Merchant traffic range generation method, information display method, device, medium and equipment

Abstract

The present application discloses a method for generating a merchant traffic range, an information display method, an apparatus, a medium and equipment, and relates to the field of Internet technology. The method includes: obtaining multiple starting point pairs between a target merchant and multiple interest areas within the capacity boundary range; estimating the preset indicator value of each starting point pair in a future preset time period through a pre-trained preset indicator estimation model, and determining a candidate starting point pair based on the estimated indicator value of each starting point pair; screening the candidate starting point pairs based on non-real-time traffic information to obtain a target starting point pair, wherein the non-real-time traffic information includes historical user demand information and merchant configuration information; generating the traffic range of the target merchant based on the target starting point pair, wherein the target merchant is visible to the user within the traffic range of the target merchant. The above method can improve the accuracy of product search or push, enhance the user experience, and avoid wasting merchant resources.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a method for generating a merchant traffic range, an information display method, an apparatus, a medium, and equipment. Background Art

[0002] Currently, on online platforms like food delivery and local life services, merchants' visibility is often directly tied to their fulfillment capabilities. For example, platforms generally adopt a "visible only if available for delivery" logic, whereby only merchants that can deliver are displayed to users, while those that cannot are uniformly filtered out.

[0003] Although the logic of "visible only when deliverable" can simplify the fulfillment process, from the user's perspective, due to geographical restrictions, they may not be able to see the merchants with which they have certain needs, resulting in unmet user needs and a damaged user experience; on the other hand, the merchant's exposure range is strictly limited to its delivery range, which will prevent its products from reaching users who really have needs, thereby wasting merchant resources. Summary of the Invention

[0004] In view of this, the embodiments of the present application provide a method for generating a merchant traffic range, an information display method, an apparatus, a medium and a device, the main purpose of which is to solve the technical problems of damaged user experience and waste of merchant resources caused by the merchant's exposure range being strictly limited to the delivery range.

[0005] According to one aspect of the present application, a method for generating a merchant traffic range is provided, the method comprising:

[0006] Within the capacity boundary, obtain multiple starting point pairs between the target merchant and multiple areas of interest;

[0007] Using a pre-trained pre-set indicator estimation model, the preset indicator value of each starting point pair in a future preset period is estimated, and candidate starting point pairs are determined based on the estimated indicator values ​​of each starting point pair;

[0008] Filtering the candidate starting point pairs based on non-real-time traffic information to obtain a target starting point pair, wherein the non-real-time traffic information includes historical user demand information and merchant configuration information;

[0009] Based on the target starting point pair, a traffic range of the target merchant is generated, wherein the target merchant is visible to the user within the traffic range of the target merchant.

[0010] Optionally, the method further includes: obtaining real-time traffic information according to a preset period, wherein the real-time traffic information includes at least one of spatiotemporal information, hotspot information, distribution resource information and manual configuration information; screening out at least one extended starting point pair within the capacity boundary based on the real-time traffic information; and adjusting the traffic range of the target merchant based on the extended starting point pair to obtain a real-time adjusted traffic range.

[0011] Optionally, the method further includes: adjusting the traffic range of the target merchant according to the traffic range corresponding to the channel where the target merchant is deployed, to obtain the traffic range of the target merchant in the channel, wherein the target merchant is deployed in at least one channel.

[0012] Optionally, the traffic range of the target merchant at least partially overlaps with the delivery range of the target merchant; then the method further includes: expanding the delivery range of the target merchant based on at least part of the target starting point pairs in the traffic range of the target merchant to obtain an expanded delivery range.

[0013] Optionally, after generating the traffic range of the target merchant based on the target starting point pair, the method further includes: storing the target starting point pairs within the traffic range of the target merchant; and / or sending the traffic range of the target merchant to the target merchant to display the traffic range of the target merchant.

[0014] Optionally, the method further includes: determining the search keywords and the user's location in response to an information search request; querying at least one candidate merchant within the traffic range of each merchant within the capacity boundary based on the search keywords and the user's location; screening the candidate merchants based on the delivery range of each candidate merchant and the user's location to obtain at least one merchant to be searched; sorting the at least one merchant to be searched based on the search keywords and user preference information, and sending the merchant information and / or product information of the sorted merchant to be searched to the user.

[0015] Optionally, the method further includes: in response to an information search request containing a demand for out-of-region delivery, determining the search keyword and the user's location; based on the search keyword and the user's location, querying at least one merchant to be searched within the traffic range of each merchant within the capacity boundary; based on the search keyword and user preference information, sorting the at least one merchant to be searched, and sending the merchant information and / or product information of the sorted merchant to be searched to the user.

[0016] Optionally, sending the sorted merchant information and / or product information of the merchants to be searched to the user includes: determining out-of-area delivery merchants based on the delivery range of each merchant to be searched and the location of the user, and marking the out-of-area delivery merchants; and sending the marked merchant information and / or product information of each merchant to be searched to the user.

[0017] Optionally, the method further includes: determining the user's location in response to an information push request, and determining the product category to be pushed based on the user preference information; querying at least one candidate merchant within the traffic range of each merchant within the capacity boundary based on the product category to be pushed and the user's location; screening the candidate merchants based on the delivery range of each candidate merchant and the user's location to obtain at least one merchant to be pushed; sorting the at least one merchant to be pushed based on the user preference information, and sending the merchant information and / or product information of the sorted merchant to be pushed to the user.

[0018] Optionally, the method further includes: in response to an out-of-region delivery demand included in the information push request, determining the user's location, and determining the product category to be pushed based on the user preference information; based on the product category to be pushed and the user's location, querying at least one merchant to be pushed within the traffic range of each merchant within the capacity boundary; based on the user preference information, sorting the at least one merchant to be pushed, and sending the merchant information and / or product information of the sorted merchant to be pushed to the user.

[0019] Optionally, sending the sorted merchant information and / or product information of the merchants to be pushed to the user includes: determining out-of-area delivery merchants based on the delivery range of each merchant to be pushed and the location of the user, and marking the out-of-area delivery merchants; and sending the marked merchant information and / or product information of each merchant to be pushed to the user.

[0020] Optionally, the training method of the preset indicator prediction model includes: within the capacity boundary range, obtaining multiple sample starting points between multiple merchants and multiple interest areas as training samples; for each of the sample starting points, using the merchant information corresponding to the sample starting point, interest area information, intersection information between merchants and interest areas, merchant information within multiple distance ranges of the interest area, and order statistics of users within the interest area at merchants within multiple distance ranges as features, and using the preset indicator value of the sample starting point in a future preset time period as a label, iteratively training the pre-constructed model; when the pre-constructed loss function reaches the preset range, stopping the model training to obtain the trained preset indicator prediction model.

[0021] Optionally, within the capacity boundary range, multiple sample starting points between multiple merchants and multiple interest areas are obtained as training samples, including: within the capacity boundary range, multiple starting point pairs are obtained between multiple merchants and multiple interest areas, and the number of orders for each of the starting point pairs is obtained; the starting point pairs whose order numbers are greater than a preset threshold are retained, and the starting point pairs whose order numbers are less than or equal to the preset threshold are sampled to obtain multiple sample starting point pairs as training samples.

[0022] Optionally, for each of the sample starting points, the merchant information, interest area information, intersection information between merchants and interest areas, merchant information within multiple distance ranges of the interest area, and order statistics of merchants within multiple distance ranges of users within the interest area corresponding to the sample starting point are used as features, including: for each of the sample starting points, respectively converting the merchant information, the interest area information and the intersection information between merchants and interest areas into merchant features, interest area features and intersection features, converting the merchant information within multiple distance ranges of the interest area into supply sequence features, and converting the order statistics of merchants within multiple distance ranges of users within the interest area into demand sequence features; performing attention processing on each merchant feature in the supply sequence features to obtain supply features, and performing attention processing on each order statistics feature in the demand sequence features to obtain demand features; and splicing or fusing the merchant features, the interest area features, the intersection features, the supply features and the demand features to obtain a feature expression of the sample starting point.

[0023] Optionally, the iterative training of the pre-constructed model includes: constructing an integrated learning model, and iteratively training the integrated learning model through the training samples to obtain an initial preset indicator estimation model as a baseline; after the training of the integrated learning model is completed, constructing a deep learning model, and expanding the training samples to expand the geographical scope covered by the training samples; based on the expanded training samples, iteratively training the deep learning model to obtain an optimized preset indicator estimation model.

[0024] Optionally, the deep learning model uses a deep neural network as the backbone network, and sequentially connects a deep factor decomposition machine module and a multi-gate hybrid expert module to the output end of the backbone network, wherein the deep factor decomposition machine module is used to learn explicit and implicit high-order cross features, and the multi-gate hybrid expert module is used to realize the multi-task prediction capability of the model.

[0025] Optionally, the deep learning model also integrates a deviation tower network and / or a generative network, wherein the deviation tower network and / or the generative network takes at least one of the city information, product category information and distance information in the training sample as input, and is used to correct the deviation values ​​of the city characteristics, product category characteristics and distance characteristics in the training sample, so that the model is simultaneously applicable to multiple different cities, product categories and distances.

[0026] According to another aspect of the present application, there is also provided an information display method, the method comprising:

[0027] In response to an information search request or an information push request initiated by a user, the merchant information and / or product information of at least one merchant is displayed, wherein the merchant is obtained by querying the traffic range of each merchant within the capacity boundary of the area where the user is located, and the traffic range of the merchant is generated based on the merchant traffic range generation method described in any of the above embodiments.

[0028] According to another aspect of the present application, a device for generating a merchant traffic range is provided, the device comprising:

[0029] A starting point pair acquisition module is used to acquire multiple starting point pairs between a target merchant and multiple interest areas within the capacity boundary;

[0030] An indicator value estimation module is used to estimate the preset indicator value of each starting point pair in a preset future period of time using a pre-trained preset indicator estimation model, and to determine candidate starting point pairs based on the estimated indicator values ​​of each starting point pair;

[0031] a starting point pair screening module, configured to screen the candidate starting point pairs based on non-real-time traffic information to obtain a target starting point pair, wherein the non-real-time traffic information includes historical user demand information and merchant configuration information;

[0032] The traffic range determination module is used to generate the traffic range of the target merchant based on the target starting point pair, wherein the target merchant is visible to the user within the traffic range of the target merchant.

[0033] According to another aspect of the present application, a storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the method for generating the above-mentioned merchant traffic range is implemented.

[0034] According to another aspect of the present application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the above-mentioned method for generating merchant traffic range when executing the program.

[0035] By means of the above technical solution, the embodiment of the present application provides a method for generating a merchant traffic range, an information display method, an apparatus, a medium and equipment. By obtaining multiple starting point pairs within the capacity boundary range and estimating the preset indicator value of each starting point pair in a preset time period in the future, a group of candidate starting point pairs that may generate orders in the future can be screened out. By screening the candidate starting point pairs through historical user demand information and merchant configuration information, a group of target starting point pairs that can meet both user needs and merchant needs can be screened out. By generating the merchant's traffic range based on the target starting point pair, and enabling users within the traffic range to search or view the corresponding merchants on the platform, the limitation of the fixed configuration delivery range on the merchant's visible range can be eliminated, so that the merchants searched or pushed by the platform can meet the user's certainty needs, thereby improving the accuracy and rationality of product search or push, and then improving the user experience. In addition, it can also enable the products of capable merchants to reach users in need, avoiding waste of merchant resources.

[0036] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0038] Figure 1 A schematic diagram showing a flow chart of a method for generating a merchant traffic range provided in an embodiment of the present application is shown;

[0039] Figure 2 A schematic diagram showing a flow chart of another method for generating a merchant traffic range provided in an embodiment of the present application is shown;

[0040] Figure 3 A schematic diagram illustrating another method for generating a merchant traffic range provided in an embodiment of the present application is shown;

[0041] Figure 4 A structural schematic diagram of a device for generating a merchant traffic range provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0042] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0043] In one embodiment, Figure 1 As shown, a method for generating a merchant traffic range is provided. The method is described by taking the application of the method to a server as an example, and includes the following steps:

[0044] Step 101: Acquire multiple starting point pairs between a target merchant and multiple interest areas within a capacity boundary.

[0045] The delivery capacity boundary refers to the maximum delivery distance within a predefined area (e.g., a city). Orders beyond this range are unavailable. For example, a city's delivery capacity boundary typically encompasses the entire city. Therefore, the delivery capacity boundary is independent of the merchant's location and is solely dependent on the predefined area's attributes (e.g., area, traffic conditions, etc.).

[0046] Furthermore, target merchants refer to businesses operating within a predefined area (such as a city). This doesn't refer to a specific merchant, but rather the merchants within the current traffic range being planned. An area of ​​interest (AOI), also known as a surface or point of interest, refers to a geographic entity within a predefined area, represented as an area on a map, such as a residential neighborhood. An origin-destination pair (OD), also known as an OD pair, refers to a combination where the merchant is the delivery origin and the user's area of ​​interest is the destination.

[0047] Specifically, when generating a traffic range for a specific merchant (i.e., the target merchant), multiple starting point pairs between the target merchant and multiple interest areas within the capacity boundary can be first obtained. The target merchant can form an starting point pair with any interest area within the capacity boundary. This starting point pair may have generated an order, or it may have never generated an order because it is outside the merchant's delivery range. Since the traffic range itself is intended to transcend the limitations of the delivery range on the merchant's visibility, multiple starting point pairs can be obtained between the merchant and all interest areas within the capacity boundary in the initial stage, including starting point pairs that have generated orders and starting point pairs that have not.

[0048] Step 102 , using a pre-trained preset indicator estimation model, estimates the preset indicator value of each starting point pair within a future preset period, and determines candidate starting point pairs based on the estimated order quantity estimation indicator value of each starting point pair.

[0049] Among them, the preset indicator prediction model can be based on a variety of model frameworks such as ensemble learning models and deep learning models, with attribute information, statistical information, portrait information and other information of multiple starting point pairs in a preset area as training samples. After iterative training, a model with reliability evaluation is obtained, which can predict the preset indicator value of any starting point pair in the future preset time period. The preset indicator value can be an indicator related to the starting point pair or a combination of multiple indicators. For example, the preset indicator value can be order volume, average order amount, merchant visit purchase rate, total order amount, a combination of order volume and average order amount, etc. This embodiment does not make specific limitations here. Furthermore, the preset time period can be set according to information such as the preset indicator value within the time period and the order refresh cycle, for example, it can be set to 15 days or 28 days.

[0050] Specifically, the preset indicator values ​​of each starting point pair obtained in step 101 within a preset future time period can be estimated using the trained preset indicator estimation model to obtain the estimated indicator values ​​of each starting point pair. Then, multiple candidate starting point pairs can be determined from the original starting point pairs based on the estimated indicator values ​​of each starting point pair. In this embodiment, the estimated indicator values ​​of each starting point can be compared with a preset threshold to determine the candidate starting point pairs, or the estimated indicator values ​​of each starting point can be sorted, and the starting points arranged in the top N can be determined as candidate starting point pairs. This embodiment does not impose any specific restrictions.

[0051] Step 103 : Screening candidate starting point pairs based on non-real-time traffic information to obtain target starting point pairs, wherein the non-real-time traffic information includes historical user demand information and merchant configuration information.

[0052] Among them, non-real-time traffic information refers to relatively stable information that will have an impact on the preset indicator values ​​in the future. Specifically, it may include historical user demand information and merchant configuration information. For example, in the future, the area where the preset area is located will enter summer. Then, based on historical user demand information, such as user search information and consumption information in the same historical period, it can be determined that users may have demand for cold drinks, desserts and snacks, seasonal products, cold dishes and other products in the future. In addition, based on merchant configuration information, such as the merchant's brand reputation, merchant level, product type, maximum order volume and other information, it can be determined to what extent the target merchant can stably provide its products.

[0053] It is understandable that at different time points, non-real-time traffic information will also change to a certain extent. For example, when the area where the preset area is located enters autumn, users' demand for various product categories will change accordingly. At the same time, merchant configuration information may also change to a certain extent due to factors such as level upgrades and product category updates. Therefore, non-real-time traffic information can only remain relatively stable for a period of time. As time goes by, the merchant's traffic range also needs to be adjusted regularly. When making adjustments, the current non-real-time traffic information can be obtained to adjust the current traffic range.

[0054] Specifically, after determining candidate starting point pairs based on the estimated index values ​​of each starting point pair, these candidate starting point pairs can be further screened based on non-real-time traffic information to obtain multiple target starting point pairs that simultaneously meet user needs and merchant supply capabilities. The screening rules (such as the information used for screening and the screening order) can be set according to actual circumstances and are not specifically limited in this embodiment. The selected target starting point pairs can not only meet the user's deterministic needs for a period of time in the future, but also ensure the merchant's order supply capabilities, thereby ensuring smooth order fulfillment.

[0055] Step 104 : generating a traffic range of a target merchant based on the target starting point pair, wherein the target merchant is visible to the user within the traffic range of the target merchant.

[0056] Specifically, after the target starting point pair is determined, the traffic range of the target merchant can be generated based on the range where the target starting point pair is located, so that the merchant's traffic range can cover the area where each target starting point is located. Among them, the target merchant is visible to users within the traffic range of the target merchant, which means that users within the traffic range can view the target merchant through various methods such as search or push, and determine whether to place an order with the target merchant, and further determine the corresponding fulfillment method, such as delivery, self-pickup, in-store, etc. It should be noted that, with reference to Figure 3 ,In a preset area, the capacity boundary range is usually larger than the flow range and delivery range of each merchant, and the flow range and delivery range of each merchant are at least partially overlapped.

[0057] In this embodiment, the traffic range of each merchant can be generated specifically according to the actual situation of the merchant. Compared with the traditional delivery range demarcation method with the merchant as the center and the delivery distance as the radius, the traffic range can not only meet the deterministic demand of users in a certain area for a certain product, but also expand the merchant's own order supply range, thereby realizing the decoupling between the merchant's visible range and the delivery range, ensuring that the generated orders can meet the needs of both users and merchants, thereby improving the user experience, and at the same time, avoiding the waste of merchant resources.

[0058] By applying the above technical solution, by obtaining multiple starting point pairs within the capacity boundary and estimating the preset indicator values ​​of each starting point pair in a preset time period in the future, a group of candidate starting point pairs that may generate orders in the future can be screened out. By screening the candidate starting point pairs through historical user demand information and merchant configuration information, a group of target starting point pairs that can meet both user and merchant needs can be screened out. By generating the merchant's traffic range based on the target starting point pair, and enabling users within the traffic range to search or view the corresponding merchants on the platform, the limitation of the fixed configuration delivery range on the merchant's visible range can be eliminated, so that the merchants searched or pushed by the platform can meet the user's certainty needs, thereby improving the accuracy and rationality of product search or push, and then improving the user experience. In addition, it can also enable the products of capable merchants to reach users in need, avoiding waste of merchant resources.

[0059] Furthermore, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, another method for generating the merchant traffic range is provided, such as Figure 2 As shown, the method includes:

[0060] Step 201 : within the capacity boundary, obtain multiple sample starting points between multiple merchants and multiple interest areas, iteratively train the pre-built model, and obtain a trained preset indicator estimation model.

[0061] Specifically, the preset indicator estimation model can be trained in the following manner: within the capacity boundary, multiple sample starting points between multiple merchants and multiple interest areas are obtained as training samples. Then, for each sample starting point, the pre-built model is iteratively trained using the merchant information corresponding to the sample starting point, interest area information, intersection information between the merchant and the interest area, merchant information within multiple distance ranges within the interest area, and order statistics of users within the interest area at merchants within multiple distance ranges. The pre-built model is then iteratively trained using the preset indicator value of the sample starting point in a preset future time period as a label. When the pre-built loss function reaches the preset range, the model training is stopped, and the trained preset indicator estimation model is obtained.

[0062] In this embodiment, multiple sampling methods can be used to obtain multiple sample starting points between multiple merchants and multiple areas of interest within the delivery capacity boundary as training samples. The starting point pairs in the training samples must cover both starting point pairs within the merchant's delivery range and starting point pairs outside the merchant's delivery range. For each sample starting point pair, information such as merchant information, area of ​​interest information, intersection information between merchants and areas of interest, merchant information within multiple distance ranges of the area of ​​interest, and order statistics for users within the area of ​​interest at merchants within multiple distance ranges can be obtained to generate feature expressions.

[0063] Merchant information can include merchant attributes (such as merchant ID, product category, and brand), order statistics (such as order volume, average order price, and visit-to-purchase rate). Area of ​​interest information can include area of ​​interest attributes (such as area ID, area type, and area), order statistics (such as order volume, average order price, visit-to-purchase rate, and number of active users). Cross-information between merchants and areas of interest can include profile information for sample starting point pairs (including product categories, order volume levels, price ranges, and distance between starting point pairs generated by the starting point pairs), and order statistics (such as order volume and average order price for the starting point pairs).

[0064] Furthermore, in terms of features, in addition to the more conventional information described above, such as merchant information, area of ​​interest information, and cross-information between merchants and areas of interest, this embodiment also focuses on characterizing supply and demand characteristics between merchants and areas of interest. This supply and demand characteristic information specifically includes merchant information within multiple distance ranges within the area of ​​interest, as well as order statistics for merchants within multiple distance ranges for users within the area of ​​interest. Within the supply and demand characteristics, by leveraging the profile information of sample starting point pairs, the similarities between different sample starting point pairs can be expressed at different granularities, and the supply and demand characteristics between merchants within the delivery area of ​​the area of ​​interest and merchants with similar characteristics outside the delivery area can be expressed. The resulting features include characteristic information such as "the number of orders from other merchants in the area of ​​interest within the same distance range, product category, and price range as the current merchant" and "the number of merchants in the area of ​​interest within the same product category as the current merchant but closer to the current merchant." In this way, even when cross-information between merchants and areas of interest is insufficient, the true characteristic information of sample starting point pairs can still be accurately characterized.

[0065] Furthermore, after obtaining the characteristic expression of the sample starting point pairs, sample labels for each sample starting point pair can be generated. In this embodiment, the training goal of the preset indicator prediction model is to find a group of starting point pairs with the largest future order increment, or the fastest growth in total order amount, or the highest growth rate in merchant visit and purchase rate within the capacity boundary. Therefore, this embodiment can select preset indicator values ​​such as order volume, total order amount, merchant visit and purchase rate within the future preset time period as the most important labels of the model. Among them, the preset time period can be set according to actual conditions. For example, the preset time period can be set to about 15 days to 30 days. By setting a suitable time period, on the one hand, it can prevent the cycle from being too short, and the preset indicator values ​​such as order volume, total order amount, merchant visit and purchase rate will be relatively sparse; on the other hand, the prediction cycle of the preset indicator values ​​related to the order can be kept at the same frequency as the refresh cycle of the order, thereby facilitating data acquisition.

[0066] In this embodiment, for sample starting point pairs with historical orders, the preset indicator values ​​for the sample starting point pairs within a certain time period can be obtained from these historical orders as sample labels. For sample starting point pairs with no historical orders or few historical orders, the preset indicator values ​​for the sample starting point pairs within a certain time period can be estimated using the supply and demand characteristics of the sample starting point pairs described above. After obtaining the feature expressions and labels for each sample starting point pair, a pre-built model (such as an ensemble learning model or a deep learning model) can be iteratively trained to obtain a trained preset indicator estimation model.

[0067] This embodiment constructs the supply and demand characteristics of the sample starting point pair as the characteristic expression of the sample, which can accurately characterize the actual situation of the sample starting point pair in terms of order supply capacity and order demand. Especially for samples that are outside the merchant's delivery range and lack orders, the supplementary supply and demand characteristics are particularly important for the characterization of the sample starting point pair, which can make the trained model more reliable and accurate.

[0068] In an optional embodiment, training samples can be constructed in the following manner: within the capacity boundary, multiple starting point pairs between multiple merchants and multiple interest areas are obtained, and the number of orders for each starting point pair is obtained. Then, the starting point pairs with an order number greater than a preset threshold are retained, and the starting point pairs with an order number less than or equal to the preset threshold are sampled to obtain multiple sample starting point pairs as training samples.

[0069] Specifically, when selecting sample starting point pairs as training samples, sample starting point pairs with orders will be relatively sparse. Therefore, all sample starting point pairs with orders or order volumes greater than a preset threshold can be retained, while sample starting point pairs without orders or order volumes less than or equal to the preset threshold can be sampled in a certain form, ultimately ensuring that at least some sample starting point pairs with orders exist in the training samples. When evaluating the effectiveness of model training, sample starting point pairs with relative distances within a certain range (such as between 3 and 7 kilometers) can be selected for random sampling and evaluation, wherein the distance between the sample starting point pairs used to evaluate the effectiveness of model training is within the same distance range as the distance between the starting point pairs that account for the majority of the model in actual application. In this embodiment, the reason for retaining close-range sample starting point pairs in the training samples is to better learn the characteristics related to merchants and areas of interest during the model training process.

[0070] This embodiment selectively constructs training samples for the model, which can not only enable the model to fully learn the characteristics related to merchants and areas of interest in the samples, but also enable the samples to cover a wider area or an area closer to the actual application of the model, thereby improving the reliability of the preset indicator estimation model, and further improving the accuracy of the model's estimation of the preset indicator value for the starting point.

[0071] In an optional embodiment, the characteristic expression of the sample can be constructed in the following manner: for each sample starting point, the merchant information, interest area information, and the intersection information between the merchant and the interest area are converted into merchant features, interest area features, and intersection features respectively; the merchant information within multiple distance ranges of the interest area is converted into supply sequence features; and the order statistics of the users within the interest area at merchants within multiple distance ranges are converted into demand sequence features. Then, attention processing is performed on each merchant feature in the supply sequence features to obtain supply features, and attention processing is performed on each order statistics feature in the demand sequence features to obtain demand features. Finally, the merchant features, interest area features, intersection features, supply features, and demand features are spliced ​​or fused to obtain the characteristic expression of the sample starting point.

[0072] In this embodiment, after obtaining multi-dimensional information about the sample starting point pair, its feature expression can be constructed using this multi-dimensional information. In this embodiment, when constructing the feature expression, each type of information can be converted into corresponding features, and then the features can be spliced ​​or fused to obtain the feature expression of the sample starting point. Alternatively, the feature expression of the sample starting point pair can be optimized by converting the supply and demand feature information into a feature sequence as described in the above embodiment.

[0073] Specifically, in the above embodiment, the supply and demand characteristics of the sample starting point pair can be described by information such as merchant information within multiple distance ranges in the area of ​​interest and order statistics of merchants within multiple distance ranges for users in the area of ​​interest. However, describing the supply and demand characteristics of the sample starting point pair only through summary statistical data such as merchant information and order statistics within multiple distance ranges will cause some detailed information in the characteristics to be lost. In order to retain as many details as possible in the supply and demand characteristic information, this embodiment introduces supply and demand sequence features, namely supply sequence features and demand sequence features, wherein the detailed features in the sequence can carry more information, thereby avoiding excessive loss of details in the supply and demand characteristics.

[0074] In this embodiment, the supply sequence feature is observed from the perspective of merchants. From their perspective, the most intuitive representation of supply is the supply of similar products around the area of ​​interest. Therefore, the information about a group of merchants in the same category and located relatively close to the area of ​​interest can be converted into a supply sequence feature. Correspondingly, the demand sequence feature is observed from the perspective of the area of ​​interest. From the user's perspective, the most intuitive representation of demand is order statistics such as order volume and total order amount. Therefore, the information about a group of merchants with high order volume or the highest total order amount around the area of ​​interest can be converted into a demand sequence feature. In this embodiment, the length of the supply sequence feature and the demand sequence feature, as well as the features contained within the sequence features, can be pre-set. For example, the length of the sequence feature can be set to 100. The sequence feature can include the merchant identifier, basic merchant attribute information (such as store opening time, price range, brand, product category, etc.), the merchant's distance from the area of ​​interest, the merchant's monthly order volume, the average order amount of the merchant in the area of ​​interest, the number of searches for the merchant in the area of ​​interest, and so on. This embodiment does not limit this.

[0075] It should be noted that the internal features of the supply sequence features and the demand sequence features are highly similar. Essentially, both describe the supply and demand around the region of interest, but each focuses on a different aspect. When using sequence features, an attention mechanism can be used to identify the most important parts of the supply and demand sequence features for the prediction target (i.e., target attention). The weights of the supply and demand features for merchants at each sample starting point are then calculated. Finally, the characteristic representation of the sample starting point can be obtained by concatenating or fusing the two sequence features with the converted merchant features, the region of interest features, and the cross features.

[0076] This embodiment converts supply and demand features into supply and demand sequence features, performs attention processing on the supply and demand sequence features, and then splices the supply and demand sequence features with features of other dimensions into the feature expression of the sample starting point. This can retain more detailed information in the supply and demand features, thereby improving the accuracy of the feature expression of the sample starting point, and further improving the overall reliability and accuracy of the preset indicator estimation model.

[0077] In an optional embodiment, the preset indicator prediction model can be trained in the following manner: in the initial stage, an integrated learning model is constructed, and the integrated learning model is iteratively trained through training samples to obtain an initial preset indicator prediction model as a baseline. After the integrated learning model training is completed, a deep learning model is constructed, and the training samples are expanded to expand the geographical scope covered by the training samples, and the deep learning model is iteratively trained based on the expanded training samples to obtain an optimized preset indicator prediction model.

[0078] Specifically, after constructing sample features, to more quickly verify the model's effectiveness, you can build an ensemble learning model as a baseline at the beginning of model training. For example, you can build an xgboost regression model to quickly verify the model's performance. Subsequently, as samples are continuously generated, you can observe that as the distance between starting point pairs increases, the number of starting point pairs increases exponentially, but the number of starting point pairs with orders decreases. Based on this, to support the learning of large amounts of sample data and more flexible modeling methods, the model can be iterated from ensemble learning models to deep learning models to obtain a better-performing preset indicator estimation model.

[0079] In this embodiment, during the model training process, the models at each stage can be evaluated offline and online. Among them, the order recall volume of the top-k starting point pairs can be used as an evaluation indicator during offline evaluation; during actual online evaluation, the ratio of orders outside the delivery range to all orders can be used as an evaluation indicator to evaluate the operating effect of the model. After evaluation, by combining the deep learning model with the supply and demand sequence characteristics, the order recall rate of the top 5% OD in offline testing will be improved compared to using only the deep learning model. It can be seen that the model structure and feature expression are constantly improving in the process of continuous training and iteration, and the model performance is also constantly improving.

[0080] This embodiment can improve the performance of the preset indicator estimation model by continuously expanding the number of samples during the model training process and updating the model structure while expanding the samples, thereby improving its estimation accuracy.

[0081] In an optional embodiment, the deep learning model uses a deep neural network as the backbone network, and sequentially connects a deep factor decomposition machine module and a multi-gate hybrid expert module to the output end of the backbone network, wherein the deep factor decomposition machine module is used to learn explicit and implicit high-order cross features, and the multi-gate hybrid expert module is used to realize the multi-task prediction capability of the model.

[0082] Specifically, the backbone network of the deep learning model can use a deep neural network (DNN). During the initial training phase of the deep learning model, offline evaluation revealed that its performance was no better than that of the ensemble learning model. Considering that ensemble learning models often involve more feature crosstalk, this embodiment introduces a deep factorization machine (xDeepFM) module and a multi-gate mixture-of-experts (MMoE) module into the deep learning model. The xDeepFM module explicitly captures high-order interactions between features while preserving low-order feature interactions, thereby improving model performance. The xDeepFM module automatically learns complex feature combinations without requiring extensive manual feature engineering, enhancing feature crosstalk and thus achieving better model performance in offline evaluations. Furthermore, the MMoE module enables different tasks to utilize different expert knowledge as needed, thereby more effectively capturing the correlations and differences between tasks. By introducing a multi-gate hybrid expert module into the deep learning model, multiple related tasks can be solved simultaneously, and the overall performance of the model can be improved by sharing information between tasks, thereby improving the model's multi-task prediction ability.

[0083] In an optional embodiment, a deviation tower network and / or a generative network are also integrated into the deep learning model, wherein the deviation tower network and / or the generative network take at least one of the city information, product category information and distance information in the training sample as input, and are used to correct the deviation values ​​of the city characteristics, product category characteristics and distance characteristics in the training sample, so that the model is simultaneously applicable to multiple different cities, product categories and distances.

[0084] Specifically, to make the trained pre-set indicator estimation model applicable to multiple cities, the sample can be supplemented with information about the city where the starting point is located, such as information about the city's supply and consumption preferences. A bias tower network can then be added to the model. This bias tower network then processes information such as city, distance, and product category to capture and address bias in the data, making the model applicable to multiple cities and scenarios. In this embodiment, the bias tower network can serve as an independent component to correct for bias in the data, allowing the main task network to focus more on learning more useful feature interactions and complex patterns. This approach can improve the overall performance and accuracy of the model. Furthermore, a generative network (PGN, Pointer Generator Network) can be used instead of the bias tower network structure to adaptively generate appropriate parameters for different cities, categories, and distances, avoiding the complex and redundant work of training and maintaining multiple models due to differences in urban scenarios.

[0085] Step 202: Acquire multiple starting point pairs between a target merchant and multiple interest areas within the capacity boundary.

[0086] Specifically, after the training of the preset indicator estimation model is completed, a corresponding traffic range can be generated for each merchant within the capacity boundary. When generating a traffic range for a certain merchant (i.e., the target merchant), multiple starting point pairs between the target merchant and multiple interest areas can be obtained within the capacity boundary. Among them, the target merchant can form a starting point pair with any interest area within the capacity boundary. This starting point pair may have generated an order, or may have never generated an order because it exceeds the merchant's delivery range. Since the traffic range itself is to break through the restrictions of the delivery range on the merchant's visible range, therefore, in the initial stage, multiple starting point pairs can be obtained between the merchant and all interest areas within the capacity boundary, including starting point pairs that have generated orders and starting point pairs that have not generated orders.

[0087] Step 203 : Preset indicator values ​​of each starting point pair in a future preset period are estimated using a pre-trained preset indicator estimation model, and candidate starting point pairs are determined based on the estimated indicator values ​​of each starting point pair.

[0088] Specifically, through the trained preset indicator estimation model, the preset indicator value of each starting point pair in the future preset time period can be estimated to obtain the estimated indicator value of each starting point pair. Then, based on the estimated indicator value of each starting point pair, multiple candidate starting point pairs can be determined from the original starting point pairs. In this embodiment, the estimated indicator value of each starting point can be compared with a preset threshold to determine the candidate starting point pair, or the estimated indicator value of each starting point can be sorted, and the starting points arranged in the topN can be determined as candidate starting point pairs. This embodiment does not impose specific restrictions.

[0089] In this embodiment, by using the preset indicator prediction model trained in step 201 to estimate the preset indicator values ​​of each starting point in the future preset time period, the traffic range can be expanded to the area with the largest future order increment, the fastest growth in total order amount, or the highest growth rate in merchant visit and purchase rate, thereby avoiding the problem of increased transportation cost caused by indiscriminate expansion of the distribution range. At the same time, it can also avoid the problem of waste of resources and transportation capacity caused by sparse orders in distant areas.

[0090] Step 204 : Screen candidate starting point pairs based on non-real-time traffic information to obtain target starting point pairs, wherein the non-real-time traffic information includes historical user demand information and merchant configuration information.

[0091] Specifically, after determining candidate starting point pairs based on the estimated index values ​​of each starting point pair, these candidate starting point pairs can be further screened based on non-real-time traffic information to obtain multiple target starting point pairs that simultaneously meet user needs and merchant supply capabilities. The screening rules (such as the information used for screening and the screening order) can be set according to actual circumstances and are not specifically limited in this embodiment. The selected target starting point pairs can not only meet the user's deterministic needs for a period of time in the future, but also ensure the merchant's order supply capabilities, thereby ensuring smooth order fulfillment.

[0092] Step 205 : generating a traffic range of a target merchant based on the target starting point pair, wherein the target merchant is visible to the user within the traffic range of the target merchant.

[0093] Specifically, after the target starting point pair is determined, the traffic range of the target merchant can be generated based on the range where the target starting point pair is located, so that the merchant's traffic range can cover the area where each target starting point is located. Among them, the target merchant is visible to users within the traffic range of the target merchant, which means that users within the traffic range can view the target merchant through various methods such as search or push, and determine whether to place an order with the target merchant, and further determine the corresponding fulfillment method, such as delivery, self-pickup, in-store, etc. It should be noted that, with reference to Figure 3,In a preset area, the capacity boundary range is usually larger than the flow range and delivery range of each merchant, and the flow range and delivery range of each merchant are at least partially overlapped.

[0094] Step 206: storing the target starting point pairs within the traffic range of the target merchant, and / or sending the traffic range of the target merchant to the target merchant to display the traffic range of the target merchant.

[0095] Specifically, after the traffic range of the target merchant is generated, the target starting point within the traffic range of the target merchant can be stored, that is, the target merchant and the traffic range of the target merchant are mapped and stored. In this way, it is convenient for merchants or operators to query the current traffic range of the target merchant. In addition, the traffic range of the target merchant can also be pushed to the target merchant, or in response to a viewing request initiated by the merchant, the traffic range can be pushed or displayed to the target merchant, so that the target merchant can view the area and range covered by the product, so that it can adjust its own product sales strategy in a targeted manner and avoid wasting merchant resources.

[0096] Step 207: periodically obtain real-time traffic information, and expand or adjust the traffic range of the target merchant based on the real-time traffic information to obtain a real-time adjusted traffic range.

[0097] Specifically, after the traffic range for a target merchant is established, it can be adjusted in real time based on real-time traffic information. Real-time traffic information refers to real-time, changing information that affects current order volume. This information can include temporal and spatial information, hotspot information, delivery resource information, and manually configured information. For example, if the current time period is morning, based on the current temporal and spatial information, it can be determined that users are currently demanding items such as breakfast, coffee, and beverages. In this case, relatively distant origin pairs can be adjusted to fall within the traffic range of the corresponding merchant, resulting in a real-time adjusted traffic range. Similarly, if it is Valentine's Day, distant origin pairs can be expanded to fall within the traffic range of merchants selling flowers. If it is peak delivery time and delivery resources are limited, some distant origin pairs will be temporarily excluded from the merchant's traffic range. If operators adjust a certain origin pair to fall within the traffic range of a merchant, the configured origin pair can be added to the corresponding merchant's traffic range.

[0098] In this embodiment, for any merchant (i.e., the target merchant), its traffic range can be periodically adjusted based on either non-real-time traffic information or real-time information. The difference between the two is that adjustments to a merchant's traffic range based on non-real-time traffic information occur over a longer period (e.g., 15 to 30 days), and the adjusted traffic range can be stored or displayed to the merchant, resulting in a relatively stable range. Adjustments to a merchant's traffic range based on real-time traffic information occur over a shorter period (e.g., 15 to 30 minutes), and the adjusted traffic range is typically not stored or displayed to the merchant. The range changes rapidly but not significantly. Furthermore, whether adjustments are made based on non-real-time or real-time traffic information, adjustments are made only within the capacity boundaries of the merchant's area and do not exceed those boundaries.

[0099] Step 208: Adjust the traffic range of the target merchant according to the traffic range corresponding to the channel where the target merchant is placed, and obtain the traffic range of the target merchant in each channel.

[0100] Specifically, currently, the instant delivery feature can be implemented not only within dedicated applications but also through mini-programs, web pages, and other methods within various applications (i.e., delivery channels) such as instant messaging, mapping, and payment apps. However, different applications (i.e., delivery channels) correspond to different traffic ranges. Once the traffic range for a target merchant is determined, the merchant's traffic range can be adaptively adjusted to suit the merchant's delivery channels, ensuring that the adjusted traffic range is compatible with the merchant's current delivery channels. In this embodiment, the target merchant has at least one delivery channel, meaning that the target merchant can be displayed or pushed within at least one application.

[0101] Step 209 : Expand the delivery range of the target merchant based on at least some of the target starting point pairs in the traffic range of the target merchant to obtain an expanded delivery range.

[0102] Specifically, after the traffic range of the target merchant is determined, the user's visible range of the merchant will change. Merchants that users could not see before can now be viewed by users through search or push, etc. Therefore, the order coverage of the user's interest area will also change. If the visit rate of users in a certain interest area to the target merchant exceeds a certain threshold, these starting points can be adjusted to the delivery range of the target merchant. In this way, users in the corresponding area will find it easier to view the corresponding merchant, and after the delivery range is adjusted, the transportation resources will also be adjusted accordingly, which can make the order delivery in the corresponding area smoother. Figure 3,For any merchant, its traffic range and delivery range at least partially overlap, and both the traffic range and delivery range are smaller than the capacity boundary range of the merchant's area.

[0103] Step 210 , in response to the information search request, query the merchants to be searched within the traffic range of each merchant within the capacity boundary, and send the merchant information and / or product information of the merchants to be searched to the user.

[0104] Specifically, after the merchant's traffic range and delivery range are determined, when the server receives an information search request sent by the client, it can respond to the information search request, first determine the user's search keyword and the user's location, and then based on the search keyword and the user's location, query at least one candidate merchant within the traffic range of each merchant within the capacity boundary, and then filter the candidate merchants based on the delivery range of each candidate merchant and the user's location to obtain at least one merchant to be searched, and finally sort the at least one merchant to be searched based on the search keyword and user preference information, and send the merchant information and / or product information of the sorted merchant to the user.

[0105] In this embodiment, in response to receiving an information search request, the server may first search for corresponding merchants within the traffic range of each merchant to determine at least one candidate merchant. At this time, the starting point pair formed by the user's interest area and the candidate merchant is within the traffic range of the candidate merchant. After the candidate merchant is determined, the candidate merchants can be screened by the delivery range to obtain at least one merchant to be searched. In this embodiment, the interest area of ​​the user who initiated the information search request is within both the traffic range and the delivery range of the merchant to be searched, that is, the merchant to be searched is the result of the intersection of the traffic range and the delivery range of each merchant. In this way, it can be ensured that the displayed merchants to be searched meet the user's needs and the merchant's supply capacity, and it can also be ensured that the order is within the transportation coverage, thereby ensuring the sequential execution of the order, thereby improving the user experience and avoiding the waste of merchant resources.

[0106] In an optional embodiment, the merchants to be searched can also be determined in the following manner: first, in response to the information search request containing out-of-region delivery requirements, the search keywords and the user's location are determined; then, based on the search keywords and the user's location, at least one merchant to be searched is queried within the traffic range of each merchant within the capacity boundary; finally, based on the search keywords and user preference information, at least one merchant to be searched is sorted, and the merchant information and / or product information of the sorted merchant to be searched is sent to the user.

[0107] In this embodiment, the out-of-area delivery requirement can be selected by the user when initiating an information search request, pre-configured by the user in the application software, or automatically executed by the application software when operating in a certain mode. This embodiment does not impose any specific limitations. In this scenario, the server can respond to the information search request by directly searching for merchants within the traffic range of each merchant and push the found merchants to the user. In this way, users can place orders for products outside their delivery range, thereby meeting their need for certainty and expanding the merchant's order coverage, thereby improving the merchant's resource utilization.

[0108] Furthermore, the information search method proposed in this embodiment can also be combined with the information search method proposed in the previous embodiment. For example, if the estimated index value of the candidate merchant screened out within the merchant's traffic range is greater than the preset threshold, it will be directly determined as the merchant to be searched; otherwise, it will be further determined whether the user's interest area is within the delivery range of the candidate merchant. If it is within the delivery range, it will be determined as the merchant to be searched, otherwise, it will be screened out. In this way, a certain proportion of merchants with high order volumes or high total order amounts but exceeding the delivery range can be retained and pushed to users, thereby meeting the user's certainty needs (that is, when the user explicitly searches for the merchant, the merchant will not be displayed because it exceeds the delivery range). At the same time, the merchant's order range can be increased to avoid wasting merchant resources.

[0109] In an optional embodiment, the merchants to be searched can also be processed as follows: based on the delivery range of each merchant to be searched and the user's location, out-of-area delivery merchants are determined, and the out-of-area delivery merchants are marked, and the marked merchant information and / or product information of each merchant to be searched is sent to the user.

[0110] In this embodiment, when the delivery range of a merchant being searched for exceeds the delivery range of the user's area of ​​interest, the server can identify and mark such merchants and push the marked results to the user, so that the user can view out-of-area delivery information when browsing merchant information, such as viewing an out-of-area delivery icon or prompt information. In this way, users can be forewarned that a merchant has exceeded the delivery range and can choose whether to place an order based on their needs, thereby improving product selection efficiency.

[0111] Step 211 , in response to the information push request, query the merchants to be pushed within the traffic range of each merchant within the capacity boundary, and send the merchant information and / or product information of the merchants to be pushed to the user.

[0112] Specifically, when the server receives an information push request sent by the client, it can respond to the information push request, first determine the user's location, and determine the product category to be pushed based on the user's preference information, and then, based on the product category to be pushed and the user's location, query at least one candidate merchant within the traffic range of each merchant within the capacity boundary, and then screen the candidate merchants based on the delivery range of each candidate merchant and the user's location to obtain at least one merchant to be pushed, and finally, sort the at least one merchant to be pushed based on the user's preference information, and send the merchant information and / or product information of the sorted merchant to the user.

[0113] In an optional embodiment, the merchants to be pushed can also be determined in the following manner: in response to the out-of-zone delivery requirement included in the information push request, the user's location is determined, and the category of goods to be pushed is determined based on the user's preference information, and then, based on the category of goods to be pushed and the user's location, at least one merchant to be pushed is queried within the traffic range of each merchant within the capacity boundary, and finally, based on the user's preference information, at least one merchant to be pushed is sorted, and the merchant information and / or product information of the sorted merchant to be pushed is sent to the user.

[0114] In an optional embodiment, the merchants to be pushed can also be processed as follows: based on the delivery range of each merchant to be pushed and the user's location, out-of-area delivery merchants are determined and the out-of-area delivery merchants are marked; the marked merchant information and / or product information of each merchant to be pushed is sent to the user.

[0115] The implementation method of step 211 is similar to the implementation method of step 210. The corresponding implementation methods and technical effects can be found in step 210 and will not be repeated here. It should be noted that in this embodiment, no matter whether the user initiates an information search request or an information push request, the server first obtains the merchant information within the traffic range of each merchant, and then sends it to the user after sorting and other processing. In addition to the above steps, other information processing methods, such as merchant sorting processing, etc. can be executed using existing models or methods, and this embodiment does not make specific restrictions here. Compared with the existing method of searching for merchant information within the delivery range of each merchant, the method provided in this embodiment can effectively improve the matching degree between the product and the user and the merchant, thereby improving the user experience on the one hand and improving the merchant's resource utilization on the other hand.

[0116] By applying the technical solution of this embodiment, it is possible to get rid of the limitations of the fixed-configuration delivery range on the visible range of merchants, so that the merchants searched or pushed by the platform can meet the user's certainty needs, thereby improving the accuracy and rationality of product search or push, and further improving the user experience. In addition, it can also enable the products of capable merchants to reach users in need, avoiding waste of merchant resources.

[0117] In one embodiment, an information display method is also provided, which includes the following steps: in response to an information search request or an information push request initiated by a user, displaying the merchant information and / or product information of at least one merchant, wherein the displayed merchant is obtained based on a traffic range query of each merchant within the capacity boundary of the user's area, and the merchant's traffic range can be generated based on the method described in any of the above embodiments.

[0118] In the above embodiment, when a user initiates an information search request or an information push request through the client, the server can query the corresponding merchant within the traffic range of each merchant within the capacity boundary of the user's area, and push the merchant to the user after filtering and sorting. In this embodiment, the user's location is within the traffic range of the pushed merchant, that is, each displayed merchant is visible to the user within their respective traffic ranges. Therefore, users within the merchant's traffic range can view the merchant through various methods such as search or push, and determine whether to place an order with the merchant, and then further determine the corresponding fulfillment method, such as delivery, self-pickup, in-store, etc. Among them, the capacity boundary range of the user's area is larger than the traffic range and delivery range of each merchant, and the traffic range and delivery range of each merchant at least partially overlap.

[0119] In this embodiment, the traffic range of each merchant can be generated in a targeted manner based on the actual situation of the merchant. Compared with the traditional method of defining the delivery range with the merchant as the center and the delivery distance as the radius, the traffic range can not only meet the deterministic demand of users for a certain commodity in a certain area, but also expand the merchant's own order supply range, thereby achieving the decoupling between the merchant's visible range and the delivery range, ensuring that the generated orders can meet the needs of both users and merchants at the same time, thereby improving the user experience, and at the same time, avoiding the waste of merchant resources. In this embodiment, the method for generating the traffic range of each merchant can refer to any of the above embodiments, and this embodiment will not be repeated here.

[0120] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, storage, and display, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. In addition, the numbers corresponding to the various steps in the above embodiments serve only as identifiers and do not limit the order in which the steps are executed. The order in which the steps are executed in each embodiment can be set according to actual circumstances.

[0121] Further, as Figures 1 to 3 In the specific implementation of the method, the embodiment of the present application provides a device for generating a merchant traffic range, such as Figure 4 As shown, the device includes:

[0122] The starting point pair acquisition module 31 may be used to acquire multiple starting point pairs between a target merchant and multiple interest areas within the capacity boundary;

[0123] The indicator value estimation module 32 is configured to estimate the preset indicator value of each starting point pair within a preset future period of time using a pre-trained preset indicator estimation model, and determine candidate starting point pairs based on the estimated indicator values ​​of each starting point pair;

[0124] A starting point pair screening module 33 is configured to screen the candidate starting point pairs based on non-real-time traffic information to obtain a target starting point pair, wherein the non-real-time traffic information includes historical user demand information and merchant configuration information;

[0125] The traffic range determination module 34 may be configured to generate a traffic range of the target merchant based on the target starting point pair, wherein the target merchant is visible to the user within the traffic range of the target merchant.

[0126] In a specific application scenario, the device also includes a traffic range adjustment module, which is used to obtain real-time traffic information according to a preset period, wherein the real-time traffic information includes at least one of time and space information, hotspot information, distribution resource information and manual configuration information; based on the real-time traffic information, at least one extended starting point pair is screened out within the capacity boundary; based on the extended starting point pair, the traffic range of the target merchant is adjusted to obtain a real-time adjusted traffic range.

[0127] In a specific application scenario, the traffic range adjustment module is also used to adjust the traffic range of the target merchant according to the traffic range corresponding to the channel where the target merchant is deployed, so as to obtain the traffic range of the target merchant in the channel, wherein the target merchant has at least one channel where the target merchant is deployed.

[0128] In a specific application scenario, the traffic range of the target merchant at least partially overlaps with the delivery range of the target merchant; the device also includes a delivery range adjustment module; the delivery range adjustment module is used to expand the delivery range of the target merchant based on at least part of the target starting point pairs in the traffic range of the target merchant to obtain the expanded delivery range.

[0129] In a specific application scenario, the device also includes a traffic range storage module and a traffic range push module, wherein the traffic range storage module is used to store the target starting point pairs within the traffic range of the target merchant; and / or the traffic range push module is used to send the traffic range of the target merchant to the target merchant to display the traffic range of the target merchant.

[0130] In a specific application scenario, the device also includes a merchant information query module, wherein the merchant information query module is used to determine the search keyword and the user's location in response to an information search request; based on the search keyword and the user's location, query at least one candidate merchant within the traffic range of each merchant within the capacity boundary; based on the delivery range of each candidate merchant and the user's location, screen the candidate merchants to obtain at least one merchant to be searched; based on the search keyword and user preference information, sort the at least one merchant to be searched, and send the merchant information and / or product information of the sorted merchant to the user.

[0131] In a specific application scenario, the merchant information query module is also used to determine the search keywords and the user's location in response to the information search request containing out-of-region delivery requirements; based on the search keywords and the user's location, query at least one merchant to be searched within the traffic range of each merchant within the capacity boundary; based on the search keywords and user preference information, sort the at least one merchant to be searched, and send the merchant information and / or product information of the sorted merchant to be searched to the user.

[0132] In a specific application scenario, the merchant information query module is also used to determine out-of-area delivery merchants based on the delivery range of each merchant to be searched and the user's location, and mark the out-of-area delivery merchants; and send the marked merchant information and / or product information of each merchant to be searched to the user.

[0133] In a specific application scenario, the merchant information query module is also used to respond to an information push request, determine the user's location, and determine the product category to be pushed based on the user preference information; based on the product category to be pushed and the user's location, query at least one candidate merchant within the traffic range of each merchant within the capacity boundary; based on the delivery range of each candidate merchant and the user's location, screen the candidate merchants to obtain at least one merchant to be pushed; based on the user preference information, sort the at least one merchant to be pushed, and send the merchant information and / or product information of the sorted merchant to be pushed to the user.

[0134] In a specific application scenario, the merchant information query module is also used to respond to the out-of-region delivery demand contained in the information push request, determine the user's location, and determine the product category to be pushed based on the user preference information; based on the product category to be pushed and the user's location, query at least one merchant to be pushed within the traffic range of each merchant within the capacity boundary; based on the user preference information, sort the at least one merchant to be pushed, and send the merchant information and / or product information of the sorted merchant to be pushed to the user.

[0135] In a specific application scenario, the merchant information query module is also used to determine out-of-area delivery merchants based on the delivery range of each merchant to be pushed and the user's location, and mark the out-of-area delivery merchants; and send the marked merchant information and / or product information of each merchant to be pushed to the user.

[0136] In a specific application scenario, the device also includes an estimation model training module, wherein the estimation model training module is used to obtain multiple sample starting points between multiple merchants and multiple interest areas within the capacity boundary as training samples; for each of the sample starting points, the merchant information corresponding to the sample starting point, the interest area information, the intersection information between the merchant and the interest area, the merchant information within multiple distance ranges of the interest area, and the order statistics information of the users within the interest area at the merchants within multiple distance ranges are used as features, and the preset indicator value of the sample starting point in the future preset time period is used as a label to iteratively train the pre-constructed model; when the pre-constructed loss function reaches the preset range, the model training is stopped to obtain the trained preset indicator estimation model.

[0137] In a specific application scenario, the estimation model training module is also used to obtain multiple starting point pairs between multiple merchants and multiple interest areas within the capacity boundary, and obtain the number of orders for each of the starting point pairs; retain the starting point pairs whose order numbers are greater than a preset threshold, and sample the starting point pairs whose order numbers are less than or equal to the preset threshold to obtain multiple sample starting point pairs as training samples.

[0138] In a specific application scenario, the estimation model training module is also used to convert the merchant information, the interest area information and the cross-information between the merchant and the interest area into merchant features, interest area features and cross-features for each sample starting point, convert the merchant information within multiple distance ranges of the interest area into supply sequence features, and convert the order statistics information of users within the interest area at merchants within multiple distance ranges into demand sequence features; perform attention processing on each merchant feature in the supply sequence features to obtain supply features, and perform attention processing on each order statistical feature in the demand sequence features to obtain demand features; and splice or fuse the merchant features, the interest area features, the cross-features, the supply features and the demand features to obtain the feature expression of the sample starting point.

[0139] In a specific application scenario, the estimation model training module is also used to construct an integrated learning model, and iteratively train the integrated learning model through the training samples to obtain an initial preset indicator estimation model as a baseline; after the integrated learning model training is completed, a deep learning model is constructed, and the training samples are expanded to expand the geographical scope covered by the training samples; based on the expanded training samples, the deep learning model is iteratively trained to obtain an optimized preset indicator estimation model.

[0140] In a specific application scenario, the deep learning model uses a deep neural network as the backbone network, and sequentially connects a deep factor decomposition machine module and a multi-gate hybrid expert module to the output end of the backbone network. The deep factor decomposition machine module is used to learn explicit and implicit high-order cross features, and the multi-gate hybrid expert module is used to realize the multi-task prediction capability of the model.

[0141] In a specific application scenario, the deep learning model also integrates a deviation tower network and / or a generative network, wherein the deviation tower network and / or the generative network takes at least one of the city information, product category information and distance information in the training sample as input, and is used to correct the deviation values ​​of the city characteristics, product category characteristics and distance characteristics in the training sample, so that the model can be applied to multiple different cities, product categories and distances at the same time.

[0142] It should be noted that for other corresponding descriptions of the functional units involved in the device for generating a merchant traffic range provided in the embodiment of the present application, please refer to Figures 1 to 3 The corresponding description in the method will not be repeated here.

[0143] The embodiment of the present application also provides a computer device, which can be specifically a personal computer, a server, a network device, etc. The computer device includes a bus, a processor, a memory and a communication interface, and may also include an input and output interface and a display device. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store location information. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the steps in each method embodiment are implemented.

[0144] Those skilled in the art will understand that the structure of the above-mentioned computer device is only a partial structure related to the solution of the present application and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components, or combine certain components, or have a different component arrangement.

[0145] In one embodiment, a computer-readable storage medium is provided. The computer-readable storage medium may be non-volatile or volatile, and stores a computer program thereon. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0146] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0147] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, graphics processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, and the like.

[0148] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0149] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for generating a merchant traffic range, characterized in that: The method comprises: Within the capacity boundary, obtain multiple starting point pairs between the target merchant and multiple areas of interest; Using a pre-trained pre-set indicator estimation model, the preset indicator value of each starting point pair in a future preset period is estimated, and candidate starting point pairs are determined based on the estimated indicator values ​​of each starting point pair; Filtering the candidate starting point pairs based on non-real-time traffic information to obtain a target starting point pair, wherein the non-real-time traffic information includes historical user demand information and merchant configuration information; Based on the target starting point pair, generating a traffic range of the target merchant, wherein the target merchant is visible to the user within the traffic range of the target merchant, and the traffic range of the target merchant at least partially overlaps with the delivery range of the target merchant; Among them, the preset indicator estimation model uses the merchant characteristics, interest area characteristics, cross-features of merchants and interest areas, supply characteristics, and demand characteristics corresponding to the sample starting point within the capacity boundary range as the feature expression, and is iteratively trained with the preset indicator value of the sample starting point in the future preset time period as the label. The preset indicator value is at least one of the order volume, average order amount, merchant visit rate, and total order amount; the supply feature is obtained by attention processing of each merchant feature in the supply sequence feature, and the supply sequence feature is obtained by converting the merchant information within multiple distance ranges of the interest area; the demand feature is obtained by attention processing of each order statistical feature in the demand sequence feature, and the demand sequence feature is obtained by converting the order statistical information of merchants within multiple distance ranges of users in the interest area.

2. The method for generating a merchant traffic range according to claim 1, characterized in that: The method further comprises: acquiring real-time traffic information according to a preset period, and screening at least one extended starting point pair within the transportation capacity boundary based on the real-time traffic information; adjusting the traffic range of the target merchant based on the extended starting point pair to obtain a real-time adjusted traffic range, wherein the real-time traffic information includes at least one of spatiotemporal information, hotspot information, distribution resource information, and manual configuration information; and / or, Adjusting the traffic range of the target merchant according to the traffic range corresponding to the channel in which the target merchant is placed, to obtain the traffic range of the target merchant in the channel, wherein the target merchant is placed in at least one channel; and / or Expanding the delivery range of the target merchant based on at least some of the target starting point pairs in the traffic range of the target merchant to obtain an expanded delivery range, wherein the traffic range of the target merchant at least partially overlaps with the delivery range of the target merchant; and / or, Storing target starting point pairs within the traffic range of the target merchant; and / or, The traffic range of the target merchant is sent to the target merchant to display the traffic range of the target merchant.

3. The method for generating a merchant traffic range according to claim 1 or 2, characterized in that: The method further comprises: In response to an information search request, a search keyword and a user's location are determined; based on the search keyword and the user's location, at least one candidate merchant is searched within the traffic range of each merchant within the capacity boundary; based on the delivery range of each candidate merchant and the user's location, the candidate merchants are screened to obtain at least one merchant to be searched; based on the search keyword and user preference information, the at least one merchant to be searched is sorted, and merchant information and / or product information of the sorted merchant to be searched is sent to the user; and / or, In response to an out-of-region delivery demand included in an information search request, a search keyword and a user location are determined; based on the search keyword and the user location, at least one merchant to be searched is queried within the traffic range of each merchant within the capacity boundary; based on the search keyword and user preference information, the at least one merchant to be searched is sorted, and the merchant information and / or product information of the sorted merchant to be searched is sent to the user.

4. The method for generating a merchant traffic range according to claim 3, characterized in that: The sending of the sorted merchant information and / or product information of the merchants to be searched to the user includes: Determine out-of-area delivery merchants based on the delivery range of each merchant to be searched and the user's location, and mark the out-of-area delivery merchants; The marked merchant information and / or product information of each merchant to be searched is sent to the user.

5. The method for generating a merchant traffic range according to claim 1 or 2, characterized in that: The method further comprises: In response to an information push request, determining the user's location and determining the product category to be pushed based on the user's preference information; based on the product category to be pushed and the user's location, querying at least one candidate merchant within the traffic range of each merchant within the capacity boundary; based on the delivery range of each candidate merchant and the user's location, screening the candidate merchants to obtain at least one merchant to be pushed; sorting the at least one merchant to be pushed based on the user's preference information, and sending the merchant information and / or product information of the sorted merchant to be pushed to the user; and / or, In response to an out-of-region delivery demand included in an information push request, the user's location is determined, and the product category to be pushed is determined based on the user's preference information; based on the product category to be pushed and the user's location, at least one merchant to be pushed is queried within the traffic range of each merchant within the capacity boundary; based on the user's preference information, the at least one merchant to be pushed is sorted, and the merchant information and / or product information of the sorted merchant to be pushed is sent to the user.

6. The method for generating a merchant traffic range according to claim 5, characterized in that: The step of sending the sorted merchant information and / or product information to be pushed to the user includes: Determine out-of-area delivery merchants based on the delivery range of each merchant to be pushed and the user's location, and mark the out-of-area delivery merchants; The marked merchant information and / or product information of each merchant to be pushed is sent to the user.

7. The method for generating a merchant traffic range according to claim 1, characterized in that: The training method of the preset indicator prediction model includes: Within the capacity boundary, obtaining multiple sample starting points between multiple merchants and multiple interest areas as training samples; For each sample starting point, the pre-built model is iteratively trained using the merchant information corresponding to the sample starting point, information about the area of ​​interest, intersection information between merchants and the area of ​​interest, merchant information within multiple distance ranges of the area of ​​interest, and order statistics of merchants within multiple distance ranges of users within the area of ​​interest as features, and using the preset indicator value of the sample starting point in a preset future time period as a label; When the pre-built loss function reaches the preset range, the model training is stopped and the pre-trained preset indicator estimation model is obtained.

8. The method for generating a merchant traffic range according to claim 7, characterized in that: The step of obtaining a plurality of sample starting points between a plurality of merchants and a plurality of interest areas as training samples within the capacity boundary includes: Within the capacity boundary, obtaining a plurality of starting point pairs between a plurality of merchants and a plurality of interest areas, and obtaining the order quantity of each of the starting point pairs; The starting point pairs whose order quantity is greater than a preset threshold are retained, and the starting point pairs whose order quantity is less than or equal to the preset threshold are sampled to obtain multiple sample starting point pairs as training samples.

9. The method for generating a merchant traffic range according to claim 7 or 8, characterized in that: The iterative training of the pre-built model includes: Constructing an integrated learning model, and iteratively training the integrated learning model using the training samples to obtain an initial preset indicator estimation model as a baseline; After the ensemble learning model is trained, a deep learning model is constructed, and the training samples are expanded to expand the geographical scope covered by the training samples; Based on the expanded training samples, the deep learning model is iteratively trained to obtain an optimized preset indicator estimation model.

10. The method for generating a merchant traffic range according to claim 9, characterized in that: The deep learning model uses a deep neural network as a backbone network, and sequentially connects a deep factorization machine module and a multi-gate hybrid expert module to the output end of the backbone network, wherein the deep factorization machine module is used to learn explicit and implicit high-order cross features, and the multi-gate hybrid expert module is used to realize the multi-task prediction capability of the model; and / or, The deep learning model also integrates a deviation tower network and / or a generative network, wherein the deviation tower network and / or the generative network takes at least one of the city information, product category information and distance information in the training sample as input, and is used to correct the deviation values ​​of the city characteristics, product category characteristics and distance characteristics in the training sample, so that the model is simultaneously applicable to multiple different cities, product categories and distances.

11. An information display method, characterized in that: The method comprises: In response to an information search request or an information push request initiated by a user, the merchant information and / or product information of at least one merchant is displayed, wherein the merchant is obtained by querying the traffic range of each merchant within the capacity boundary of the area where the user is located, and the traffic range of the merchant is generated based on the merchant traffic range generation method described in any one of claims 1 to 10.

12. A device for generating a merchant traffic range, characterized in that: The device comprises: A starting point pair acquisition module is used to acquire multiple starting point pairs between a target merchant and multiple interest areas within the capacity boundary; An indicator value estimation module is used to estimate the preset indicator value of each starting point pair in a preset future period of time using a pre-trained preset indicator estimation model, and to determine candidate starting point pairs based on the estimated indicator values ​​of each starting point pair; a starting point pair screening module, configured to screen the candidate starting point pairs based on non-real-time traffic information to obtain a target starting point pair, wherein the non-real-time traffic information includes historical user demand information and merchant configuration information; a traffic range determination module, configured to generate a traffic range of the target merchant based on the target starting point pair, wherein the target merchant is visible to the user within the traffic range of the target merchant, and the traffic range of the target merchant at least partially overlaps with the delivery range of the target merchant; Among them, the preset indicator estimation model uses the merchant characteristics, interest area characteristics, cross-features of merchants and interest areas, supply characteristics, and demand characteristics corresponding to the sample starting point within the capacity boundary range as the feature expression, and is iteratively trained with the preset indicator value of the sample starting point in the future preset time period as the label. The preset indicator value is at least one of the order volume, average order amount, merchant visit rate, and total order amount; the supply feature is obtained by attention processing from each merchant feature in the supply sequence feature, and the supply sequence feature is converted from merchant information within multiple distance ranges in the interest area; the demand feature is obtained by attention processing from each order statistical feature in the demand sequence feature, and the demand sequence feature is converted from order statistical information of merchants within multiple distance ranges of users in the interest area.

13. A 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 11 is implemented.

14. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 11 is implemented.

Images