For each first historical click behavior data, if the total generalization score corresponding to the first historical click behavior data is less than the preset generalization score, the first historical click behavior data is determined as the second historical click behavior data.

[0048] In one possible design, the first historical click behavior data also includes the click time when the target driver clicked on the first cargo information;

[0049] Determining the second behavior sequence based on the at least one second historical click behavior data, or the at least one second historical click behavior data and the first behavior sequence, includes:

[0050] If the number of at least one second historical click behavior data is greater than or equal to a second preset number, based on the click time corresponding to the first historical click behavior data, the second preset number of second historical click behavior data whose click time is closest to the current time is determined as the second behavior sequence; wherein, the second preset number is less than the first preset number;

[0051] If the number of at least one second historical click behavior data is less than the second preset number, a third historical click behavior data is determined based on the first behavior sequence and the at least one second historical click behavior data; wherein, the third historical click behavior data is the first historical click behavior data in the first behavior sequence other than the at least one second historical click behavior data;

[0052] If the number of third historical click behavior data is greater than the target number, based on the total generalization score corresponding to each third historical click behavior data, the target number of third historical click behavior data are selected as fourth historical click behavior data in ascending order of total generalization score, and the second historical click behavior data and the fourth historical click behavior data are determined as the second behavior sequence; wherein, the target number is the difference between the second preset number and the number of at least one second historical click behavior data;

[0053] If the number of third historical click behavior data is less than or equal to the target number, the second historical click behavior data and the third historical click behavior data are determined as the second behavior sequence.

[0054] Secondly, this application provides a source recommendation device, the device including modules for performing the method described in the first aspect or various possible designs of the first aspect.

[0055] Thirdly, this application provides an electronic device, including: a memory and at least one processor;

[0056] The memory stores computer-executed instructions;

[0057] The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method described in the first aspect or various possible designs of the first aspect.

[0058] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed, implement the method described in the first aspect or various possible designs of the first aspect.

[0059] Fifthly, this application provides a computer program product, which includes computer program code that, when run on a computer, causes the computer to implement the method described in the first aspect or various possible designs of the first aspect.

[0060] In a sixth aspect, this application provides a chip, comprising: an interface circuit and a logic circuit, wherein the interface circuit is configured to receive signals from other chips outside the chip and transmit them to the logic circuit, or to send signals from the logic circuit to other chips outside the chip, and the logic circuit is configured to implement the method described in the first aspect or various possible designs of the first aspect.

[0061] This application provides a freight recommendation method, apparatus, and electronic device. In this method, a first preset number of historical click behavior data corresponding to the target driver are obtained by receiving a real-time search request from the target driver, and a first behavior sequence is generated. Based on this, each piece of first historical click behavior data is generalized to the real-time search request, and a total generalization score is calculated. Then, the first historical click behavior data is filtered according to the preset generalization score to obtain second historical click behavior data with higher relevance to the real-time search request. A second behavior sequence is determined based on the second historical click behavior data. Finally, the second behavior sequence, along with second freight information, the target driver's historical behavior statistics, and real-time behavior statistics, are input into a target recommendation model to generate recommendation results sorted by relevance. The freight recommendation method provided by this application can effectively utilize second historical click behavior data, which has a longer coverage and stronger relevance to real-time requests, to model the long-term interests and periodic patterns of drivers, even when online inference performance is limited. This solves the problem in existing technologies where relying only on a small number of recent click behaviors leads to insufficient accuracy in freight recommendations and difficulty in meeting the actual needs of drivers. Attached Figure Description

[0062] Figure 1 A flowchart illustrating a method for recommending sources of goods provided in an embodiment of this application;

[0063] Figure 2 A flowchart illustrating another method for recommending sources of goods provided in this application embodiment;

[0064] Figure 3 A flowchart illustrating another method for recommending sources of goods provided in this application embodiment;

[0065] Figure 4 A flowchart illustrating another method for recommending sources of goods provided in this application embodiment;

[0066] Figure 5 A flowchart illustrating another method for recommending sources of goods provided in this application embodiment;

[0067] Figure 6 A flowchart illustrating another method for recommending goods provided in this application embodiment;

[0068] Figure 7 This is a schematic diagram of the structure of a source recommendation device provided in an embodiment of this application;

[0069] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0071] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims and drawings of this application are intended to cover non-exclusive inclusion.

[0072] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of the phrase "embodiment" in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0073] In this article, the term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists, A and B can exist simultaneously, and B exists. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0074] Furthermore, the terms "first," "second," etc., in the specification and claims of this application or in the aforementioned drawings are used to distinguish different objects rather than to describe a specific order, and may explicitly or implicitly include one or more of the features.

[0075] In the description of this application, unless otherwise stated, "multiple" and "at least two" mean two or more (including two), and similarly, "multiple groups" and "at least two groups" mean two or more (including two groups).

[0076] In the description of this application, it should be noted that, unless otherwise explicitly specified and limited, the terms "connected" and "linked" should be interpreted broadly. For example, "connected" or "linked" can refer not only to a physical connection, but also to an electrical connection or a signal connection. For instance, it can be a direct connection, i.e., a physical connection, or an indirect connection through at least one intermediate component, as long as the circuit is connected. It can also refer to the internal connection between two components. A signal connection can refer not only to a signal connection through a circuit, but also to a signal connection through a medium, such as radio waves. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0077] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, different technical features in this application can be combined with each other.

[0078] Millions of drivers are active on online freight platforms every day, generating a large amount of behavioral data, such as each driver's clicks, phone calls, and transactions regarding freight orders. This behavioral data provides an important data foundation for online freight platforms to make targeted freight recommendations to drivers in the future.

[0079] Modeling and utilizing user behavior sequences is a key area of ​​research in recommender algorithms. Early research primarily employed statistical Markov chain methods, followed by the development of the GRU4Rec model based on recurrent neural networks, and in recent years, attention-based neural network models have emerged. The evolution of these technologies generally shows a trend from capturing short-term interests to mining long-term interests, from coarse-grained to fine-grained processing, and from serial processing to parallel computing. Modeling long-term interests typically requires a large time window or even the entire historical behavior sequence, with the number of behaviors potentially reaching tens of thousands or even millions, resulting in high computational costs and limited online inference performance.

[0080] In logistics and transportation scenarios, drivers' freight-finding behavior exhibits a clear cyclical pattern due to their common need to return to their place of residence and the fact that some drivers have fixed offline orders. Short-distance drivers typically have a freight-finding cycle of 1 to 2 days, while long-distance drivers may have a cycle as long as a week or even half a month. Therefore, incorporating driver behavioral sequence information is crucial in the driver-freight matching process of online freight platforms. However, limited by current online inference performance, the sequence features input to the model in actual business systems are typically only the driver's most recent 20 click records. Analysis shows that this range only covers a freight-finding cycle of about 0.35 days, which differs significantly from the driver's actual freight-finding cycle. Therefore, because the model's input sequence is short, it can only reflect the driver's short-term behavioral characteristics, making it difficult to accurately model the driver's stable cyclical patterns and long-term interests. This results in poor accuracy of freight recommendation results, failing to fully meet the actual needs of drivers.

[0081] In summary, in freight recommendation methods, the poor accuracy of freight recommendation results is a problem that urgently needs to be solved due to the high overhead of processing long behavioral sequences of drivers and the difficulty in modeling long-term interests.

[0082] Next, through some specific embodiments and accompanying drawings, this application will describe in detail how it solves the problem that the accuracy of the cargo recommendation results is poor due to the large processing overhead of long behavior sequences of drivers and the difficulty in modeling long-term interests in the above-mentioned cargo recommendation method.

[0083] Figure 1 This is a flowchart illustrating a method for recommending sources of goods provided in an embodiment of this application. Figure 1 As shown, the source recommendation method provided in this application embodiment specifically includes S101 to S106, and S101 to S106 will be described in detail below.

[0084] It should be noted that the execution subject of the supply recommendation method provided in this application embodiment can be an electronic device. The electronic device can be a mobile phone, a laptop computer, etc., and this embodiment does not specifically limit it.

[0085] S101. Receive the real-time search request entered by the target driver on the online freight platform.

[0086] It should be noted that target drivers can input the corresponding filter criteria for their real-time search request through interactive controls on the client or web interface of the online freight platform.

[0087] Specifically, target drivers can directly enter the filter criteria corresponding to their real-time search request into the search box on the client or web interface of the online freight platform.

[0088] Target drivers can also select the corresponding filter criteria for their real-time search request through the selection box on the client or web interface of the online freight platform.

[0089] Target drivers can also directly select historically saved filter criteria or frequently used routes on the online freight platform, and the platform will automatically generate corresponding real-time search requests.

[0090] In one possible embodiment, the real-time search request includes the first departure point and the first destination selected by the target driver.

[0091] When the target driver only wants to find transportation tasks in the general direction and does not care about additional constraints such as vehicle parameters and tonnage, the target driver only needs to enter the first departure point and the first destination he expects when entering a real-time search request.

[0092] In this embodiment, the filtering conditions input by the target driver are relatively simple, mainly used to match cargo information between the target driver's departure location and destination location.

[0093] In another possible embodiment, the real-time search request includes a first origin and a first destination, as well as at least one first vehicle length, at least one first vehicle type, and / or a target tonnage range selected by the target driver.

[0094] It should be noted that implementing a search request may include a first departure point, a first destination, and at least one first vehicle length; it may also include a first departure point, a first destination, at least one first vehicle length, and at least one first vehicle type; it may also include a first departure point, a first destination, at least one first vehicle length, at least one first vehicle type, and a target tonnage range; it may also include a first departure point, a first destination, and at least one first vehicle type; it may also include a first departure point, a first destination, at least one first vehicle type, and a target tonnage range; it may also include a first departure point, a first destination, and a target tonnage range. This embodiment does not specifically limit these aspects.

[0095] In this embodiment, the screening criteria input by the target driver are more refined, which can more accurately constrain the recall range of the cargo source.

[0096] S102. Obtain the first behavior sequence corresponding to the target driver. The first behavior sequence includes a first preset number of first historical click behavior data. Each first historical click behavior data includes the second departure point, second destination, second vehicle length, second vehicle type and cargo tonnage corresponding to the first cargo information.

[0097] The first source of goods information refers to the source of goods that the target driver clicked on within the first preset historical time period.

[0098] The first preset historical time period can be set by the platform administrator according to actual needs, and this embodiment does not impose specific limitations on it. For example, the first preset historical time period is the most recent 30 days.

[0099] It should be noted that when target drivers browse the published freight information on the online freight platform, they will click on the freight information that interests them to view the details of the freight information. The freight information that the target driver clicks on is the first freight information.

[0100] By constructing a first behavior sequence using the first source of goods information within a first preset historical time period, it is possible to extract the real records of the interaction between the target driver and the source of goods information within the limited first preset historical time period, avoiding interference from invalid or noisy data; at the same time, it can ensure that the constructed first behavior sequence covers the target driver's main interests and freight-finding habits within the first preset historical time period.

[0101] The first behavior sequence consists of the first historical click behavior data related to the target driver.

[0102] It should be noted that the online freight platform extracts M cargo information from the target driver's click logs within the first preset historical time period, in chronological order from most recent to oldest. The platform then parses and processes each of the M cargo information to obtain M first historical click behavior data, forming a first behavior sequence. At this point, the length of the first behavior sequence is M.

[0103] Where M represents the first preset quantity, and M is a positive integer. M can be set by the platform administrator according to actual needs, and this embodiment does not impose specific limitations on it.

[0104] For example, analysis of driver behavior logs from online freight platforms reveals that, taking driver click behavior as an example, a behavior sequence of length 20, ordered from most recent to oldest, only covers 0.35 days of freight search time. As the length of the behavior sequence increases, the coverage gradually improves; a behavior sequence of length 150 covers 4.1 days of freight search time, while a behavior sequence of length 200 covers 5.3 days. Considering that most drivers on online freight platforms have a round-trip transportation cycle of less than 2 days, the time period covered by a click behavior sequence of length 150 is sufficient to cover at least two transportation cycles for most drivers, while maintaining reasonable online computing overhead. Therefore, the first preset quantity M is set to 150, meaning the first behavior sequence includes the 150 historical click behavior data of the target driver most recently requested in the real-time search request.

[0105] It should be noted that if the number of first source information within the first preset historical time period is less than the first preset quantity, the first behavior sequence is generated based on the first historical click behavior data corresponding to all first source information within the first preset historical time period.

[0106] For any first source of cargo information, the second vehicle length corresponding to the first source of cargo information is the vehicle length required in the first source of cargo information; the second vehicle type corresponding to the first source of cargo information is the vehicle type required in the first source of cargo information; and the cargo tonnage corresponding to the first source of cargo information is the cargo weight recorded in the first source of cargo information.

[0107] In this embodiment, because the exposure behavior data corresponding to the target driver is noisy and the transaction behavior data corresponding to the target driver is too sparse, it is not friendly to newly registered drivers on the online freight platform. Therefore, the first behavior sequence includes the target driver's historical click behavior data, rather than the target driver's corresponding exposure behavior data or transaction behavior data. Generating the first behavior sequence using the target driver's historical click behavior data can reflect the target driver's true interest and preference for freight sources while ensuring sufficient data volume, thereby providing a reliable data foundation for subsequent generalization processing and avoiding modeling bias caused by interference from exposure data or insufficient transaction data.

[0108] S103. Based on the real-time search request, perform generalization processing on each first historical click behavior data to obtain the total generalization score corresponding to each first historical click behavior data.

[0109] It should be noted that the electronic device extracts the conditions such as the first departure point, first destination, at least one first vehicle length, at least one first vehicle type, and target tonnage range contained in the real-time search request, and applies preset rules to compare the extracted first departure point, first destination, at least one first vehicle length, at least one first vehicle type, and target tonnage range with the second departure point, second destination, second vehicle length, second vehicle type, and cargo tonnage in each first historical click behavior data. The comparison results are quantified into a numerical form (total generalization score), which is used to represent the correlation between the first historical click behavior data and the real-time search request.

[0110] S104. Based on the preset generalization score and the total generalization score corresponding to each first historical click behavior data, filter the first preset number of first historical click behavior data to obtain at least one second historical click behavior data.

[0111] The preset generalization score can be set by the platform administrator according to actual needs, and this embodiment does not impose specific limitations on it. For example, the preset generalization score is 7 points.

[0112] It should be noted that the number of second historical click behavior data is N, where N is a positive integer and 1 < N ≤ M.

[0113] In one possible embodiment, the method steps shown in S104 can be implemented by Sa, which will be described in detail below.

[0114] Sa, for each first historical click behavior data, if the total generalization score corresponding to the first historical click behavior data is less than the preset generalization score, the first historical click behavior data is determined as the second historical click behavior data.

[0115] The total generalization score corresponding to the first historical click behavior data is used to measure the degree of difference between the first historical click behavior data and the real-time search request. The larger the total generalization score corresponding to the first historical click behavior data, the greater the difference between the first source information corresponding to the first historical click behavior data and the real-time search request; the smaller the total generalization score corresponding to the first historical click behavior data, the closer the first source information corresponding to the first historical click behavior data is to the real-time search request, and the higher the relevance.

[0116] It should be noted that the second historical click behavior data is the first historical click behavior data whose total generalization score is less than the preset generalization score. By selecting the first historical click behavior data with a total generalization score less than the preset generalization score as the second historical click behavior data, it is ensured that the second historical click behavior data in the second behavior sequence has a high correlation with the real-time search request. This avoids historical click behavior data that differs significantly from the real-time search request from entering the subsequent processing flow, making the features of the second behavior sequence input into the target recommendation model more concentrated and accurate. This reduces the interference of invalid data on the reasoning of the target recommendation model and ensures that the limited sequence length can carry more information closely related to the current needs of the target driver, thereby improving the stability and accuracy of the target recommendation model in generating recommendation results.

[0117] S105. Determine the second behavior sequence based on at least one second historical click behavior data, or at least one second historical click behavior data and the first behavior sequence.

[0118] In one possible embodiment, the method steps shown in S105 can be implemented by S1051 to S1056, which are described in detail below.

[0119] S1051. Determine whether the number of second historical click behavior data is greater than or equal to the second preset number.

[0120] The second preset quantity is P, where P is a positive integer and P > 10. P can be set by the platform administrator according to actual needs, and this embodiment does not impose a specific limitation on it. For example, P is 20.

[0121] It should be noted that when N≥P, the method steps shown in S1052 are executed; when N<P, the method steps shown in S1053 are executed.

[0122] In this embodiment, the second preset quantity is the target length parameter of the input sequence of the target recommendation model configured by the system, which is used to limit the upper limit of the length of the second row sequence.

[0123] S1052. When the number of at least one second historical click behavior data is greater than or equal to a second preset number, based on the click time corresponding to the first historical click behavior data, the second preset number of second historical click behavior data whose click time is closest to the current moment are determined as the second behavior sequence.

[0124] The second preset quantity is less than the first preset quantity, i.e., P < M.

[0125] It should be noted that each first historical click behavior data also includes the click time of the first cargo information corresponding to the first historical click behavior data of the target driver.

[0126] When N≥P, the electronic device sorts the N second historical click behavior data based on the click time corresponding to each second historical click behavior data, and selects the first P second historical click behavior data from the closest to the farthest according to the proximity of the click time to the current time. The set of the selected first P second historical click behavior data is determined as the second behavior sequence. The unselected (NP) second historical click behavior data are not included in the second behavior sequence.

[0127] S1053. If the number of at least one second historical click behavior data is less than the second preset number, determine the third historical click behavior data based on the first behavior sequence and at least one second historical click behavior data.

[0128] The third historical click behavior data is the first historical click behavior data in the first behavior sequence, excluding at least one second historical click behavior data.

[0129] It should be noted that when N < P, the M first historical click data points include N second historical click data points and Q third historical click data points, where M = N + Q. Q is a positive integer greater than or equal to 1.

[0130] S1054. Determine whether the number of third historical click behavior data is greater than the target number.

[0131] The target quantity is the difference between the second preset quantity and the quantity of at least one second historical click behavior data. The target quantity is H, where H = PN.

[0132] It should be noted that when Q > H, the method steps shown in S1055 are executed; when Q ≤ H, the method steps shown in S1056 are executed.

[0133] S1055. When the number of third historical click behavior data is greater than the target number, based on the total generalization score corresponding to each third historical click behavior data, select the target number of third historical click behavior data as the fourth historical click behavior data in order of increasing total generalization score, and determine the second historical click behavior data and the fourth historical click behavior data as the second behavior sequence.

[0134] It should be noted that when Q > H, the Q third historical click behavior data are sorted in ascending order of total generalization score, and the top H third historical click behavior data are determined as the fourth historical behavior data. Then, the set of N second historical click behavior data and H fourth historical behavior data is determined as the second behavior sequence.

[0135] S1056. If the number of third historical click behavior data is less than or equal to the target number, the second historical click behavior data and the third historical click behavior data are determined as the second behavior sequence.

[0136] It should be noted that when Q≤H, the set of N second historical click behavior data and H third historical click behavior data is determined as the second behavior sequence, that is, the second behavior sequence is the first behavior sequence.

[0137] In the second row sequence, all historical click data are arranged in order of the click time from most recent to oldest.

[0138] In this embodiment, the quantity of second historical click behavior data is judged, and when the quantity of second historical click behavior data is insufficient, a second behavior sequence is generated by supplementing the first behavior sequence with fourth or third historical click behavior data that has high relevance. When the quantity of second historical click behavior data is too large, a second behavior sequence is generated based on second historical click behavior data that is closer to the current time. This ensures that a second behavior sequence that meets the preset length requirement and takes into account both timeliness and relevance can always be output, thereby ensuring the data integrity and consistency of the input target recommendation model.

[0139] S106. The target recommendation model is used to process the second behavior sequence and the second cargo information, as well as the historical behavior statistics and real-time behavior statistics of the target driver, to obtain the recommendation results corresponding to the real-time search request. The recommendation results include multiple third cargo information, which are arranged in descending order of relevance between each third cargo information and the real-time search request.

[0140] The second source of cargo information refers to cargo information that has not yet been accepted for orders on the online freight platform. There are multiple second source of cargo information, and each second source of cargo information includes information such as the third origin, the third destination, the expected delivery time, the expected vehicle length, the expected vehicle type, and the number of times the cargo has been viewed since it was posted.

[0141] Among them, the target driver's historical behavior statistics include the number of times the target driver clicked on cargo information with the origin as the first departure point and the destination as the first destination in the past 7 days.

[0142] Among them, the real-time behavior statistics of the target driver are overall statistical data generated based on the target driver's operation behavior in a recent period of time, which is used to reflect the target driver's immediate freight search preferences and operating habits.

[0143] It should be noted that the real-time behavior statistics of the target driver include the first time elapsed since the target driver last clicked on a secondary freight category, the second time elapsed since the target driver last clicked on freight information with the same origin as the third origin, the third time elapsed since the target driver last clicked on freight information with the same origin as the third origin and the same destination as the third destination, the number of times the target driver clicked on freight information with the same vehicle length as the desired delivery vehicle in the last 30 minutes, the number of times the target driver clicked on freight information with the same vehicle type as the desired delivery vehicle in the last 5 minutes, and the number of times freight information with the same origin as the third origin and the same destination as the third origin was exposed to the target driver by the online freight platform in the last 30 minutes.

[0144] Specifically, secondary cargo categories refer to the second-level classifications used on online freight platforms when cargo information is categorized according to its nature. Online freight platforms typically set up multi-level categories for cargo information to facilitate management and retrieval. For example, the primary cargo category might be "parts," the secondary category might be further refined to "regular parts," the tertiary category could be "general parts," and the quaternary category could be "auto parts." Through this hierarchical classification system, secondary cargo categories can provide a more refined division based on primary categories, accurately describing the type and characteristics of cargo, thereby reflecting drivers' preferences in specific cargo categories.

[0145] The third departure point involved in the real-time behavior statistics of the target driver is any third departure point included in any of the multiple second cargo source information.

[0146] In this embodiment, the real-time behavioral statistics of the target driver can reflect the target driver's immediate cargo-finding preferences and operational characteristics within a shorter time scale, providing the target recommendation model with input data that is more closely aligned with real-time search requests.

[0147] It should be noted that the target recommendation model includes a first sub-model and a second sub-model. The method steps shown in S106 can be implemented through S1061 and S1062, which will be explained in detail below.

[0148] S1061. The second behavior sequence is processed using the first sub-model to obtain the target driver's preference information.

[0149] The first sub-model is a neural network model based on the attention mechanism.

[0150] The second line sequence serves as the input to the first sub-model. The first sub-model analyzes and processes the second line sequence, outputting the target driver's preference information.

[0151] It should be noted that the second behavior sequence consists of filtered second historical click behavior data, or second historical click behavior data and third historical click behavior data, or second historical click behavior data and fourth historical click behavior data. The second behavior sequence can cover the target driver's long-term freight-finding behavior characteristics while ensuring data relevance, and capture freight-finding habits that are closer to the target driver's actual cycle, so that the preference information extracted by the first sub-model can better reflect the target driver's stable periodic interests and long-term preferences.

[0152] S1062. The second sub-model is used to process the target driver's preference information and the second source of goods, as well as the target driver's historical behavior statistics and real-time behavior statistics, to obtain the recommendation results corresponding to the real-time search request.

[0153] The second sub-model is the Hierarchical Information Extraction Network (HiNet) model.

[0154] It should be noted that the method of processing the target driver's preference information, the second source of goods information, the target driver's historical behavior statistics and real-time behavior statistics through the second sub-model to obtain the recommendation results corresponding to the real-time search request is an existing method, and will not be described in detail in this embodiment.

[0155] This application provides a freight recommendation method. It receives real-time search requests from target drivers, acquires a first preset number of historical click behavior data corresponding to the target driver, and generates a first behavior sequence. Based on this, each piece of first historical click behavior data is generalized to the real-time search request, and a total generalization score is calculated. Then, the first historical click behavior data is filtered according to the preset generalization score to obtain second historical click behavior data with higher relevance to the real-time search request. A second behavior sequence is determined based on the second historical click behavior data. Finally, the second behavior sequence, along with second freight information, the target driver's historical behavior statistics, and real-time behavior statistics, are input into a target recommendation model to generate recommendation results sorted by relevance. The freight recommendation method provided by this application can effectively utilize second historical click behavior data, which has a longer coverage and stronger relevance to real-time requests, to model the long-term interests and periodic patterns of drivers, even when the online inference performance of the target recommendation model is limited. This solves the problem in existing technologies where relying only on a small number of recent click behaviors leads to insufficient accuracy in freight recommendations and difficulty in meeting the actual needs of drivers.

[0156] In the above embodiments, the electronic device needs to perform generalization processing on each first historical click behavior data based on a real-time search request to obtain the total generalization score corresponding to each first historical click behavior data. Next, the specific process by which the electronic device performs generalization processing on any first historical click behavior data based on a real-time search request to obtain the total generalization score corresponding to that first historical click behavior data will be described in detail.

[0157] In one possible embodiment, the method steps shown in S103 can be implemented by S1031 to S1036, which are described in detail below.

[0158] S1031. For each first historical click behavior data, determine the first generalization score based on the second origin and the first origin.

[0159] Among them, the first generalization score Used to characterize the degree of deviation of the second departure point from the first departure point.

[0160] S1032. Determine the second generalization score based on the second destination and the first destination.

[0161] S1033. Determine the third generalization score based on the second vehicle length and at least one first vehicle length.

[0162] S1034. Determine the fourth generalization score based on the second vehicle type and at least one first vehicle type.

[0163] Among them, the fourth generalization score Used to characterize the degree of difference between the second vehicle model and the first vehicle model defined in the real-time search request.

[0164] S1035. Determine the fifth generalization score based on the tonnage of the source cargo and the target tonnage range.

[0165] Among them, the fifth generalization score Used to characterize the degree of difference between the tonnage of the source cargo and the target tonnage range defined in the real-time search request.

[0166] S1036. The sum of the first generalization score, the second generalization score, the third generalization score, the fourth generalization score, and the fifth generalization score is determined as the total generalization score corresponding to the first historical click behavior data.

[0167] In this embodiment, by calculating the generalization scores for five dimensions—origin, destination, vehicle length, vehicle type, and tonnage—and merging them into a total generalization score, the multi-dimensional correlation between the first historical click behavior data and the real-time search request can be quantified. This allows for a more accurate distinction between historical click behavior data with high relevance and historical click behavior data that is irrelevant or has significant deviations during subsequent filtering. This effectively solves the problem in the prior art that it relies only on a small number of recent clicks and cannot fully reflect the driver's periodic interests and actual needs. It provides a foundation of input sequences that cover a longer time window and have higher relevance for the target recommendation model.

[0168] In the above embodiments, the electronic device needs to determine a first generalization score based on the second origin and the first origin, and a second generalization score based on the second destination and the first destination. Next, the specific processes by which the electronic device determines the first generalization score based on the second origin and the first origin, and the second generalization score based on the second destination and the first destination, will be described in detail.

[0169] Figure 3 This is a flowchart illustrating another method for recommending sources of goods provided in an embodiment of this application. Figure 3 As shown, in one possible embodiment, the supply recommendation method provided in this application further includes implementations Sc1 to Sc5, which are described in detail below.

[0170] Sc1. Determine whether the first target location and the second target location are consistent.

[0171] It should be noted that if the first target location and the second target location are the same, the method steps shown in Sc2 shall be executed; if the first target location and the second target location are different, the method steps shown in Sc3 shall be executed.

[0172] Wherein, the first destination is the first departure point or the first destination, the second destination is the second departure point or the second destination, and the target generalization score is the first generalization score or the second generalization score.

[0173] If the first destination is the first departure point, and the second destination is the second departure point, then the target generalization score is the first generalization score.

[0174] If the first destination is the first destination, and the second destination is the second destination, then the target generalization score is the second generalization score.

[0175] Sc2. If the first target location and the second target location are the same, the target generalization score is set to zero.

[0176] It should be noted that if the first departure point and the second departure point are the same, .

[0177] If the first destination and the second destination are the same, .

[0178] Sc3. If the first destination and the second destination are inconsistent, determine whether the first destination is the detailed location selected by the target driver.

[0179] If the first destination is the detailed location selected for the target driver, execute the method steps shown in Sc4; if the first destination is the prefecture-level administrative region selected for the target driver, execute the method steps shown in Sc5.

[0180] Sc4. Given that the first target location is the detailed location selected by the target driver, the target generalization score is determined as the ratio of the target empty driving distance to the segmented fixed generalization distance.

[0181] The detailed location selected by the target driver is the precise geographical location chosen by the target driver in the real-time search request or automatically located by electronic devices.

[0182] For example, similar to the "My Location" function in common navigation software, it can obtain the current location through the positioning module of an electronic device, accurate to the specific street, building, or residential level. Precise geographic location differs from administrative divisions, offering higher accuracy and making it suitable for accurately calculating the empty driving distance between a first destination and a second destination.

[0183] Among them, the target empty driving distance is the straight-line distance between the second target location and the first target location, and the segmented fixed generalized distance is related to the target empty driving distance.

[0184] It should be noted that if the first departure point and the second departure point are different, and the first departure point is the specific location selected by the target driver, , This represents the straight-line distance (target empty driving distance) between the second departure point and the first departure point. express The corresponding segmented fixed generalization distance.

[0185] If the first destination and the second destination are different, and the first destination is the specific location selected by the target driver, , This represents the straight-line distance (target empty driving distance) between the second destination and the first destination. express The corresponding segmented fixed generalization distance.

[0186] It should be noted that the electronic device has a pre-stored mapping table that records the correspondence between different empty driving distance ranges and their corresponding segmented fixed generalization distances. Once the target empty driving distance is obtained, the electronic device can determine the segmented fixed generalization distances related to the target empty driving distance by looking up this mapping table.

[0187] For example, the mapping table specifies that: when the target empty distance is less than or equal to 20 km, the segmented fixed generalization distance is 20 km; when the target empty distance is 50 km or more and greater than 20 km, the segmented fixed generalization distance is 50 km; when the target empty distance is 70 km or more and greater than 50 km, the segmented fixed generalization distance is 70 km. When the target empty distance is 35 km, the segmented fixed generalization distance is 50 km; when the target empty distance is 60 km, the segmented fixed generalization distance is 70 km.

[0188] Sc5. If the first target location is the prefecture-level administrative region selected by the target driver, the target generalization score is determined as the ratio of the target empty driving distance to the first distance.

[0189] The prefecture-level administrative region selected by the target driver refers to the geographical area chosen by the driver in the real-time search request, based on administrative divisions, such as a city or district level. Unlike detailed locations, prefecture-level administrative regions have a larger spatial scope, typically covering several urban areas or towns. Their positioning accuracy is relatively lower, making them more suitable for representing a driver's demand for cargo within a large area. Calculating empty driving distance based on this range can reflect the driver's intention to find cargo on a larger scale, without relying on specific location points.

[0190] The first distance is the sum of a fixed distance and a preset ratio of transportation distance, and the transportation distance is the distance between the second departure point and the second destination.

[0191] It should be noted that when the first destination is the first departure point, the preset ratio is the first preset ratio; when the first destination is the first final destination, the preset ratio is the second preset ratio; the first preset ratio is less than the second preset ratio.

[0192] The fixed distance, the first preset ratio, and the second preset ratio can all be set by the platform administrators themselves; this embodiment does not impose any specific limitations on this.

[0193] For example, the first preset ratio is 10%, and the second preset ratio is 20%.

[0194] It should be noted that if the first departure point and the second departure point are different, and the first departure point is the prefecture-level administrative region selected by the target driver, , Indicates a fixed distance. Indicates the first preset ratio. This indicates the distance (transportation distance) between the second point of origin and the second destination.

[0195] If the first destination and the second destination are different, and the first destination is the prefecture-level administrative region selected by the target driver, , This indicates the second preset ratio.

[0196] In this embodiment, the spatial deviation between the first source of goods and the real-time search request can be reasonably quantified under location matching conditions of different granularities, providing a data basis for subsequent calculation of the total generalization score, and solving the problem in the prior art that it is difficult to accurately reflect the driver's true periodic interest based on only a small number of recent click behaviors.

[0197] In the above embodiments, the electronic device needs to determine a third generalization score based on the second vehicle length and at least one first vehicle length. The specific process by which the electronic device determines the third generalization score based on the second vehicle length and at least one first vehicle length will be described in detail below.

[0198] Figure 4 This is a flowchart illustrating another method for recommending sources of goods provided in an embodiment of this application. Figure 4 As shown, in one possible embodiment, the method steps shown in S1033 can be implemented by Sd01 to Sd10, which will be described in detail below.

[0199] Sd01, Determine whether the real-time search request includes the first vehicle commander.

[0200] If the real-time search request includes the first vehicle length, execute the method steps shown in Sd02; if the real-time search request does not include the first vehicle length, execute the method steps shown in Sd09.

[0201] Sd02, Determine whether there is a second vehicle length among at least one first vehicle length.

[0202] If a second vehicle length exists among at least one first vehicle length, the method steps shown in Sd03 are executed; if a second vehicle length does not exist among at least one first vehicle length, the method steps shown in Sd04 are executed.

[0203] Sd03, Determine the third generalization score to be zero.

[0204] It should be noted that if the real-time search request includes a first vehicle length, and at least one of the first vehicle lengths contains a second vehicle length, it means that the second vehicle length corresponding to the first historical click behavior data is completely consistent with the first vehicle length set by the target driver in the real-time search request. In this case, it can be determined that the first historical click behavior data has the highest matching degree with the real-time search request in the vehicle length dimension, and therefore is determined to be... .

[0205] Sd04. Determine whether there is a second vehicle length among at least one behavior vehicle length corresponding to the target driver.

[0206] Among them, the behavioral vehicle length corresponding to the target driver is related to the third vehicle length corresponding to the third source of goods and the fourth vehicle length corresponding to the fourth source of goods.

[0207] The third type of cargo information consists of cargo information that the target driver accepted within the second preset historical time period. The fourth type of cargo information consists of cargo information that the target driver clicked on within the second preset historical time period.

[0208] The duration of the second preset historical period is longer than the duration of the first preset historical period. The second preset historical period can be set by the platform administrator, and this embodiment does not impose specific limitations on it. For example, the second preset historical period is the most recent three months.

[0209] It should be noted that the target driver's behavioral vehicle length refers to a vehicle length parameter that reflects the target driver's actual operational preferences within a second preset historical time period. The determination of the behavioral vehicle length does not depend on the target driver's first vehicle length selected in the real-time search request, but rather on the statistical extraction of the target driver's actual interactive behavior on the online freight platform.

[0210] Specifically, the vehicle lengths that appear frequently in the third source of goods information can be used as one of the target driver's behavioral vehicle lengths to reflect the vehicle length requirements that the target driver has actually carried in the historical transaction records; the vehicle lengths that appear more frequently in the fourth source of goods information than the set frequency can also be used as one of the target driver's behavioral vehicle lengths to reflect the target driver's interests and preferences during the browsing phase; and if the target driver clicks on the fourth source of goods information and triggers the feedback behavior of making a phone call, the vehicle length corresponding to the fourth source of goods information containing the feedback of making a phone call can be used as one of the target driver's behavioral vehicle lengths, thereby more accurately depicting the target driver's vehicle length preferences.

[0211] In addition, the system can comprehensively calculate the three types of behaviors—transaction, click, and phone call—according to preset weighting rules, and ultimately determine the vehicle length of at least one behavior corresponding to the target driver.

[0212] It should be noted that if a second vehicle length exists in at least one action vehicle length, the method steps shown in Sd05 shall be executed; if a second vehicle length does not exist in at least one action vehicle length, the method steps shown in Sd06 shall be executed.

[0213] Sd05. If a second vehicle length exists in at least one behavior vehicle length, determine the third generalization score as the second preset value.

[0214] The second preset value can be set by the platform administrator, and this embodiment does not impose any specific limitations on it.

[0215] For example, the second preset value is 1.1.

[0216] If a second vehicle length exists in at least one behavioral vehicle length, it indicates that the second vehicle length corresponding to the first historical click behavior data is consistent with the target driver's actual operation or preferred vehicle length within the second preset historical time period. In this case, it can be determined that the first historical click behavior data did not directly match the first vehicle length set in the real-time search request in terms of vehicle length dimension, but it is consistent with the target driver's historical behavioral characteristics, therefore it is determined that... .

[0217] Sd06. If there is no second vehicle length among at least one vehicle length, determine whether the second vehicle length is less than the target first vehicle length. Wherein, the target first vehicle length is the smallest first vehicle length among at least one first vehicle length.

[0218] If the second vehicle length is less than the target first vehicle length, execute the method steps shown in Sd07; if the second vehicle length is greater than or equal to the target first vehicle length, execute the method steps shown in Sd08.

[0219] Sd07. If the second vehicle length is less than the target first vehicle length, determine the third generalization score as the third preset value.

[0220] The third preset value can be set by the platform administrator, and this embodiment does not impose specific limitations on it. For example, the third preset value is 2.2.

[0221] If the second vehicle length is less than the target first vehicle length, it means that the second vehicle length corresponding to the first historical click behavior data failed to match the first vehicle length set in the real-time search request, and the second vehicle length is less than the minimum first vehicle length required by the target driver in the real-time search request. In this case, it can be determined that the first historical click behavior data has a certain difference from the real-time search request in the vehicle length dimension, but the degree of difference is relatively acceptable, therefore it is determined that... .

[0222] Sd08, if the second vehicle length is greater than or equal to the target first vehicle length, determine the third generalization score as the fourth preset value.

[0223] The fourth preset value is greater than the third preset value. The fourth preset value can be set by the platform administrator, and this embodiment does not impose specific limitations on it. For example, the fourth preset value is 3.1.

[0224] If the second vehicle length is greater than or equal to the target first vehicle length, it indicates that the second vehicle length corresponding to the first historical click behavior data failed to match the first vehicle length set in the real-time search request. Furthermore, if the second vehicle length is not less than the minimum first vehicle length required by the driver in the real-time search request, it can be determined that the first historical click behavior data has a significant difference from the real-time search request in terms of vehicle length. This difference is more pronounced than when the second vehicle length is less than the target first vehicle length, therefore, it is determined that… .

[0225] Sd09. Determine whether there is a second vehicle length among at least one behavior vehicle length corresponding to the target driver.

[0226] If the real-time search request does not include the first vehicle length, and there is a second vehicle length among at least one action vehicle length, execute the method steps shown in Sd03; if the real-time search request does not include the first vehicle length, and there is no second vehicle length among at least one action vehicle length, execute the method steps shown in Sd10.

[0227] It should be noted that if the real-time search request does not include the first vehicle length, and at least one behavior vehicle length includes the second vehicle length, the third generalization score is determined to be zero.

[0228] Sd10, Determine the third generalization score as the first preset value.

[0229] The first preset value can be set by the platform administrator, and this embodiment does not impose specific limitations on it. For example, the first preset value is 2.1.

[0230] If the real-time search request does not include the first vehicle length, and the second vehicle length is not present in at least one of the target driver's behavioral vehicle lengths, it indicates that the second vehicle length corresponding to the first historical click behavior data is neither constrained by the real-time search request's limitations nor matches the target driver's historical preferred vehicle length. In this case, it can be determined that the first historical click behavior data has a low correlation with the real-time search request in terms of vehicle length, therefore... .

[0231] In the above embodiments, the electronic device needs to determine a fourth generalization score based on the second vehicle model and at least one first vehicle model. The specific process by which the electronic device determines the fourth generalization score based on the second vehicle model and at least one first vehicle model will be described in detail below.

[0232] Figure 5 This is a flowchart illustrating another method for recommending sources of goods provided in an embodiment of this application. Figure 5 As shown, in one possible embodiment, the method steps shown in S1034 can be implemented by Se01 to Se07, which will be described in detail below.

[0233] Se01, Determine whether the real-time search request includes the first vehicle model.

[0234] If the real-time search request includes the first vehicle model, execute the method steps shown in Se02; if the real-time search request does not include the first vehicle model, execute the method steps shown in Se06.

[0235] Se02, determine whether a second model exists in at least one first model.

[0236] If a second vehicle model is present in at least one first vehicle model, the method steps shown in Se03 are performed; if a second vehicle model is not present in at least one first vehicle model, the method steps shown in Se04 are performed.

[0237] Se03, determine that the fourth generalization score is zero.

[0238] It should be noted that if the real-time search request includes a first vehicle model, and at least one of the first vehicle models contains a second vehicle model, it means that the second vehicle model corresponding to the first historical click behavior data is completely consistent with the first vehicle model set by the target driver in the real-time search request. In this case, it can be determined that the first historical click behavior data has the highest matching degree with the real-time search request in terms of vehicle model, and therefore is determined to be... .

[0239] Se04. Determine whether a second vehicle type exists in at least one of the vehicle types corresponding to the target driver's behavior.

[0240] If a second vehicle model exists in at least one vehicle model, perform the method steps shown in Se03; if a second vehicle model does not exist in at least one vehicle model, perform the method steps shown in Se05.

[0241] If the real-time search request includes a first vehicle type, and at least one of the first vehicle types does not contain a second vehicle type, but at least one behavioral vehicle type contains a second vehicle type, it indicates that although the second vehicle type corresponding to the first historical click behavior data is inconsistent with the first vehicle type in the real-time search request, the second vehicle type corresponding to the first historical click behavior data matches the vehicle type preference shown by the target driver in historical behavior. In this case, it can be determined that the first historical click behavior data has a high correlation with the target driver's interest needs in the vehicle type dimension, and therefore it is determined that... .

[0242] Se05, Determine the fourth generalization score as the fifth preset value.

[0243] The fifth preset value can be set by the platform administrator, and this embodiment does not impose specific limitations on it. For example, the fifth preset value is 2.2.

[0244] If the real-time search request includes a first vehicle type, and at least one of the first vehicle types does not include a second vehicle type, and at least one behavioral vehicle type does not include a second vehicle type, it indicates that the second vehicle type corresponding to the first historical click behavior data neither matches the first vehicle type set in the real-time search request nor conforms to the vehicle type preference shown by the target driver in historical behavior. In this case, it can be determined that the first historical click behavior data has a low relevance to the real-time search request in terms of vehicle type dimension, therefore, it is determined that... .

[0245] Se06. Determine whether a second vehicle type exists in at least one of the vehicle types corresponding to the target driver's behavior.

[0246] If a second vehicle model exists in at least one vehicle model, perform the method steps shown in Se03; if a second vehicle model does not exist in at least one vehicle model, perform the method steps shown in Se07.

[0247] If the real-time search request does not include the first vehicle type, and at least one of the vehicle types in the behavioral search includes the second vehicle type, it indicates that the second vehicle type corresponding to the first historical click behavior data matches the vehicle type preference shown by the target driver in the historical behavior. In this case, it can be determined that the first historical click behavior data has a high correlation with the target driver's interest needs in terms of vehicle type, and therefore it is determined that... .

[0248] Se07, Determine the fourth generalization score as the sixth preset value.

[0249] The sixth preset value can be set by the platform administrator, and this embodiment does not impose specific limitations on it. For example, the sixth preset value is 1.1.

[0250] If the real-time search request does not include the first vehicle type, and at least one of the vehicle types in the behavior does not include the second vehicle type, it indicates that the second vehicle type corresponding to the first historical click behavior data is neither constrained by the vehicle type in the real-time search request nor matches the vehicle type preference shown by the target driver in historical behavior. In this case, it can be determined that the first historical click behavior data lacks effective matching basis in the vehicle type dimension, therefore, it is determined that... .

[0251] In the above embodiments, the electronic device needs to determine the fifth generalization score based on the source weight and the target weight range. The specific process by which the electronic device determines the fifth generalization score based on the source weight and the target weight range will be described in detail below.

[0252] Figure 6 This is a flowchart illustrating another method for recommending goods provided in an embodiment of this application. Figure 6 As shown, in one possible embodiment, the method steps shown in S1035 can be implemented by Sf01 to Sf05, which will be described in detail below.

[0253] Sf01, Determine whether the real-time search request includes the target tonnage range.

[0254] If the real-time search request includes the target tonnage range, execute the method steps shown in Sf02; if the real-time search request does not include the target tonnage range, execute the method steps shown in Sf05.

[0255] Sf02. Determine whether the weight of the goods is within the target weight range.

[0256] If the weight of the goods is within the target weight range, execute the method steps shown in Sf03; if the weight of the goods is outside the target weight range, execute the method steps shown in Sf04.

[0257] Sf03, Determine the fifth generalization score to be zero.

[0258] If the tonnage of the cargo falls within the target tonnage range, it means that the tonnage of the cargo corresponding to the first historical click behavior data is completely consistent with the target tonnage range set by the target driver in the real-time search request. In this case, it can be determined that the first historical click behavior data has the highest matching degree with the real-time search request in the tonnage dimension, and therefore is determined to be valid. .

[0259] Sf04. Determine that the fifth generalization score is the sum of the fixed value and the second ratio, where the second ratio is the ratio of the source weight to the first weight.

[0260] Where the rated load is greater than or equal to the second ton, the first ton is the rated load; where the rated load is less than the second ton, the first ton is the second ton; and the second ton is the maximum ton corresponding to the target ton range.

[0261] It should be noted that the approved load capacity can be calculated based on the vehicle parameters registered by the target driver. Specifically, the basic parameters of the target driver's vehicle, such as vehicle length, model, and number of axles, are first obtained and then matched with a pre-stored standard load capacity comparison table for vehicles on the online freight platform. This standard load capacity comparison table can be established by the online freight platform based on relevant regulations or experience data from the transportation industry, and is used to reflect the standard approved load capacity values ​​corresponding to different combinations of vehicle length, model, and number of axles.

[0262] For example, when the vehicle is a van, 9.6 meters long and has 2 axles, the corresponding rated load in the reference table can be set to the first value; when the vehicle is a flatbed, 13 meters long and has 3 axles, the corresponding rated load in the reference table can be set to the second value.

[0263] If the tonnage of the cargo is outside the target tonnage range, it indicates that the tonnage of the cargo corresponding to the first historical click behavior data does not meet the load requirement (target tonnage range) set by the target driver in the real-time search request. Therefore, it can be determined that the first historical click behavior data differs from the real-time search request in terms of tonnage, and thus... , Indicates a fixed value. This represents the second ratio. Indicates the tonnage of the goods. Indicates the first ton.

[0264] Based on the rated load ≥Second ton weight In this case, ;exist In this case, .

[0265] Sf05. Determine the fifth generalization score as the sum of a fixed value and the first ratio, where the first ratio is the ratio of the tonnage of the cargo source to the approved load capacity of the target driver's vehicle.

[0266] If the real-time search request does not include the target tonnage range, it indicates that the target driver did not specify the tonnage condition of the cargo during the search. In this case, it is impossible to directly determine the match between the cargo tonnage corresponding to the first historical click behavior data and the target driver's needs. Therefore, it can be determined that the first historical click behavior data lacks explicit constraints in the tonnage dimension. , This indicates the first ratio.

[0267] Figure 7 This is a schematic diagram of a source recommendation device provided in an embodiment of this application. Figure 7 As shown, the product recommendation device 700 provided in this embodiment includes a receiving module 701, an acquisition module 702, a first processing module 703, a screening module 704, a determination module 705, and a second processing module 706.

[0268] The receiving module 701 is used to receive a real-time search request input by the target driver on the online freight platform; wherein the real-time search request includes the first departure point and the first destination selected by the target driver; or, the real-time search request includes the first departure point and the first destination, as well as at least one first vehicle length, at least one first vehicle type and / or target tonnage range selected by the target driver.

[0269] The acquisition module 702 is used to acquire the first behavior sequence corresponding to the target driver. The first behavior sequence includes a first preset number of first historical click behavior data. Each first historical click behavior data includes the second departure point, second destination, second vehicle length, second vehicle type and cargo tonnage corresponding to the first cargo information. The first cargo information is the cargo information clicked by the target driver within the first preset historical time period.

[0270] The first processing module 703 is used to perform generalization processing on each first historical click behavior data based on the real-time search request, so as to obtain the total generalization score corresponding to each first historical click behavior data.

[0271] The filtering module 704 is used to filter a first preset number of first historical click behavior data based on a preset generalization score and the total generalization score corresponding to each first historical click behavior data, so as to obtain at least one second historical click behavior data.

[0272] The determination module 705 is used to determine a second behavior sequence based on at least one second historical click behavior data, or at least one second historical click behavior data and a first behavior sequence.

[0273] The second processing module 706 is used to process the second behavior sequence and the second cargo information, as well as the historical behavior statistics and real-time behavior statistics of the target driver, using the target recommendation model to obtain the recommendation results corresponding to the real-time search request. The recommendation results include multiple third cargo information, which are arranged in descending order of relevance between each third cargo information and the real-time search request. Among them, the second cargo information is the cargo information that has not yet been accepted on the online freight platform.

[0274] It should be understood that the corresponding processes performed by each module have been described in detail in the above method embodiments, and will not be repeated here for the sake of brevity.

[0275] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 8 As shown, the electronic device 800 provided in this embodiment includes a memory 801 and a processor 802.

[0276] The memory 801 can be a separate physical unit, connected to the processor 802 via a bus 803. Alternatively, the memory 801 and processor 802 can be integrated and implemented in hardware. The memory 801 stores program instructions, which the processor 802 calls to execute the operations performed by the electronic device in any of the above method embodiments.

[0277] Optionally, when some or all of the methods in the above embodiments are implemented by software, the electronic device 800 may also include only the processor 802. A memory 801 for storing programs is located outside the electronic device 800, and the processor 802 is connected to the memory via circuits / wires to read and execute the programs stored in the memory. The processor 802 may be a central processing unit (CPU), a network processor (NP), or a combination of a CPU and an NP. The processor 802 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0278] The memory 801 may include volatile memory, such as random-access memory (RAM); the memory may also include non-volatile memory, such as flash memory, hard disk drive (HDD) or solid-state drive (SSD); the memory may also include a combination of the above types of memory.

[0279] For example, this application provides a chip including: an interface circuit and a logic circuit. The interface circuit is used to receive signals from other chips outside the chip and transmit them to the logic circuit, or to send signals from the logic circuit to other chips outside the chip. The logic circuit is used to perform the operations performed by the electronic device in the above method embodiments.

[0280] For example, this application provides a computer-readable storage medium having computer program instructions stored thereon, which are executed by the processor of an electronic device to cause the electronic device to perform the operations performed by the electronic device in the above method embodiments.

[0281] For example, this application provides a computer program product that, when run on an electronic device, causes the electronic device to perform the operations performed by the electronic device in the above method embodiments.

[0282] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for recommending suppliers, characterized in that, The method includes: Receive a real-time search request input by a target driver on an online freight platform; wherein the real-time search request includes a first departure point and a first destination selected by the target driver; or, the real-time search request includes the first departure point and the first destination, as well as at least one first vehicle length, at least one first vehicle type, and / or a target tonnage range selected by the target driver; Obtain the first behavior sequence corresponding to the target driver. The first behavior sequence includes a first preset number of first historical click behavior data. Each first historical click behavior data includes the second departure point, second destination, second vehicle length, second vehicle type, and cargo tonnage corresponding to the first cargo information, as well as the click time when the target driver clicked the first cargo information. The first cargo information is the cargo information clicked by the target driver within the first preset historical time period. Based on the real-time search request, each of the first historical click behavior data is generalized to obtain the total generalization score corresponding to each of the first historical click behavior data. Based on the preset generalization score and the total generalization score corresponding to each of the first historical click behavior data, the first preset number of first historical click behavior data are filtered to obtain at least one second historical click behavior data. A second behavior sequence is determined based on the at least one second historical click behavior data, or the at least one second historical click behavior data and the first behavior sequence; The second behavior sequence and the second cargo information, as well as the historical and real-time behavior statistics of the target driver, are processed using a target recommendation model to obtain the recommendation result corresponding to the real-time search request. The recommendation result includes multiple third cargo information, which are arranged in descending order of relevance between each third cargo information and the real-time search request. The second cargo information is the cargo information that has not yet been accepted on the online freight platform. The step of determining the second behavior sequence based on the at least one second historical click behavior data, or the at least one second historical click behavior data and the first behavior sequence, includes: If the number of at least one second historical click behavior data is greater than or equal to a second preset number, based on the click time corresponding to the first historical click behavior data, the second preset number of second historical click behavior data whose click time is closest to the current time is determined as the second behavior sequence; wherein, the second preset number is less than the first preset number; If the number of at least one second historical click behavior data is less than the second preset number, a third historical click behavior data is determined based on the first behavior sequence and the at least one second historical click behavior data; wherein, the third historical click behavior data is the first historical click behavior data in the first behavior sequence other than the at least one second historical click behavior data; If the number of third historical click behavior data is greater than the target number, based on the total generalization score corresponding to each third historical click behavior data, the target number of third historical click behavior data are selected as fourth historical click behavior data in ascending order of total generalization score, and the second historical click behavior data and the fourth historical click behavior data are determined as the second behavior sequence; wherein, the target number is the difference between the second preset number and the number of at least one second historical click behavior data; If the number of third historical click behavior data is less than or equal to the target number, the second historical click behavior data and the third historical click behavior data are determined as the second behavior sequence.

2. The method according to claim 1, characterized in that, The generalization process, based on the real-time search request, is performed on each of the first historical click behavior data to obtain a total generalization score corresponding to each of the first historical click behavior data, including: For each of the first historical click behavior data, a first generalization score is determined based on the second origin and the first origin; Determine the second generalization score based on the second destination and the first destination; The third generalization score is determined based on the second vehicle length and the at least one first vehicle length; A fourth generalization score is determined based on the second vehicle model and the at least one first vehicle model; The fifth generalization score is determined based on the source weight and the target weight range; The sum of the first generalization score, the second generalization score, the third generalization score, the fourth generalization score, and the fifth generalization score is determined as the total generalization score corresponding to the first historical click behavior data.

3. The method according to claim 2, characterized in that, The method further includes: If the first destination and the second destination are the same, the target generalization score is determined to be zero; wherein the first destination is the first departure point or the first destination, the second destination is the second departure point or the second destination, and the target generalization score is the first generalization score or the second generalization score; if the first destination is the first departure point, the second destination is the second departure point, and the target generalization score is the first generalization score; if the first destination is the first destination, the second destination is the second destination, and the target generalization score is the second generalization score. When the first target location and the second target location are inconsistent, and the first target location is the detailed location selected by the target driver, the target generalization score is determined to be the ratio of the target empty driving distance to the segmented fixed generalization distance; wherein, the target empty driving distance is the straight-line distance between the second target location and the first target location, and the segmented fixed generalization distance is related to the target empty driving distance; If the first destination and the second destination are different, and the first destination is a prefecture-level administrative region selected by the target driver, the target generalization score is determined to be the ratio of the target empty driving distance to the first distance; wherein, the first distance is the sum of a fixed distance and a preset proportion of the transportation distance, and the transportation distance is the distance between the second departure point and the second destination.

4. The method according to claim 3, characterized in that, The step of determining the third generalization score based on the second vehicle length and the at least one first vehicle length includes: If the second vehicle length meets the first preset condition, the third generalization score is determined to be zero; The first preset condition includes any one of the following: The real-time search request includes the first vehicle length, and the second vehicle length exists among the at least one first vehicle length; The real-time search request does not include the first vehicle length, and the second vehicle length is present in at least one of the behavioral vehicle lengths corresponding to the target driver; the behavioral vehicle length corresponding to the target driver is related to the third vehicle length corresponding to the third cargo information and the fourth vehicle length corresponding to the fourth cargo information; the third cargo information is the cargo information that the target driver accepted orders for within a second preset historical time period; the fourth cargo information is the cargo information that the target driver clicked on within the second preset historical time period. If the real-time search request does not include the first vehicle length, and the second vehicle length is not among the at least one behavior vehicle length, the third generalization score is determined to be a first preset value; If the real-time search request includes the first vehicle length, the second vehicle length is not present in the at least one first vehicle length, and the second vehicle length is present in the at least one behavioral vehicle length, then the third generalization score is determined to be the second preset value. If the real-time search request includes the first vehicle length, and the second vehicle length is not present in either the at least one first vehicle length or the at least one behavioral vehicle length, determine whether the second vehicle length is less than the target first vehicle length, wherein the target first vehicle length is the smallest first vehicle length among the at least one first vehicle length; If the second vehicle length is less than the target first vehicle length, the third generalization score is determined to be a third preset value; If the second vehicle length is greater than or equal to the target first vehicle length, the third generalization score is determined to be a fourth preset value; wherein the fourth preset value is greater than the third preset value.

5. The method according to claim 4, characterized in that, The step of determining the fourth generalization score based on the second vehicle model and the at least one first vehicle model includes: If the second vehicle model meets the second preset condition, the fourth generalization score is determined to be zero; The second preset condition includes any one of the following: The real-time search request includes the first vehicle model, and the second vehicle model exists among the at least one first vehicle model; The real-time search request includes the first vehicle model, the second vehicle model not existing in at least one of the first vehicle models, and the second vehicle model existing in at least one of the behavior vehicle models corresponding to the target driver; The real-time search request does not include the first vehicle type, and the second vehicle type exists among the at least one behavioral vehicle type; wherein, the behavioral vehicle type corresponding to the target driver is related to the third vehicle type corresponding to the third cargo source information and the fourth vehicle type corresponding to the fourth cargo source information; If the real-time search request includes the first vehicle model, and the second vehicle model is not present in either the at least one first vehicle model or the at least one behavioral vehicle model, then the fourth generalization score is determined to be the fifth preset value. If the real-time search request does not include the first vehicle model, and the second vehicle model is not among the at least one behavioral vehicle model, the fourth generalization score is determined to be the sixth preset value.

6. The method according to claim 3, characterized in that, The determination of the fifth generalization score based on the source weight and the target weight range includes: If the real-time search request includes the target tonnage range and the tonnage of the source cargo is within the target tonnage range, the fifth generalization score is determined to be zero. If the real-time search request does not include the target tonnage range, the fifth generalization score is determined to be the sum of a fixed value and a first ratio, where the first ratio is the ratio of the tonnage of the cargo source to the approved load capacity of the target driver's vehicle. If the real-time search request includes the target tonnage range and the source tonnage is outside the target tonnage range, the fifth generalization score is determined to be the sum of the fixed value and the second ratio, where the second ratio is the ratio of the source tonnage to the first tonnage; if the approved load is greater than or equal to the second tonnage, the first tonnage is the approved load; if the approved load is less than the second tonnage, the first tonnage is the second tonnage; the second tonnage is the maximum tonnage corresponding to the target tonnage range.

7. The method according to claim 1, characterized in that, The first preset number of first historical click behavior data are filtered based on a preset generalization score and the total generalization score corresponding to each first historical click behavior data to obtain at least one second historical click behavior data, including: For each first historical click behavior data, if the total generalization score corresponding to the first historical click behavior data is less than the preset generalization score, the first historical click behavior data is determined as the second historical click behavior data.

8. A source recommendation device, characterized in that, The device includes: The receiving module receives a real-time search request input by the target driver on the online freight platform; wherein the real-time search request includes the first departure point and the first destination selected by the target driver; or, the real-time search request includes the first departure point and the first destination, as well as at least one first vehicle length, at least one first vehicle type and / or target tonnage range selected by the target driver. The acquisition module acquires a first behavior sequence corresponding to the target driver. The first behavior sequence includes a first preset number of first historical click behavior data. Each first historical click behavior data includes a second departure point, a second destination, a second vehicle length, a second vehicle type, and a cargo tonnage corresponding to the first cargo information, as well as the click time when the target driver clicked on the first cargo information. The first cargo information is the cargo information clicked by the target driver within a first preset historical time period. The first processing module performs generalization processing on each of the first historical click behavior data based on the real-time search request to obtain the total generalization score corresponding to each of the first historical click behavior data. The filtering module filters the first preset number of first historical click behavior data based on the preset generalization score and the total generalization score corresponding to each first historical click behavior data, to obtain at least one second historical click behavior data. The determining module determines a second behavior sequence based on the at least one second historical click behavior data, or the at least one second historical click behavior data and the first behavior sequence; The second processing module uses a target recommendation model to process the second behavior sequence, the second cargo information, and the historical and real-time behavior statistics of the target driver to obtain the recommendation result corresponding to the real-time search request. The recommendation result includes multiple third cargo information items, which are arranged in descending order of relevance between each third cargo information item and the real-time search request. The second cargo information item is the cargo information that has not yet been accepted on the online freight platform. The determining module is specifically used to: when the number of at least one second historical click behavior data is greater than or equal to a second preset number, based on the click time corresponding to the first historical click behavior data, determine the second preset number of second historical click behavior data with the click time closest to the current moment as the second behavior sequence; wherein, the second preset number is less than the first preset number; If the number of at least one second historical click behavior data is less than the second preset number, a third historical click behavior data is determined based on the first behavior sequence and the at least one second historical click behavior data; wherein, the third historical click behavior data is the first historical click behavior data in the first behavior sequence other than the at least one second historical click behavior data; If the number of third historical click behavior data is greater than the target number, based on the total generalization score corresponding to each third historical click behavior data, the target number of third historical click behavior data are selected as fourth historical click behavior data in ascending order of total generalization score, and the second historical click behavior data and the fourth historical click behavior data are determined as the second behavior sequence; wherein, the target number is the difference between the second preset number and the number of at least one second historical click behavior data; If the number of third historical click behavior data is less than or equal to the target number, the second historical click behavior data and the third historical click behavior data are determined as the second behavior sequence.

9. An electronic device, characterized in that, include: Memory and at least one processor; The memory stores computer-executed instructions; The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method as described in any one of claims 1 to 7.

Images