Logistics information recommendation method, apparatus and device, and computer storage medium

Abstract

The embodiment of the invention provides a logistics information recommendation method and device, electronic equipment and a computer storage medium. The method comprises the steps that a prediction data set is acquired; predicting the prediction data set by using a pre-trained first depth model to obtain an output vector of a current bid invitation line side and an output vector of a carrier side; predicting the output vector of the current bid invitation line side and the output vector of the carrier side by using a pre-trained second depth model to obtain a prediction result of the current bid invitation line; the prediction result comprises a bid winning prediction probability of each carrier in a plurality of carriers; and based on the prediction result of the current bid invitation line, determining a recommendation list meeting the bid invitation requirement.

Description

Technical Field

[0001] The present application relates to the field of logistics technology, and in particular to a logistics information recommendation method, device, electronic device and computer storage medium. Background Art

[0002] At present, when logistics platforms bid for routes, they lack the logic of matching routes with carriers; the general practice adopted during bidding is to notify the carriers of the full amount. In addition, a few special routes require manual experience judgment before subsequent bidding. When judging through manual experience, the main reference factors are the changes in the volume of cargo on the route and the carrier's carrying capacity. However, land transportation of logistics can include trunk lines, branch lines, ferries, and transfer stations. If it is extended to the route level dimension, the number of route sections that need to be bid is very large, and the candidate pool of carriers is very large. Business personnel have limited energy to sort out the matching of routes and carriers; how to accurately find a number of carriers whose matching degree meets the bidding requirements for the current bidding routes is an urgent problem to be solved. Summary of the invention

[0003] The present application provides a logistics information recommendation method, device, electronic device and computer storage medium.

[0004] The technical solution of this application is implemented as follows:

[0005] The present application provides a logistics information recommendation method, the method comprising:

[0006] Acquire a prediction data set; the prediction data set includes current bidding route data and multiple carrier data;

[0007] Using the pre-trained first deep model to predict the prediction data set, and obtaining an output vector of the current bidding line side and an output vector of the carrier side;

[0008] Using the pre-trained second deep model, the output vector of the current bidding route side and the output vector of the carrier side are predicted to obtain a prediction result of the current bidding route; the prediction result includes a predicted probability of winning the bid for each of the multiple carriers;

[0009] Based on the prediction results of the current bidding route, a recommended list that meets the bidding requirements is determined.

[0010] In some embodiments, the predicting the prediction data set using the pre-trained first deep model includes:

[0011] Extracting features from the prediction data set to obtain current bidding line-side features and carrier-side features;

[0012] Performing feature processing and screening on the current bidding line side features and the carrier side features respectively to obtain processed current bidding line side features and carrier side features;

[0013] The pre-trained first deep model is used to predict the processed current bidding line side features and carrier side features.

[0014] In some embodiments, the predicting the output vector of the current bidding route side and the output vector of the carrier side by using the pre-trained second deep model includes:

[0015] Determine the current cross-features between the current bidding route-side features and the carrier-side features;

[0016] Performing feature processing and screening on the current cross feature to obtain processed current cross feature;

[0017] Splicing the processed current intersection feature, the output vector of the current bidding route side, and the output vector of the carrier side to obtain a spliced ​​feature;

[0018] The pre-trained second deep model is used to predict the spliced ​​features.

[0019] In some embodiments, the prediction data set also includes other logistics-related data of the current bidding route during the carrier time, and the processed current cross feature, the output vector of the current bidding route side, and the output vector of the carrier side are spliced ​​to obtain the spliced ​​feature, including:

[0020] Extracting features from other logistics-related data in the prediction data set to obtain related party features;

[0021] The processed current cross feature, the related party feature, the output vector of the current bidding line side and the output vector of the carrier side are spliced ​​to obtain the spliced ​​feature.

[0022] In some embodiments, determining a recommended list that meets the bidding requirements based on the prediction result of the current bidding route includes:

[0023] The forecast results are filtered according to set business rules to obtain the recommended list that meets the bidding requirements; the recommended list includes at least one carrier.

[0024] In some embodiments, the method further comprises:

[0025] Obtaining basic logistics data; the basic logistics data includes historical bidding route data and winning carrier data corresponding to the historical bidding routes;

[0026] Performing feature processing on the basic logistics data to obtain a first sample set; the first sample set includes features related to historical bidding routes and successful bidders;

[0027] Dividing the first sample set to obtain a first training sample set and a first verification sample set;

[0028] The first initial deep model is trained using the first training sample set, and the first initial deep model is verified and fine-tuned using the first verification sample set to obtain a trained first deep model.

[0029] In some embodiments, the basic logistics data also includes other logistics-related data of the historical bidding route during the transportation time, and the method further includes:

[0030] Performing feature processing on the basic logistics data to obtain a second sample set; the second sample set includes features related to historical bidding routes, successful bidders, and historical related parties;

[0031] Dividing the second sample set to obtain a second training sample set and a second verification sample set;

[0032] The second initial deep model is trained using the second training sample set, and the second initial deep model is verified and fine-tuned using the second verification sample set to obtain a trained second deep model.

[0033] The embodiment of the present application also proposes a logistics information recommendation device, which includes an acquisition module, a first prediction module, a second prediction module and a training module, wherein:

[0034] An acquisition module, used to acquire a prediction data set; the prediction data set includes current bidding route data and multiple carrier data;

[0035] A first prediction module is used to predict the prediction data set using a pre-trained first deep model to obtain an output vector of the current bidding line side and an output vector of the carrier side;

[0036] A second prediction module is used to predict the output vector of the current bidding route side and the output vector of the carrier side by using the pre-trained second deep model to obtain a prediction result of the current bidding route; the prediction result includes a predicted probability of winning the bid for each of the multiple carriers;

[0037] The determination module is used to determine a recommended list that meets the bidding requirements based on the prediction results of the current bidding line.

[0038] An embodiment of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a logistics information recommendation method provided by one or more of the aforementioned technical solutions is implemented.

[0039] An embodiment of the present application provides a computer storage medium, which stores a computer program; after the computer program is executed, it can implement the logistics information recommendation method provided by one or more of the above-mentioned technical solutions.

[0040] The embodiments of the present application propose a logistics information recommendation method, device, electronic device and computer storage medium, the method comprising: obtaining a prediction data set; the prediction data set comprises current bidding route data and multiple carrier data; using a pre-trained first deep model to predict the prediction data set to obtain an output vector on the current bidding route side and an output vector on the carrier side; using a pre-trained second deep model to predict the output vector on the current bidding route side and the output vector on the carrier side to obtain a prediction result of the current bidding route; the prediction result comprises a predicted probability of winning the bid for each of multiple carriers; based on the prediction result of the current bidding route, determining a recommended list that meets the bidding requirements.

[0041] It can be seen that the embodiment of the present application does not require manual matching of routes and carriers, but instead uses a two-layer deep model to predict the predicted probability of winning the bid of each carrier matching the current bidding route, and can screen out various high-quality carriers that meet the bidding requirements based on the predicted probability of winning the bid of each carrier. In this way, business personnel only need to focus on inviting bids from several carriers that meet the bidding requirements, and there is no need to invite bids from all carriers, thereby reducing the failure rate; in addition, the various carriers recommended by the above model will generate relatively low quotations or have a large room for bargaining, thereby improving the overall bidding efficiency and reducing the carrier cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a flow chart of a logistics information recommendation method in an embodiment of the present application;

[0043] Figure 2 is a schematic structural diagram of a first initial depth model in an embodiment of the present application;

[0044] Figure 3 This is a schematic diagram of a process of inviting bids from carriers in an embodiment of the present application;

[0045] Figure 4 It is a schematic diagram of the composition structure of the logistics information recommendation device according to an embodiment of the present application;

[0046] Figure 5 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] The present application will be further described in detail below in conjunction with the accompanying drawings and Examples. It should be understood that the embodiments provided herein are only used to explain the present application and are not intended to limit the present application. In addition, the embodiments provided below are partial embodiments for implementing the present application, rather than providing all embodiments for implementing the present application. In the absence of conflict, the technical solutions recorded in the embodiments of the present application can be implemented in any combination.

[0048] It should be noted that, in the embodiments of the present application, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a method or apparatus including a series of elements includes not only the elements explicitly recorded, but also includes other elements not explicitly listed, or also includes elements inherent to the implementation of the method or apparatus. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other related elements (such as steps in a method or units in a device, for example, a unit may be a portion of a circuit, a portion of a processor, a portion of a program or software, etc.) in the method or apparatus including the element.

[0049] The term "and / or" herein is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, I and / or J may indicate the existence of I alone, I and J at the same time, and J alone. In addition, the term "at least one" herein indicates any one of a plurality of or any combination of at least two of a plurality of. For example, at least one of I, J, and R may indicate any one or more elements selected from the set consisting of I, J, and R.

[0050] For example, the logistics information recommendation method provided in the embodiment of the present application includes a series of steps, but the logistics information recommendation method provided in the embodiment of the present application is not limited to the recorded steps. Similarly, the logistics information recommendation device provided in the embodiment of the present application includes a series of modules, but the logistics information recommendation device provided in the embodiment of the present application is not limited to including the modules explicitly recorded, and may also include modules required to obtain relevant task data or perform processing based on task data.

[0051] The embodiments of the present application can be applied to a computer system composed of terminal devices and / or servers, and can operate with many other general or special computing system environments or configurations. Here, the terminal device can be a thin client, a thick client, a handheld or laptop device, a microprocessor-based system, a set-top box, a programmable consumer electronic product, a network personal computer, a small computer system, etc., and the server can be a small computer system, a large computer system, and a distributed cloud computing technology environment including any of the above systems, etc.

[0052] Electronic devices such as terminal devices and servers can implement corresponding functions through the execution of program modules. Generally, program modules can include routines, programs, target programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in a distributed cloud computing environment, in which tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules can be located on local or remote computing system storage media including storage devices.

[0053] At present, land transportation of logistics mainly includes trunk lines, branch lines, ferries, and transfer stations. If it is extended to the route level dimension, the number of tenders required is very large, and the candidate pool of carriers is very large. Business personnel have limited energy to sort out routes and match carriers; bidding work is often cyclical and volatile. For example, in a year's bidding work, there are obvious peaks in March and August. The high bidding pressure leads to inaccurate matching, resulting in the invited carriers having no carrying capacity, and may also increase bidding costs. At the same time, the factors considered during matching are limited, and it is impossible to fully consider the fluctuations in cargo volume, oil price adjustments, route changes, etc. It is highly dependent on manual experience, and it is difficult to accurately find the carrier with the best quotation.

[0054] In view of the above technical problems, the following embodiments are proposed.

[0055] In some embodiments of the present application, the logistics information recommendation method can be implemented using a processor in a logistics information recommendation device, and the above-mentioned processor can be at least one of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, and a microprocessor.

[0056] Figure 1 is a flow chart of a logistics information recommendation method in an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0057] Step 100: Obtain prediction data set.

[0058] In the embodiment of the present application, the prediction data set may include the current bidding route data and multiple carrier data, and may also include other logistics-related data of the current bidding route during the transportation time. The current bidding route indicates the target route that needs to be bid currently; the transportation time refers to the time period during which the current bidding route needs to transport goods, for example, it may be the next six months, twelve months, etc. after the successful bidding.

[0059] Exemplarily, the current bidding route data may include the route type, route mileage, average daily cargo volume in the future period, starting and ending areas, cities to which the starting and ending areas belong, provinces to which the starting and ending cities belong, business type, required vehicle models, etc. of the current bidding route; multiple carrier data may include the level of each carrier, driver scale, total vehicle scale, proportion of each model, main business type, bidding preference, etc.; among them, the level of the carrier can be determined based on information such as the carrier's registered capital, place of registration, and registration time; other logistics-related data may include the current off-season and peak season conditions, bargaining time, price per kilometer of friendly companies, current oil prices, oil prices in the future period, the average settlement price per kilometer of each carrier at each level between provinces, the average settlement price per kilometer of each carrier at each level between regions, the average settlement price per kilometer of each carrier at each level between cities, etc.; here, the three average settlement prices can be calculated based on relevant data from the last three months or six months.

[0060] Exemplarily, some data in the prediction data set can be directly obtained from the corresponding data source; for example, the line type, line mileage and other data of the current bidding line can be obtained from the line data source; multiple carrier data can be obtained from each carrier data source.

[0061] It should be noted that the average daily cargo volume in the future and the oil price in the future will play a key role in the bargaining reduction of the current bidding route in the later period; therefore, these two parts of data need to be obtained before conducting model prediction.

[0062] Exemplarily, the embodiment of the present application does not limit the method of obtaining these two parts of data; for example, an autoregressive model and a long short-term memory (LSTM) recursive neural network can be used to predict the average daily cargo volume and oil price based on the historical time period, respectively, and the prediction results of the two sub-models are fused in an average weighted manner to finally obtain the average daily cargo volume in the future period and the oil price in the future period; here, the historical time period can be the past three months, six months or twelve months, etc.; the future period can be the next three months, six months or twelve months, etc.

[0063] Exemplarily, after acquiring the above data in the prediction data set, a preliminary cleaning and integration operation will be performed; here, the cleaning and integration operation includes but is not limited to filling in missing data, removing abnormal data, converting data, and other operations on the above acquired data.

[0064] Understandably, in order to ensure the validity of the data in the data source, each time a version of the bargaining results is obtained, the data sources involved can be updated in a timely manner, and the relevant data in the prediction data set can be updated synchronously to ensure the accuracy of subsequent prediction results.

[0065] Step 101: Use the pre-trained first deep model to predict the prediction data set to obtain the output vector of the current bidding line side and the output vector of the carrier side.

[0066] Exemplarily, the first deep model can be a twin-tower model based on a semantic similarity model (Deep Structured Semantic Models, DSSM), or other variant twin-tower models, such as an MV-DSSM model, an LSTM-DSSM model, etc., or other types of deep models, such as DeepFM (Deep Factorization Machines), etc.; the twin-tower model is taken as an example for explanation below.

[0067] In an embodiment of the present application, the first deep model is trained through the following steps: obtaining basic logistics data; performing feature processing on the basic logistics data to obtain a first sample set; the first sample set includes features related to historical bidding routes and winning carriers; dividing the first sample set to obtain a first training sample set and a first verification sample set; using the first training sample set to train the first initial deep model, and using the first verification sample set to verify and fine-tune the first initial deep model to obtain a trained first deep model.

[0068] Exemplarily, the basic logistics data may include historical bidding route data, successful carrier data corresponding to the historical bidding routes, and other logistics-related data of the historical bidding routes during the carrier period.

[0069] Exemplarily, the historical bidding route data may include relevant data of each historical bidding route in the historical time period; for example, when the historical time period is the past three months, the historical bidding route data may include the route type, route mileage, average daily cargo volume, starting and ending areas, cities to which the starting and ending areas belong, provinces to which the starting and ending cities belong, business type, required vehicle model, contract price, etc. of each historical bidding route in the past three months; the winning carrier data corresponding to the historical bidding routes indicates the level, driver scale, total vehicle scale, proportion of each vehicle model, main business type, bidding preference, etc. of each carrier that won the bid for each historical bidding route; other logistics-related data of the historical bidding routes during the carrier time may include the off-season and peak season conditions of each historical bidding route in the historical time period, bargaining time, price per kilometer of friendly companies, oil price, average settlement price per kilometer of each level of each carrier between provinces, average settlement price per kilometer of each level of each carrier between regions, average settlement price per kilometer of each level of each carrier between cities, etc.

[0070] Similarly, some of the data included in the above basic logistics data can be directly obtained from the corresponding data source, and preliminary cleaning and integration operations will be performed after all the data are obtained, which will not be repeated here.

[0071] Exemplarily, feature processing of basic logistics data to obtain a first sample set may include: feature extraction and classification processing of the basic logistics data to obtain historical bidding line-side features and winning carrier-side features; feature processing and screening of the historical bidding line-side features and winning carrier-side features respectively to obtain processed historical bidding line-side features and processed winning carrier-side features, which together constitute the first sample set.

[0072] Here, feature processing and screening processing means that methods such as feature correlation analysis or principal component analysis can be used to screen out several key features from multiple original features; for example, more than 60 key features can be screened out from more than 100 original features.

[0073] Exemplarily, after obtaining the first sample set, the first sample set can be divided according to a set ratio to obtain a first training sample set and a first verification sample set; here, the setting ratio can be set according to actual conditions, and the embodiments of the present application are not limited to this; for example, the setting ratio can be 8:2 or 7:3, etc.

[0074] In an embodiment of the present application, after the first initial depth model is constructed, training the first initial depth model using the first training sample set may include: inputting the first training sample set into the first initial depth model, determining a first loss function of the first initial depth model; and training the first initial depth model using the first loss function.

[0075] See also Figure 2 The first initial deep model includes a line-side encoder and a carrier-side encoder; illustratively, inputting the first training sample set into the first initial deep model, and determining the first loss function of the first initial deep model may include: inputting the training sample set into the first initial deep model, and determining the output vector of each historical bidding line in the first training sample set according to the line-side encoder; determining the output vector of each successful carrier in the first training sample set according to the carrier-side encoder; determining the first similarity between the output vector of each historical bidding line and the output vector of each successful carrier; and determining the first loss function according to the first similarity.

[0076] Here, the first similarity may be cosine similarity or other similarities, and this embodiment of the present application does not limit this.

[0077] Exemplarily, after obtaining the first similarity, the softmax function is used to convert the first similarity into a posterior probability P(C + |L), as shown in formula (1):

[0078]

[0079] Among them, L represents each historical bidding line, C + Represents a positive sample, C' represents one of the samples, γ is the smoothing factor of the softmax function, C is a set of one positive sample and N-1 negative samples, R() represents the cosine similarity function, and L and C are both dense feature vector representations.

[0080] Exemplarily, during the training phase, the first loss function L(Λ) is minimized by maximum likelihood estimation, as shown in formula (2):

[0081]

[0082] Here, Λ represents the parameters of the neural network, P(D + |Q) represents the probability of the corresponding winning carrier (i.e., positive sample) under each historical bidding route. The model can be optimized by stochastic gradient descent; that is, the goal of the loss function is to maximize the probability of positive sample prediction. Finally, through the above training, separate output vector representations of L and C can be obtained. In particular, the carrier routes of each winning carrier in a historical time period (for example, the past six months or twelve months) can also be used as sequence features and input into the first initial deep model for training to further improve the training effect of the model.

[0083] In the embodiment of the present application, after the training of the pre-constructed first initial depth model is completed, the first initial depth model is further verified and fine-tuned using the first verification sample set to obtain a trained first depth model.

[0084] In some embodiments, using the first verification sample set to verify and fine-tune the first initial deep model to obtain a trained first deep model may include: determining the output vector of each historical bidding line in the first verification sample set according to the line-side encoder; determining the output vector of each successful carrier in the first verification sample set according to the carrier-side encoder; determining the second similarity between the output vector of each historical bidding line and the output vector of each successful carrier; determining the second loss function based on the second similarity; and using the second loss function to verify and fine-tune the first initial deep model to obtain the trained first deep model.

[0085] Exemplarily, the implementation method of determining the second loss function to verify and fine-tune the first initial depth model is similar to the implementation method of determining the first loss function to train the first initial depth model, which is not repeated here.

[0086] It can be seen that the embodiment of the present application is conducive to improving the accuracy of model training by training and verifying the first deep model. In addition, the first deep model used in the embodiment of the present application can be a dual-tower model, which has a high degree of parallelism and effectively reduces the time cost of model training.

[0087] Exemplarily, after obtaining the trained first deep model, using the model to predict the prediction data set can include: extracting features from the prediction data set to obtain current bidding line side features and carrier side features; performing feature processing and screening on the current bidding line side features and carrier side features respectively to obtain processed current bidding line side features and carrier side features; and using the pre-trained first deep model to predict the processed current bidding line side features and carrier side features.

[0088] Here, feature extraction is performed on the prediction data set, which is equivalent to feature extraction of the current bidding line data and multiple carrier data in the prediction data set. It should be noted that the implementation method of feature extraction and other feature processing on the prediction data set to obtain the processed current line side features and carrier side features is similar to the above-mentioned feature processing of the basic logistics data, which will not be repeated here.

[0089] It can be understood that when the first deep model is a dual-tower model, the output vector of the current bidding line side and the output vector of the carrier side can be obtained by predicting the processed current line side features and carrier side features through the pre-trained first deep model.

[0090] Step 102: Use the pre-trained second deep model to predict the output vector of the current bidding route side and the output vector of the carrier side to obtain the prediction result of the current bidding route.

[0091] Exemplarily, the second deep model may be a deep neural network (DNN) model, or may be other variant DNN models, for example, a DNN model combined with various dropout methods, etc., which is not limited in the embodiments of the present application. The following description will be given by taking the DNN model as an example.

[0092] In an embodiment of the present application, the second deep model is trained through the following steps: feature processing is performed on the basic logistics data to obtain a second sample set; the second sample set includes features related to historical bidding routes, winning carriers, and historical related parties; the second sample set is divided to obtain a second training sample set and a second verification sample set; the second initial deep model is trained using the second training sample set, and the second initial deep model is verified and fine-tuned using the second verification sample set to obtain a trained second deep model.

[0093] Exemplarily, since the basic logistics data also includes other logistics-related data of historical bidding routes during the transportation period, feature extraction of the basic logistics data can obtain not only the historical bidding route side features and the winning carrier side features, but also the historical related party features.

[0094] In some embodiments, feature processing is performed on basic logistics data to obtain a second sample set, which may include: determining historical cross-features between historical bidding line-side features and winning carrier-side features; and performing feature processing and screening on the historical cross-features and historical related party features, respectively, to obtain processed historical cross-features and historical related party features, which together constitute the second sample set.

[0095] Exemplarily, historical cross-features may include: the number of bidding carriers in the area where the historical bidding routes are located, the bargaining success rate of each winning carrier for all bidding routes in the historical time period (number of successes / total number of bargaining times), the bargaining reduction statistics of each winning carrier for all bidding routes in the historical time period (median, maximum reduction, minimum reduction), and the bargaining reduction statistics of each winning carrier for routes with the same flow direction as each historical bidding route (median, maximum reduction, minimum reduction); historical related party features can be determined based on other logistics-related data of the historical bidding routes during the carrier time.

[0096] It should be noted that the second sample set is used to train the second depth model. Since the process of training the first depth model with the second sample set is similar to that of the first sample set mentioned above, it will not be repeated here.

[0097] It can be seen that in the embodiment of the present application, the representation vectors of different carriers and various historical bidding routes are preliminarily learned through the first deep model. On this basis, more cross-features and other-party features are further combined through the second deep model to improve the prediction performance of the model, thereby being able to recommend better carriers for the current bidding routes.

[0098] In some embodiments, after obtaining the trained second deep model, the output vector of the current bidding line side and the output vector of the carrier side are predicted using the pre-trained second deep model, which can include: determining the current cross-features between the current bidding line side features and the carrier side features; performing feature processing and screening on the current cross-features to obtain processed current cross-features; splicing the processed current cross-features, the output vector of the current bidding line side, and the output vector of the carrier side to obtain spliced ​​features; and predicting the spliced ​​features using the pre-trained second deep model.

[0099] Exemplarily, cross-features may include: the number of bidding carriers in the area where the current bidding route is located, the bargaining success rate of each carrier for all bidding routes in the historical time period (number of successes / total number of bargaining times), the bargaining reduction statistics of each carrier for all bidding routes in the historical time period (median, maximum reduction, minimum reduction), and the bargaining reduction statistics of each winning carrier for routes in the same direction as the current bidding route (median, maximum reduction, minimum reduction).

[0100] In some embodiments, the processed current cross-features, the output vector of the current bidding line side, and the output vector of the carrier side are spliced ​​to obtain the spliced ​​features, which may include: extracting features from other logistics-related data in the prediction data set to obtain related party features; splicing the processed current cross-features, related party features, the output vector of the current bidding line side, and the output vector of the carrier side to obtain the spliced ​​features.

[0101] It can be understood that by predicting the spliced ​​features through the pre-trained second deep model, the predicted probability of winning the bid for each of the multiple carriers can be obtained, that is, the predicted probability of each carrier winning the bid for the current bidding route.

[0102] It can be seen that in the prediction process of the embodiment of the present application, the first deep model is used to predict the current bidding line side characteristics and the carrier side characteristics, and on this basis, the second deep model is used to further predict the cross-features and other party characteristics between the current bidding line side characteristics and the carrier side characteristics, which can effectively improve the accuracy of model prediction.

[0103] Step 103: Based on the prediction results of the current bidding route, determine a recommended list that meets the bidding requirements.

[0104] In some embodiments, determining a recommended list that meets the bidding requirements based on the prediction results of the current bidding line may include: filtering the prediction results according to set business rules to obtain a recommended list that meets the bidding requirements.

[0105] Exemplarily, the set business rules may be rules related to a blacklist; the recommended list includes at least one carrier that meets the bidding requirements.

[0106] In the embodiment of the present application, after obtaining the prediction results of the current bidding route according to the above steps, the first K carriers with the highest predicted probability of winning the bid can be selected from the prediction results, where K is an integer greater than zero; it can be understood that the higher the predicted probability of winning the bid of the carrier, the higher the matching degree between the carrier and the current bidding route. Then, the selected carriers are filtered according to the set business rules, that is, it is determined in turn whether these carriers exist in the blacklist. If so, the corresponding carriers are filtered out, indicating that the carrier does not meet the bidding requirements; otherwise, the carrier is retained, indicating that the carrier meets the bidding requirements.

[0107] It can be seen that after the above filtering process, various carriers that meet the bidding requirements can be obtained; finally, the business personnel send invitation information to these carriers, and the invitation information includes the bidding time to invite these carriers to bid on time. Here, there is no limitation on the way of sending the invitation information, for example, it can be sent by phone or email.

[0108] An embodiment of the present application proposes a logistics information recommendation method, which includes: obtaining a prediction data set; the prediction data set includes current bidding route data and multiple carrier data; using a pre-trained first deep model to predict the prediction data set to obtain an output vector on the current bidding route side and an output vector on the carrier side; using a pre-trained second deep model to predict the output vector on the current bidding route side and the output vector on the carrier side to obtain a prediction result for the current bidding route; the prediction result includes a predicted probability of winning the bid for each of the multiple carriers; based on the prediction result of the current bidding route, determining a recommended list that meets the bidding requirements. It can be seen that the embodiment of the present application does not require manual matching of routes and carriers, but instead uses a two-layer deep model to predict the predicted probability of winning the bid of each carrier matching the current bidding route, and can screen out various high-quality carriers that meet the bidding requirements based on the predicted probability of winning the bid of each carrier. In this way, business personnel only need to focus on inviting bids from several carriers that meet the bidding requirements, and there is no need to invite bids from all carriers, thereby reducing the failure rate; in addition, the various carriers recommended by the above model will generate relatively low quotations or have a large room for bargaining, thereby improving the overall bidding efficiency and reducing the carrier cost.

[0109] In order to better reflect the purpose of the present application, further explanation is given based on the above embodiments of the present application.

[0110] Figure 3 FIG. 1 is a flow chart of inviting bids from carriers according to an embodiment of the present application. Figure 3 As shown, the method comprises the following steps:

[0111] Step 200: Obtain a prediction data set.

[0112] Exemplarily, the prediction data set may include current bidding route data, multiple carrier data, and other logistics-related data of the current bidding route during the carrier time; the contents of each item of data in the prediction data set have been described in the above embodiments and will not be repeated here.

[0113] Step 201: Predict the carrier's bid winning probability based on the prediction data set.

[0114] Exemplarily, the prediction data set is predicted by using the pre-trained first deep model and the second deep model to obtain the predicted probability of winning the bid for each carrier for the current bidding route. The specific process has been described in the above embodiment and will not be repeated here.

[0115] Step 202: Send invitation to bid information.

[0116] For example, after obtaining the predicted probability of winning the bid for the current bidding route, the predicted results can be filtered according to the set business rules to obtain a recommended list that meets the bidding requirements; then, the business personnel send invitation information to each carrier on the recommended list.

[0117] Step 203: The carrier bids.

[0118] For example, after receiving the invitation to bid information sent by the business personnel, the carrier will bid on time according to the bidding time in the invitation to bid information.

[0119] Step 204: Multiple rounds of bargaining.

[0120] Exemplarily, each carrier participating in the bidding conducts multiple rounds of bargaining on the current bidding route.

[0121] Step 205: Win the bid.

[0122] Exemplarily, after successfully winning the bid, the client enters into a bidding contract with the carrier that has successfully won the bid.

[0123] Figure 4 is a schematic diagram of the structure of the logistics information recommendation device according to an embodiment of the present application. Figure 4 As shown, the device includes: an acquisition module 300, a first prediction module 301, a second prediction module 302 and a determination module 303, wherein:

[0124] An acquisition module 300 is used to acquire a prediction data set; the prediction data set includes current bidding route data and multiple carrier data;

[0125] A first prediction module 301 is used to predict the prediction data set using a pre-trained first deep model to obtain an output vector of the current bidding line side and an output vector of the carrier side;

[0126] A second prediction module 302 is used to predict the output vector of the current bidding route side and the output vector of the carrier side using the pre-trained second deep model to obtain a prediction result of the current bidding route; the prediction result includes a predicted probability of winning the bid for each of the multiple carriers;

[0127] The determination module 303 is used to determine a recommended list that meets the bidding requirements based on the prediction result of the current bidding line.

[0128] In some embodiments, the first prediction module 301 is used to:

[0129] Extracting features from the prediction data set to obtain current bidding line-side features and carrier-side features;

[0130] Performing feature processing and screening on the current bidding line side features and the carrier side features respectively to obtain processed current bidding line side features and carrier side features;

[0131] The pre-trained first deep model is used to predict the processed current bidding line side features and carrier side features.

[0132] In some embodiments, the second prediction module 302 is used to:

[0133] Determine the current cross-features between the current bidding route-side features and the carrier-side features;

[0134] Performing feature processing and screening on the current cross feature to obtain processed current cross feature;

[0135] Splicing the processed current intersection feature, the output vector of the current bidding route side, and the output vector of the carrier side to obtain a spliced ​​feature;

[0136] The pre-trained second deep model is used to predict the spliced ​​features.

[0137] In some embodiments, the prediction data set also includes other logistics-related data of the current bidding route during the transportation time. The second prediction module 302 is further used to:

[0138] Extracting features from other logistics-related data in the prediction data set to obtain related party features;

[0139] The processed current cross feature, the related party feature, the output vector of the current bidding line side and the output vector of the carrier side are spliced ​​to obtain the spliced ​​feature.

[0140] In some embodiments, the determining module 303 is used to:

[0141] The forecast results are filtered according to set business rules to obtain the recommended list that meets the bidding requirements; the recommended list includes at least one carrier.

[0142] In some embodiments, the apparatus further comprises a training model, wherein the training module is configured to:

[0143] Obtaining basic logistics data; the basic logistics data includes historical bidding route data and winning carrier data corresponding to the historical bidding routes;

[0144] Performing feature processing on the basic logistics data to obtain a first sample set; the first sample set includes features related to historical bidding routes and successful bidders;

[0145] Dividing the first sample set to obtain a first training sample set and a first verification sample set;

[0146] The first initial deep model is trained using the first training sample set, and the first initial deep model is verified and fine-tuned using the first verification sample set to obtain a trained first deep model.

[0147] In some embodiments, the basic logistics data also includes other logistics-related data of the historical bidding routes during the transportation time. The training module is also used to:

[0148] Performing feature processing on the basic logistics data to obtain a second sample set; the second sample set includes features related to historical bidding routes, successful bidders, and historical related parties;

[0149] Dividing the second sample set to obtain a second training sample set and a second verification sample set;

[0150] The second initial deep model is trained using the second training sample set, and the second initial deep model is verified and fine-tuned using the second verification sample set to obtain a trained second deep model.

[0151] In practical applications, the acquisition module 300, the first prediction module 301, the second prediction module 302, the determination module 303 and the training module can all be implemented by a processor located in an electronic device, and the processor can be at least one of an ASIC, a DSP, a DSPD, a PLD, a FPGA, a CPU, a controller, a microcontroller, and a microprocessor.

[0152] In addition, each functional module in this embodiment can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or software functional modules.

[0153] If the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment is essentially or the part that contributes to the relevant technology or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the method of this embodiment. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc. Various media that can store program codes.

[0154] Specifically, the computer program instructions corresponding to a logistics information recommendation method in this embodiment can be stored on a storage medium such as a CD, a hard disk, or a USB flash drive. When the computer program instructions corresponding to a logistics information recommendation method in the storage medium are read or executed by an electronic device, any logistics information recommendation method in the aforementioned embodiments is implemented.

[0155] Based on the same technical concept as the above embodiments, see Figure 5 , which shows an electronic device 400 provided in an embodiment of the present application, which may include: a memory 401 and a processor 402; wherein,

[0156] Memory 401, used for storing computer programs and data;

[0157] The processor 402 is used to execute the computer program stored in the memory to implement any logistics information recommendation method of the aforementioned embodiments.

[0158] In practical applications, the memory 401 may be a volatile memory, such as RAM; or a non-volatile memory, such as ROM, flash memory, hard disk drive (HDD) or solid-state drive (SSD); or a combination of the above types of memory, and provide instructions and data to the processor 402.

[0159] The processor 402 may be at least one of ASIC, DSP, DSPD, PLD, FPGA, CPU, controller, microcontroller, and microprocessor. It is understandable that for different logistics information recommendation devices, the electronic device used to implement the processor function may also be other, which is not specifically limited in the embodiments of the present application.

[0160] In some embodiments, the functions or modules included in the device provided in the embodiments of the present application can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments. For the sake of brevity, it will not be repeated here.

[0161] The above description of various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other, and for the sake of brevity, they will not be repeated herein.

[0162] The methods disclosed in the various method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0163] The features disclosed in the various product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0164] The features disclosed in the various method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0165] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of hardware embodiments, software embodiments, or embodiments in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.

[0166] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0167] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0168] The above are only preferred embodiments of the present application and are not intended to limit the protection scope of the present application.

Claims

1. A logistics information recommendation method, characterized in that: The method comprises: Acquire a prediction data set; the prediction data set includes current bidding route data and multiple carrier data; Using the pre-trained first deep model to predict the prediction data set, and obtaining an output vector of the current bidding line side and an output vector of the carrier side; Using the pre-trained second deep model, the output vector of the current bidding route side and the output vector of the carrier side are predicted to obtain a prediction result of the current bidding route; the prediction result includes a predicted probability of winning the bid for each of the multiple carriers; Based on the prediction results of the current bidding route, a recommended list that meets the bidding requirements is determined.

2. The method according to claim 1, characterized in that The predicting the prediction data set by using the pre-trained first deep model includes: Extracting features from the prediction data set to obtain current bidding line-side features and carrier-side features; Performing feature processing and screening on the current bidding line side features and the carrier side features respectively to obtain processed current bidding line side features and carrier side features; The pre-trained first deep model is used to predict the processed current bidding line side features and carrier side features.

3. The method according to claim 2, characterized in that The predicting the output vector of the current bidding route side and the output vector of the carrier side by using the pre-trained second deep model includes: Determine the current cross-features between the current bidding route-side features and the carrier-side features; Performing feature processing and screening on the current cross feature to obtain processed current cross feature; Splicing the processed current intersection feature, the output vector of the current bidding route side, and the output vector of the carrier side to obtain a spliced ​​feature; The pre-trained second deep model is used to predict the spliced ​​features.

4. The method according to claim 3, characterized in that The prediction data set also includes other logistics-related data of the current bidding route during the carrier time. The processed current cross feature, the output vector of the current bidding route side, and the output vector of the carrier side are spliced ​​to obtain the spliced ​​features, including: Extracting features from other logistics-related data in the prediction data set to obtain related party features; The processed current cross feature, the related party feature, the output vector of the current bidding line side and the output vector of the carrier side are spliced ​​to obtain the spliced ​​feature.

5. The method according to claim 1, characterized in that The step of determining a recommended list that meets the bidding requirements based on the prediction result of the current bidding line includes: The forecast results are filtered according to set business rules to obtain the recommended list that meets the bidding requirements; the recommended list includes at least one carrier.

6. The method according to claim 1, characterized in that The method further comprises: Obtaining basic logistics data; the basic logistics data includes historical bidding route data and winning carrier data corresponding to the historical bidding routes; Performing feature processing on the basic logistics data to obtain a first sample set; the first sample set includes features related to historical bidding routes and successful bidders; Dividing the first sample set to obtain a first training sample set and a first verification sample set; The first initial deep model is trained using the first training sample set, and the first initial deep model is verified and fine-tuned using the first verification sample set to obtain a trained first deep model.

7. The method according to claim 6, characterized in that The basic logistics data also includes other logistics-related data of the historical bidding route during the transportation time. The method also includes: Performing feature processing on the basic logistics data to obtain a second sample set; the second sample set includes features related to historical bidding routes, successful bidders, and historical related parties; Dividing the second sample set to obtain a second training sample set and a second verification sample set; The second initial deep model is trained using the second training sample set, and the second initial deep model is verified and fine-tuned using the second verification sample set to obtain a trained second deep model.

8. A logistics information recommendation device, characterized in that: The device comprises: An acquisition module, used to acquire a prediction data set; the prediction data set includes current bidding route data and multiple carrier data; A first prediction module is used to predict the prediction data set using a pre-trained first deep model to obtain an output vector of the current bidding line side and an output vector of the carrier side; A second prediction module is used to predict the output vector of the current bidding route side and the output vector of the carrier side by using the pre-trained second deep model to obtain a prediction result of the current bidding route; the prediction result includes a predicted probability of winning the bid for each of the multiple carriers; The determination module is used to determine a recommended list that meets the bidding requirements based on the prediction results of the current bidding line.

9. An electronic device, characterized in that: The device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the method according to any one of claims 1 to 7 is implemented when the processor executes the program.

10. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images