A method, device, equipment and readable medium for address recommendation

Abstract

The present application relates to an address recommendation method, device, equipment and readable medium, wherein the address recommendation method comprises: obtaining a comparison address and an address to be queried input by a target object, wherein the comparison address is a delivery address of a historical order, and the address to be queried is a delivery address to be queried; inputting the address to be queried and the comparison address into a preset address matching model for matching to obtain a matching result; determining a target address that meets preset conditions from the comparison address according to the matching result, and recommending the target address to the target object. The present application obtains an address matching model by training the delivery address of the historical order as a training sample, and completes the matching of the delivery address. The present application can perform address clustering based only on the delivery information of the express delivery, and recommend addresses to the target object.

Description

Technical Field

[0001] The present application relates to the field of address clustering for logistics and express delivery, and in particular to an address recommendation method, device, equipment and readable medium. Background Art

[0002] With the development of the express delivery industry, the volume of logistics has increased, which means that more manpower and material resources are needed to handle the delivery work. However, the current delivery address is still filled in manually, and the irregular filling of the delivery address has brought a lot of workload in determining the accurate address, and also increased the pressure of logistics. In the related technology, the clustering method is a partition clustering algorithm or a density-based clustering algorithm, both of which are based on address data containing longitude and latitude information and some address tag data for clustering. However, in the absence of longitude and latitude information, it is impossible to perform address clustering based on address information and recommend addresses to users.

[0003] With respect to the above-mentioned problem that "in the absence of latitude and longitude information, it is impossible to cluster addresses based on address information and recommend addresses to users", no effective solution has been proposed yet. Summary of the invention

[0004] The present application provides an address recommendation method, apparatus, device and readable medium to solve the above-mentioned technical problem of "in the absence of longitude and latitude information, it is impossible to perform address clustering based on address information and recommend addresses to users".

[0005] According to one aspect of an embodiment of the present application, the present application provides an address recommendation method, including: obtaining a comparison address and an address to be queried input by a target object, wherein the comparison address is a delivery address of a historical order, and the address to be queried is a delivery address to be queried; matching the address to be queried and the comparison address input into a preset address matching model to obtain a matching result; determining a target address that meets preset conditions from the comparison address according to the matching result, and recommending the target address to the target object.

[0006] Optionally, before inputting the address to be queried and the comparison address into a preset address matching model for matching, the method also includes obtaining the address matching model in the following manner: obtaining a sample set, wherein the sample set includes a training sample set and a verification set, and both the training sample set and the verification set are pre-processed delivery addresses; inputting the training sample set into a pre-trained model for training to obtain an intermediate model; verifying the intermediate model using the verification set, and when the recognition accuracy of the intermediate model for the verification set reaches a target threshold, determining the intermediate model as the address matching model; when the recognition accuracy of the intermediate model for the verification set does not reach the target threshold, continuing to train the intermediate model using the training sample set until the recognition accuracy of the intermediate model for the verification set reaches the target threshold, determining the intermediate model as the address matching model.

[0007] Optionally, when the recognition accuracy of the intermediate model for the verification set does not reach the target threshold, continuing to use the training sample set to train the intermediate model includes: determining the loss value of the objective function based on the recognition results of the intermediate model for the training sample set and the verification set; continuing to use the training sample set to train the intermediate model, and adjusting the internal parameters of the intermediate model according to the loss value during the training process until the recognition accuracy of the intermediate model for the verification set reaches the target threshold, and determining the intermediate model as the address matching model.

[0008] Optionally, obtaining the sample set includes: obtaining the address area and delivery person data corresponding to the delivery address; dividing the delivery addresses with the same address area and the same delivery person data into a first sample set, dividing the delivery addresses with the same address area and different delivery person data into a second sample set, dividing the delivery addresses with different address areas and the same delivery person data into a third sample set, and dividing the delivery addresses with different address areas and different delivery person data into a fourth sample set.

[0009] Optionally, the internal parameters include a first probability interval, a second probability interval, a third probability interval and a fourth probability interval; wherein the first probability interval is the first initial parameter of the pre-trained model corresponding to the first sample set, the second probability interval is the second initial parameter of the pre-trained model corresponding to the second sample set, the third probability interval is the third initial parameter of the pre-trained model corresponding to the third sample set, and the fourth probability interval is the fourth initial parameter of the pre-trained model corresponding to the fourth sample set.

[0010] Optionally, the training sample set is input into a pre-trained model for training to obtain an intermediate model, including: inputting the first sample set, the second sample set, the third sample set and the fourth sample set into the pre-trained model for training to adjust the values ​​of the first probability interval, the second probability interval, the third probability interval and the fourth probability interval until the first probability interval, the second probability interval, the third probability interval and the fourth probability interval remain unchanged, thereby obtaining an intermediate model.

[0011] Optionally, determining a target address that meets preset conditions from the comparison addresses according to the matching results, and recommending the target address to the target object includes: determining a probability interval in which the matching result is located, wherein the probability interval is used to represent the degree of matching between the comparison address and the address to be queried; when the matching result is in the target probability interval, determining the target address according to the scores of each comparison address in the matching result; and recommending the target address to the target object.

[0012] Optionally, when the matching result is within the target probability interval, determining the target address according to the score of each comparison address in the matching result includes: sorting each comparison address in the matching result according to the degree of match with the address to be queried to obtain a sorting result, wherein the score includes the degree of match; and determining the comparison address before the target sorting position in the sorting result as the target address.

[0013] According to another aspect of an embodiment of the present application, the present application provides an address recommendation device, including: an acquisition module, used to acquire a comparison address and an address to be queried input by a target object, wherein the comparison address is a delivery address of a historical order, and the address to be queried is a delivery address to be queried; a matching module, used to match the address to be queried and the comparison address input into a preset address matching model to obtain a matching result; a determination module, used to determine a target address that meets preset conditions from the comparison address according to the matching result, and recommend the target address to the target object.

[0014] According to another aspect of an embodiment of the present application, the present application provides an electronic device, including a memory, a processor, a communication interface and a communication bus, wherein the memory stores a computer program that can be run on the processor, the memory and the processor communicate through the communication bus and the communication interface, and the processor implements the steps of any of the above methods when executing the computer program.

[0015] According to another aspect of an embodiment of the present application, the present application also provides a computer-readable medium having a non-volatile program code executable by a processor, wherein the program code enables the processor to execute any of the above methods.

[0016] The technical solution of the present application can be applied to the design of natural language processing using deep learning technology.

[0017] The above technical solution provided by the embodiment of the present application has the following advantages compared with the related art:

[0018] The present application provides an address recommendation method, including: obtaining a comparison address and an address to be queried input by a target object, wherein the comparison address is a delivery address of a historical order, and the address to be queried is a delivery address to be queried; inputting the address to be queried and the comparison address into a preset address matching model for matching to obtain a matching result; determining a target address that meets preset conditions from the comparison address according to the matching result, and recommending the target address to the target object. The present application obtains an address matching model by training the delivery address of the historical order as a training sample, and completes the matching of the delivery address. The present application can perform address clustering based only on the delivery information of the express delivery, and recommend addresses to the target object. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0021] Figure 1 A flowchart of an optional address recommendation method provided according to an embodiment of the present application;

[0022] Figure 2 A flowchart of an optional address matching model training method provided according to an embodiment of the present application;

[0023] Figure 3 A block diagram of an optional address recommendation device provided according to an embodiment of the present application;

[0024] Figure 4 A schematic diagram of an optional electronic device structure provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0026] In the subsequent description, the suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of the present application and have no specific meaning. Therefore, "module" and "component" can be used interchangeably.

[0027] In the related technology, there are two relatively classic clustering methods: one typical method is the partition clustering algorithm, which uses the center of a cluster to represent a cluster, that is, the cluster point selected in the iterative process is not necessarily a point in the cluster. Its purpose is to minimize the sum of squared errors SSE (Sum of Squared Error) between the data points in each cluster (a total of k) and the centroid of the cluster, which is also the evaluation criterion for evaluating the final clustering effect of the K-means algorithm; another typical method is the density-based clustering algorithm, which uses spatial indexing technology to search the neighborhood of the object, introduces concepts such as "core object" and "density reachable", and starts from the core object to form a cluster of all density-reachable objects. Simply put, it is an algorithm based on a process that continuously expands according to the density of the object. The density of an object can be judged by the number of objects close to the object.

[0028] However, the above clustering methods are all based on the fact that address data contains longitude and latitude information and some address label data. Usually, the address information in express orders is not standardized address expression information, and does not carry any longitude and latitude information. Therefore, it is difficult for couriers to determine a standard address during the actual delivery process. Without obtaining longitude and latitude information or address label data in advance, it is impossible to perform address clustering and make address recommendations to users.

[0029] In order to solve the problem mentioned in the background technology, according to one aspect of the embodiment of the present application, as Figure 1 As shown, the present application provides an address recommendation method, comprising:

[0030] Step 11, obtaining the comparison address and the address to be queried input by the target object, wherein the comparison address is the delivery address of the historical order, and the address to be queried is the delivery address to be queried;

[0031] Step 12, inputting the address to be queried and the comparison address into a preset address matching model for matching to obtain a matching result;

[0032] Step 13, determining a target address that meets preset conditions from the compared addresses according to the matching results, and recommending the target address to the target object.

[0033] Specifically, in step 11, a comparison address from historical orders is obtained as a matching object, and the address to be queried input by the target object is obtained at the same time, and then step 12 is executed to input the comparison address and the address to be queried into the address matching model. The address matching model has been pre-trained. In step 13, the address matching model can determine at least one target address with a higher matching degree of the query address in the comparison address and recommend it to the target object.

[0034] This solution uses the delivery address information of historical orders as training samples to train an address matching model. By inputting the address to be queried, the matching of the address to be queried can be completed and the comparison address with a higher matching degree can be used as the target address to recommend the address to the target object.

[0035] As an optional embodiment, Figure 2 As shown, before the address to be queried and the comparison address are input into a preset address matching model for matching, the method further includes obtaining the address matching model in the following manner:

[0036] Step 21, obtaining a sample set, wherein the sample set includes a training sample set and a verification set, and both the training sample set and the verification set are pre-processed delivery addresses;

[0037] Step 22, input the training sample set into the pre-training model for training to obtain an intermediate model;

[0038] Step 23, verifying the intermediate model using the verification set, and determining the intermediate model as the address matching model when the recognition accuracy of the intermediate model for the verification set reaches a target threshold;

[0039] Step 24, when the recognition accuracy of the intermediate model for the verification set does not reach the target threshold, continue to train the intermediate model using the training sample set until the recognition accuracy of the intermediate model for the verification set reaches the target threshold, and then determine the intermediate model as the address matching model.

[0040] For example, the embodiment of the present application can implement a supervised address matching classification model by constructing a Cross-Encoder model. The Cross-Encoder model inputs two sentences into the model together and can directly output the semantic consistency scores of the two sentences.

[0041] Specifically, the sample set obtained in step 21 includes a training sample set and a verification set. Both the training sample set and the verification set are obtained by pre-processing the delivery addresses of historical orders. The processing method includes dividing them into a training sample set and a verification set and determining the matching degree between the comparison addresses. In step 22, the training sample set is input into a pre-trained model, and the pre-trained model is trained on a large scale to obtain an intermediate model. The training method is to match the delivery addresses of the training sample set to obtain a matching result, wherein the matching result represents the matching degree between the trained delivery addresses. In step 23, the intermediate model is verified using the verification set. The verification method is the same as the above-mentioned training method. When the recognition accuracy of the intermediate model for the verification set reaches the target threshold, the intermediate model is determined as an address matching model, that is, when the error value between the matching result and the predetermined matching degree is within an acceptable range, the intermediate model is determined as an address matching model.

[0042] Optionally, the target threshold is a target matching range generated by combining a predetermined matching degree and a maximum allowable error. For example, the predetermined matching degree of address A and address B is 0.56, and the maximum allowable error is ±0.08, then the target threshold of address A and address B is 0.48-0.64, that is, when the matching result of the intermediate model for address A and address B falls between 0.48 and 0.64, the intermediate model is determined as the address matching model.

[0043] Optionally, when the recognition accuracy of the intermediate model for the verification set does not reach the target threshold, execute step 24 and continue to use the training sample set to train the intermediate model. Whenever the training of the intermediate model is completed after the training sample set is input, the intermediate model is verified using the verification set until the recognition accuracy of the intermediate model for the verification set reaches the target threshold, and the intermediate model is determined as the address matching model.

[0044] As an optional embodiment, when the recognition accuracy of the intermediate model for the verification set does not reach the target threshold, continuing to use the training sample set to train the intermediate model includes: determining the loss value of the objective function based on the recognition results of the intermediate model for the training sample set and the verification set; continuing to use the training sample set to train the intermediate model, and adjusting the internal parameters of the intermediate model according to the loss value during the training process until the recognition accuracy of the intermediate model for the verification set reaches the target threshold, and determining the intermediate model as an address matching model.

[0045] For example, the loss value of the objective function is determined based on the recognition results of the intermediate model for the training sample set and the validation set, where the objective function can be the difference between a predetermined matching degree and a matching degree obtained by the intermediate model. When the absolute value of the loss value of the objective function is not higher than the absolute value of the maximum allowable error, the intermediate model is determined as an address matching model; when the absolute value of the loss value of the objective function is higher than the absolute value of the maximum allowable error, the internal parameters of the intermediate model are adjusted according to the loss value and training continues.

[0046] Specifically, the internal parameter optimization includes the adjustment of the probability interval, the number of training times, the sample size of the training batch, the vector dimension setting, and the learning rate setting. The adjustment of the probability interval will be explained below and will not be repeated here.

[0047] For example, the number of training times can be set using the early stopping method, which is to calculate the performance of the model on the validation set during training. When the performance of the model on the validation set begins to decline, stop training. This can avoid the problem of overfitting of the model due to continued training, and also reduce unnecessary verification. For example, during the training process, the loss value keeps decreasing. When the loss value starts to increase after a certain moment, stop training and retain the model parameter value at that moment.

[0048] For example, the sample size of the training batch is adjusted according to the model training process. Too many samples mean that the time required for one training will be relatively long. When the difference between the matching result and the actual matching degree is small, it means that the recognition accuracy of the model is close to the address matching model, and the sample size of the training can be appropriately reduced. On the contrary, when the difference between the matching result and the actual matching degree is large, it means that the recognition accuracy of the model is quite different from the address matching model, and the sample size of the training can be appropriately increased. For example, when the loss value is small, 50 samples are set for training, and when the loss value is large, 100 samples are set for training.

[0049] For example, during the training process, the vector dimension is set according to the training results. The dimension ranges from dozens of dimensions to thousands of dimensions. Generally, there is a dimension with the best effect in the middle. Too low a dimension indicates insufficient representation capability, and too high a dimension is prone to overfitting. The vector dimension of this application can be 64 and 128.

[0050] For example, during the training process, the learning rate is set according to the training results. The accuracy rate determines the accuracy of the model. If it is set too small, the convergence is slow, and it is also possible to converge to a local optimal solution. If it is set too large, it will cause ups and downs, or even fail to converge. In the present application, the learning rate example can be set between 0.001 and 0.004.

[0051] As an optional embodiment, obtaining a sample set includes: obtaining the address area and delivery person data corresponding to the delivery address; dividing the delivery addresses with the same address area and the same delivery person data into a first sample set, dividing the delivery addresses with the same address area and different delivery person data into a second sample set, dividing the delivery addresses with different address areas and the same delivery person data into a third sample set, and dividing the delivery addresses with different address areas and different delivery person data into a fourth sample set.

[0052] Specifically, the delivery address of the sample set obtained includes the address area corresponding to the delivery address and the delivery person data. The address area can be extracted according to the delivery address, and the specific extraction level is determined according to the size of the matching range. For example, if the matching range is the entire Zhejiang Province, then the address area can be terminated to the district level (province-city-district). If the matching range is Hangzhou City, Zhejiang Province, then the address area can be terminated to the street level (city-district-street). In actual situations, the regional range can be determined according to actual conditions. For example, the municipality directly under the central government directly uses the city as the maximum range, so the level is appropriately reduced. The delivery person data can be any data of name, number (one person one number) or nickname. At the same time, in order to prevent interference caused by duplicate names, "name + number", "petite + name + number" or "petite + number" can be used, or the number (one person one number) can be used directly.

[0053] For example, address matching is largely based on the fact that the delivery area of ​​the same courier is similar (assuming that each courier can only deliver within a small area belonging to him, which is actually the way it is divided). Then, the courier can be fully utilized to aggregate a batch of delivery addresses, and these comparison addresses are the same sample set.

[0054] For example, the matching degree between the delivery addresses with the same address area and the same delivery person data is the highest, and the matching degree between the delivery addresses with different address areas and different delivery person data is the lowest.

[0055] As an optional embodiment, the internal parameters include a first probability interval, a second probability interval, a third probability interval and a fourth probability interval; wherein the first probability interval is the first initial parameter of the pre-trained model corresponding to the first sample set, the second probability interval is the second initial parameter of the pre-trained model corresponding to the second sample set, the third probability interval is the third initial parameter of the pre-trained model corresponding to the third sample set, and the fourth probability interval is the fourth initial parameter of the pre-trained model corresponding to the fourth sample set.

[0056] Specifically, based on the above-mentioned aggregation method of address area and deliveryman data, the initial parameters of the pre-trained model can be determined. The first initial parameter is the first probability interval corresponding to the first sample set, the second initial parameter is the second probability interval corresponding to the second sample set, the third initial parameter is the third probability interval corresponding to the third sample set, and the fourth initial parameter is the fourth probability interval corresponding to the fourth sample set. According to the correlation, the first probability interval with the same address area and the same deliveryman data has the highest matching degree, followed by the second probability interval with the same address area and different deliveryman data, followed by the third probability interval with different address areas and the same deliveryman data, and finally the fourth probability interval with different address areas and different deliveryman data. For example, the first probability interval is 0.6 to 1.0, the second probability interval is 0.5 to 0.6, the third probability interval is 0.4 to 0.5, and the fourth probability interval is 0.1 to 0.4.

[0057] As an optional embodiment, the training sample set is input into a pre-trained model for training to obtain an intermediate model, including: inputting the first sample set, the second sample set, the third sample set and the fourth sample set into the pre-trained model for training to adjust the values ​​of the first probability interval, the second probability interval, the third probability interval and the fourth probability interval until the first probability interval, the second probability interval, the third probability interval and the fourth probability interval remain unchanged, thereby obtaining an intermediate model.

[0058] Specifically, each sample set is input into the pre-training model for training, and the corresponding probability interval will be adjusted. That is to say, the first initial parameter, the second initial parameter, the third initial parameter and the fourth initial parameter of the pre-training model will change during the pre-training process. When the above parameters are adjusted to be stable and unchanged, an intermediate model is obtained.

[0059] This application uses the address information and delivery person information of historical orders as sample information for large-scale pre-training, and based on this method, constructs sample data for address matching by giving a probability distribution, and obtains an intermediate model for address matching after pre-training. The intermediate model can be trained based only on the information of the delivery address itself.

[0060] As an optional embodiment, determining a target address that meets preset conditions from the comparison addresses according to the matching results, and recommending the target address to the target object includes: determining a probability interval in which the matching result is located, wherein the probability interval is used to represent the degree of matching between the comparison address and the address to be queried; when the matching result is in the target probability interval, determining the target address according to the scores of each comparison address in the matching result; and recommending the target address to the target object.

[0061] Specifically, the target object inputs the address to be queried into the address matching model to obtain a matching result. The address matching model will identify the matching degree between the address to be queried and the comparison address, and divide the comparison address according to the probability interval where the matching degree is located. When the matching result is in the target probability interval, the target address is determined according to the score of each comparison address in the matching result and the target address is recommended to the target object, where the score represents the matching degree.

[0062] Specifically, the target probability intervals are the first probability interval, the second probability interval and the third probability interval of the address matching model, and the priority is that the first probability interval is higher than the second probability interval and higher than the third probability interval.

[0063] Optionally, when all matching results are outside the target probability interval, that is, there is no matching result in any of the first probability interval, the second probability interval and the third probability interval, the comparison address in the fourth probability interval is directly recommended to the target object as the target address.

[0064] Optionally, when all matching results are outside the target probability interval, a display module may be added to prompt the target object that there is no query result, and the target user may re-enter other addresses to be queried.

[0065] As an optional embodiment, when the matching result is within the target probability interval, determining the target address according to the score of each comparison address in the matching result includes: sorting each comparison address in the matching result according to the degree of match with the address to be queried to obtain a sorting result, wherein the score includes the degree of match; and determining the comparison address in the sorting result that is before the target sorting position as the target address.

[0066] For example, each comparison address in the matching result is sorted according to the matching degree with the address to be queried to obtain a sorting result. The sorting method can be ascending or descending.

[0067] Specifically, the number of target addresses can be set according to demand, and the target sorting position is determined by the number of target addresses plus 1. For example, when sorting in descending order by score, if the target object needs N (10) recommended addresses, then the 10 comparison addresses before the N+1 (11)th position are selected as target addresses and recommended to the target object.

[0068] According to another aspect of the embodiment of the present application, Figure 3 As shown, the present application provides an address recommendation device, including: an acquisition module 31, used to acquire a comparison address and an address to be queried input by a target object, wherein the comparison address is a delivery address of a historical order, and the address to be queried is a delivery address to be queried; a matching module 32, used to match the address to be queried and the comparison address input into a preset address matching model to obtain a matching result; a determination module 33, used to determine a target address that meets preset conditions from the comparison address according to the matching result, and recommend the target address to the target object.

[0069] It should be noted that the acquisition module 31 in this embodiment can be used to execute step 11 in the embodiment of the present application, the matching module 32 in this embodiment can be used to execute step 12 in the embodiment of the present application, and the determination module 33 in this embodiment can be used to execute step 13 in the embodiment of the present application.

[0070] Optionally, the address recommendation device also includes an address matching model acquisition module, which is used to: obtain a sample set before inputting the address to be queried and the comparison address into a preset address matching model for matching, wherein the sample set includes a training sample set and a verification set, and the training sample set and the verification set are both pre-processed delivery addresses; input the training sample set into a pre-trained model for training to obtain an intermediate model; verify the intermediate model using the verification set, and when the recognition accuracy of the intermediate model for the verification set reaches a target threshold, determine the intermediate model as the address matching model; when the recognition accuracy of the intermediate model for the verification set does not reach the target threshold, continue to use the training sample set to train the intermediate model until the recognition accuracy of the intermediate model for the verification set reaches the target threshold, and then determine the intermediate model as the address matching model.

[0071] Optionally, the address matching model acquisition module is also used to: determine the loss value of the objective function based on the recognition results of the intermediate model for the training sample set and the verification set; continue to use the training sample set to train the intermediate model, and adjust the internal parameters of the intermediate model according to the loss value during the training process until the recognition accuracy of the intermediate model for the verification set reaches the target threshold, and determine the intermediate model as the address matching model.

[0072] Optionally, the address matching model acquisition module is also used to: obtain the address area and delivery person data corresponding to the delivery address; divide the delivery addresses with the same address area and the same delivery person data into a first sample set, divide the delivery addresses with the same address area and different delivery person data into a second sample set, divide the delivery addresses with different address areas and the same delivery person data into a third sample set, and divide the delivery addresses with different address areas and different delivery person data into a fourth sample set.

[0073] Optionally, the internal parameters include a first probability interval, a second probability interval, a third probability interval and a fourth probability interval; wherein the first probability interval is the first initial parameter of the pre-trained model corresponding to the first sample set, the second probability interval is the second initial parameter of the pre-trained model corresponding to the second sample set, the third probability interval is the third initial parameter of the pre-trained model corresponding to the third sample set, and the fourth probability interval is the fourth initial parameter of the pre-trained model corresponding to the fourth sample set.

[0074] Optionally, the address matching model acquisition module is also used to: input the first sample set, the second sample set, the third sample set and the fourth sample set into the pre-trained model for training to adjust the values ​​of the first probability interval, the second probability interval, the third probability interval and the fourth probability interval until the first probability interval, the second probability interval, the third probability interval and the fourth probability interval remain unchanged, and an intermediate model is obtained.

[0075] Optionally, the determination module is specifically used to: determine the probability interval in which the matching result is located, wherein the probability interval is used to indicate the degree of matching between the comparison address and the address to be queried; when the matching result is in the target probability interval, determine the target address according to the score of each comparison address in the matching result; and recommend the target address to the target object.

[0076] Optionally, the determination module is further used to: sort each comparison address in the matching result according to the matching degree with the address to be queried to obtain a sorting result, wherein the score includes the matching degree; and determine the comparison address before the target sorting position in the sorting result as the target address.

[0077] It should be noted here that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments.

[0078] According to another aspect of the embodiment of the present application, Figure 4 As shown, the present application provides an electronic device, including a memory 61, a processor 62, a communication interface 63 and a communication bus 64, the memory 61 stores a computer program that can be run on the processor 62, the memory 61 and the processor 62 communicate through the communication bus 64 and the communication interface 63, and the processor 62 implements the steps of the above method when executing the computer program.

[0079] The memory and processor in the above electronic device communicate via a communication bus and a communication interface. The communication bus may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus may be divided into an address bus, a data bus, a control bus, etc.

[0080] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0081] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0082] According to another aspect of an embodiment of the present application, the present application also provides a computer-readable medium having a non-volatile program code executable by a processor, wherein the program code enables the processor to execute any of the above methods.

[0083] Optionally, in an embodiment of the present application, a computer-readable medium is configured to store program code for the processor to perform the following steps: obtaining a comparison address and an address to be queried input by the target object, wherein the comparison address is a delivery address of a historical order, and the address to be queried is a delivery address to be queried; matching the address to be queried and the comparison address input into a preset address matching model to obtain a matching result; determining a target address that meets preset conditions from the comparison address according to the matching result, and recommending the target address to the target object.

[0084] The present application adopts an address recommendation method, including: obtaining a comparison address and an address to be queried input by a target object, wherein the comparison address is a delivery address of a historical order, and the address to be queried is a delivery address to be queried; inputting the address to be queried and the comparison address into a preset address matching model for matching to obtain a matching result; determining a target address that meets preset conditions from the comparison address according to the matching result, and recommending the target address to the target object. The present application obtains an address matching model by training the delivery address of the historical order as a training sample, and completes the matching of the delivery address. The present application can perform address clustering based only on the delivery information of the express delivery, and recommend addresses to the target object.

[0085] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and this embodiment will not be described in detail here.

[0086] When the embodiments of the present application are specifically implemented, reference may be made to the above-mentioned embodiments, which have corresponding technical effects.

[0087] It is understood that the embodiments described herein may be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit may be implemented in one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), general purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present application, or a combination thereof.

[0088] For software implementation, the technology described herein can be implemented by a unit that performs the functions described herein. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0089] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0090] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0091] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0092] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0093] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0094] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium, including several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk or an optical disk. It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the term "include", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such a process, method, article or device. Without more constraints, an element defined by the phrase "comprising a..." does not exclude the existence of other identical elements in the process, method, article or apparatus comprising the element.

[0095] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest range consistent with the principles and novel features applied for herein.

Claims

1. An address recommendation method, characterized in that: include: Obtaining a comparison address and an address to be queried input by the target object, wherein the comparison address is the delivery address of the historical order, and the address to be queried is the delivery address to be queried; Input the address to be queried into a preset address matching model for matching to obtain a matching result; Determine a target address that meets a preset condition from the compared addresses according to the matching result, and recommend the target address to the target object; Before inputting the address to be queried into a preset address matching model for matching, the method further includes obtaining the address matching model in the following manner: obtaining a sample set, wherein the sample set includes a training sample set and a verification set, and both the training sample set and the verification set are pre-processed delivery addresses of the historical orders; inputting the training sample set into a pre-trained model for training to obtain an intermediate model; using the verification set to verify the intermediate model, and when the recognition accuracy of the intermediate model for the verification set reaches a target threshold, determining the intermediate model as the address matching model; when the recognition accuracy of the intermediate model for the verification set does not reach the target threshold, continuing to train the intermediate model using the training sample set until the recognition accuracy of the intermediate model for the verification set reaches the target threshold, determining the intermediate model as the address matching model; The acquiring of the sample set comprises: acquiring the address area and delivery person data corresponding to the delivery address of the historical order; dividing the delivery addresses with the same address area and the same delivery person data into a first sample set, dividing the delivery addresses with the same address area and different delivery person data into a second sample set, dividing the delivery addresses with different address areas and the same delivery person data into a third sample set, and dividing the delivery addresses with different address areas and different delivery person data into a fourth sample set; When the recognition accuracy of the intermediate model for the verification set does not reach the target threshold, continuing to use the training sample set to train the intermediate model includes: determining the loss value of the objective function according to the recognition results of the intermediate model for the training sample set and the verification set; continuing to use the training sample set to train the intermediate model, and adjusting the internal parameters of the intermediate model according to the loss value during the training process, until the recognition accuracy of the intermediate model for the verification set reaches the target threshold, determining the intermediate model as the address matching model; The internal parameters include a first probability interval, a second probability interval, a third probability interval and a fourth probability interval; wherein the first probability interval is a first initial parameter of the pre-trained model corresponding to the first sample set, the second probability interval is a second initial parameter of the pre-trained model corresponding to the second sample set, the third probability interval is a third initial parameter of the pre-trained model corresponding to the third sample set, and the fourth probability interval is a fourth initial parameter of the pre-trained model corresponding to the fourth sample set; The step of inputting the training sample set into a pre-trained model for training to obtain an intermediate model includes: inputting the first sample set, the second sample set, the third sample set, and the fourth sample set into the pre-trained model for training to adjust the values ​​of the first probability interval, the second probability interval, the third probability interval, and the fourth probability interval until the first probability interval, the second probability interval, the third probability interval, and the fourth probability interval remain unchanged, thereby obtaining an intermediate model.

2. The method according to claim 1, characterized in that: Determining a target address that meets a preset condition from the compared addresses according to the matching result, and recommending the target address to the target object includes: Determine a probability interval in which the matching result is located, wherein the probability interval is used to represent the matching degree between the comparison address and the address to be queried; When the matching result is within the target probability interval, determining the target address according to the score of each of the comparison addresses in the matching result; The target address is recommended to the target object.

3. The method according to claim 2, characterized in that When the matching result is within the target probability interval, determining the target address according to the score of each of the comparison addresses in the matching result includes: Sorting each of the comparison addresses in the matching result according to the matching degree with the address to be queried, to obtain a sorting result, wherein the score includes the matching degree; The comparison address before the target sorting position in the sorting result is determined as the target address.

4. An address recommendation device, implementing the address recommendation method according to any one of claims 1 to 3, characterized in that: include: An acquisition module, used to acquire a comparison address and an address to be queried input by a target object, wherein the comparison address is a delivery address of a historical order, and the address to be queried is a delivery address to be queried; A matching module, used for inputting the address to be queried into a preset address matching model for matching to obtain a matching result; A determination module is used to determine a target address that meets a preset condition from the compared addresses according to the matching result, and recommend the target address to the target object.

5. An electronic device, comprising a memory, a processor, a communication interface and a communication bus, wherein the memory stores a computer program that can be run on the processor, and the memory and the processor communicate through the communication bus and the communication interface, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 3 are implemented.

6. A computer-readable medium having a non-volatile program code executable by a processor, characterized in that: The program code enables the processor to execute the method according to any one of claims 1 to 3.

Images