CN110689402A - Method, apparatus, electronic device, and readable storage medium for recommending merchants - Google Patents

Abstract

The embodiment of the application provides a method and a device for recommending merchants, electronic equipment and a readable storage medium, aiming at enabling a server to recommend merchant information to a user more accurately and improving the operation efficiency of the user. The method comprises the following steps: obtaining user characteristics of a target user and merchant characteristics of a plurality of candidate merchants respectively; inputting the user characteristics and the merchant characteristics into a repurchase rate prediction model to obtain the repurchase rate of the target user for each candidate merchant in the candidate merchants; recommending a target merchant to the target user according to the repurchase rate of the target user for each candidate merchant in the candidate merchants, wherein the target merchant is at least one of the candidate merchants.

Description

Method, apparatus, electronic device, and readable storage medium for recommending merchants

technical field

The embodiments of the present application relate to the field of Internet technologies, and in particular, to a method, an apparatus, an electronic device, and a readable storage medium for recommending a merchant.

Background technique

With the development of Internet mining technology and the popularization of intelligent terminal devices, more and more end users connect to e-commerce platforms or O2O (Online To Offline) platforms and other servers through browsers or clients to realize online shopping, online shopping, and online shopping. Online transaction activities such as ordering meals and online ticket purchases. In the related art, the server will actively recommend business information to the user according to the user's historical browsing records, so that the user can quickly enter the homepage of the target business without performing a search operation, so as to promote the target business to obtain benefits and improve the performance of the target business. User operation efficiency.

When the current server recommends business information to users, it usually determines the target business with a higher historical click-through rate and/or historical conversion rate according to the historical click-through rate and/or historical conversion rate of each business information, and compares the target merchant's Information is recommended to users. However, with the current recommendation method, the business information recommended to the user can only satisfy the maximization of the current revenue of the server and the maximization of the user's current operation efficiency, but cannot satisfy the maximization of the long-term revenue of the server and the long-term operation efficiency of the user maximization.

SUMMARY OF THE INVENTION

Embodiments of the present application provide a method, apparatus, electronic device, and readable storage medium for recommending merchants, which aim to enable the server to more accurately recommend merchant information to users, so as to improve the long-term revenue of the server and the long-term operation efficiency of the user.

A first aspect of the embodiments of the present application provides a method for recommending a merchant, and the method includes:

Obtain the user characteristics of the target user and the respective merchant characteristics of multiple candidate merchants;

Inputting the user characteristics and the multiple merchant characteristics into a repurchase rate prediction model to obtain the repurchase rate of the target user for each candidate merchant in the multiple candidate merchants;

According to the repurchase rate of the target user for each candidate merchant of the plurality of candidate merchants, a target merchant is recommended to the target user, and the target merchant is at least one of the plurality of candidate merchants.

Optionally, the method further includes constructing a repurchase rate prediction model.

Optionally, the building a repurchase rate prediction model includes:

For a historical user who has placed an order, obtain the sequence of business browsing records of the historical user within a preset time period from the time of placing the order;

For each merchant browsing record in the merchant browsing record sequence, a training sample corresponding to the merchant browsing record is established, and the training sample includes: the merchant characteristics, user characteristics and user repurchase situations corresponding to the merchant browsing record. The reward value of , the merchant characteristics and user characteristics corresponding to the next merchant browsing record;

Construct a training sample set according to the corresponding training samples of multiple business browsing records;

Based on the training sample set, a preset reinforcement learning model is trained to obtain the repurchase rate prediction model.

Optionally, the method further includes:

For each merchant browsing record in the merchant browsing record sequence:

In the case that the historical user places an order for the merchant corresponding to the merchant's browsing record, it is determined that the reward value corresponding to the user's repurchase situation is a value greater than zero;

In the case that the historical user has not placed an order for the merchant corresponding to the merchant's browsing record, it is determined that the reward value corresponding to the user's repurchase situation is a value not greater than zero.

Optionally, in the case that the historical user places an order for the merchant corresponding to the merchant's browsing record, determining the reward value corresponding to the user's repurchase situation, including:

In the case where the historical user places an order for the merchant corresponding to the merchant's browsing record, determine the time difference between the historical user's current ordering time and the last ordering time;

According to the order time difference and the preset reward value function, the reward value corresponding to the user's repurchase situation is determined, and the reward value is negatively correlated with the order time difference.

Optionally, according to the order time difference and a preset reward value function, determine the reward value corresponding to the user's repurchase situation, including:

The reward value corresponding to the user's repurchase situation is determined according to the following formula:

Among them, r represents the reward value corresponding to the user's repurchase situation, C represents the weight adjustment coefficient, and T represents the time difference between the order placement.

Optionally, the user characteristics include at least one of the following: the user's consumption preference, the user's geographic location, a user portrait, and a behavior characteristic.

Optionally, the merchant characteristics include at least one of the following: merchant ID, category ID, delivery time, sales volume, full discount limit, and customer unit price.

A second aspect of an embodiment of the present application provides an apparatus for recommending a merchant, the apparatus comprising:

The feature obtaining module is used to obtain the user feature of the target user and the respective merchant features of multiple candidate merchants;

a repurchase rate obtaining module, configured to input the user characteristics and the multiple merchant characteristics into a repurchase rate prediction model to obtain the repurchase rate of the target user for each candidate merchant in the multiple candidate merchants;

A target merchant recommendation module, configured to recommend a target merchant to the target user according to the repurchase rate of the target user for each candidate merchant in the plurality of candidate merchants, and the target merchant is one of the plurality of candidate merchants at least one of.

Optionally, the device further includes:

The model building module is used to build a repurchase rate prediction model.

Optionally, the model building module includes:

The merchant browsing record sequence sub-module is used to obtain the merchant browsing record sequence of the historical user within the preset time period from the time of placing the order for the historical user who has placed the order;

A training sample establishment sub-module is used to establish a training sample corresponding to the merchant browsing record for each merchant browsing record in the merchant browsing record sequence, and the training sample includes: the merchant feature corresponding to the merchant browsing record, User characteristics and reward value corresponding to the user's repurchase situation, merchant characteristics and user characteristics corresponding to the next merchant browsing record;

The training sample set construction sub-module is used to construct the training sample set according to the corresponding training samples of multiple merchant browsing records;

The model training sub-module is used for training a preset reinforcement learning model based on the training sample set to obtain the repurchase rate prediction model.

Optionally, the model building module further includes:

The reward value determination sub-module is used for each merchant browsing record in the merchant browsing record sequence, in the case that the historical user places an order for the merchant corresponding to the merchant browsing record, determine the user's repurchase situation corresponding to the situation. The reward value is a value greater than zero; in the case that the historical user has not placed an order for the merchant corresponding to the merchant's browsing record, it is determined that the reward value corresponding to the user's repurchase is a value not greater than zero.

Optionally, the reward value determination submodule includes:

an order time difference determination unit, configured to determine the order time difference between the current order time of the historical user and the last order time when the historical user places an order for the merchant corresponding to the merchant browsing record;

The reward value determination unit is configured to determine the reward value corresponding to the user's repurchase situation according to the order time difference and a preset reward value function, and the reward value is negatively correlated with the order time difference.

Optionally, the reward value determination unit includes:

The reward value determination subunit is used to determine the reward value corresponding to the user's repurchase situation according to the following formula:

Among them, r represents the reward value corresponding to the user's repurchase situation, C represents the weight adjustment coefficient, and T represents the time difference between the order placement.

Optionally, the user characteristics include at least one of the following: the user's consumption preference, the user's geographic location, a user portrait, and a behavior characteristic.

Optionally, the merchant characteristics include at least one of the following: merchant ID, category ID, delivery time, sales volume, full discount limit, and customer unit price.

A third aspect of the embodiments of the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the steps in the method described in the first aspect of the present application.

A fourth aspect of the embodiments of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, where the processor implements the method described in the first aspect of the present application when executed. A step of.

Using the method for recommending merchants provided by the embodiments of the present application, the user characteristics of the target user and the respective merchant characteristics of multiple candidate merchants are first obtained, and then these characteristics are input into the pre-trained repurchase rate prediction model to obtain the target user for multiple candidate merchants. The repurchase rate of each candidate merchant in the merchant. Among them, the repurchase rate of each candidate merchant represents the long-term demand of the target user for the candidate merchant. The higher the repurchase rate, the greater the long-term demand of the target user for the candidate merchant, and the greater the probability that the target user places an order for the candidate merchant again. Finally, according to the repurchase rates of multiple candidate merchants, the target merchant is recommended to the target user, so as to meet the user's long-term needs, improve the long-term revenue of the server, and improve the user's long-term operation efficiency on the server homepage.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present application more clearly, the following briefly introduces the drawings that are used in the description of the embodiments of the present application. Obviously, the drawings in the following description are only some embodiments of the present application. , for those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative labor.

Fig. 1 is the training flow chart of the repurchase rate prediction model proposed by an embodiment of the present application;

2 is a schematic diagram of a training sample proposed by an embodiment of the present application;

3 is a schematic diagram of a method for determining a reward value proposed by an embodiment of the present application;

4 is a flowchart of a method for recommending a merchant proposed by an embodiment of the present application;

5 is a schematic diagram of interaction between a server and a client according to an embodiment of the present application;

FIG. 6 is a schematic diagram of an apparatus for recommending a merchant provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

In the field of Internet technology, after the server homepage is displayed by a browser or client user, the server will actively recommend business information to the user according to the user's historical browsing records, so that the user can quickly enter the website without performing a search operation. Target business home page, improve user operation efficiency. When the current server recommends business information to users, it usually determines the target business with a higher historical click-through rate and/or historical conversion rate according to the historical click-through rate and/or historical conversion rate of each business information, and compares the target merchant's Information is recommended to users. However, with the current recommendation method, the business information recommended to the user can only satisfy the maximization of the current revenue of the server and the maximization of the user's current operation efficiency, but cannot satisfy the maximization of the long-term revenue of the server and the long-term operation efficiency of the user maximization.

In view of this, the embodiment of the present application proposes to first obtain the user characteristics of the target user and the respective merchant characteristics of multiple candidate merchants, and then determine the repurchase rate of the target user for each of the multiple candidate merchants according to these characteristics. Finally, according to the repurchase rates of multiple candidate merchants, the target merchant is recommended to the target user, so as to meet the user's long-term needs, improve the long-term revenue of the server, and improve the user's long-term operation efficiency on the server homepage.

In addition, in order to more intelligently implement the above method proposed by the embodiment of the present application, the embodiment of the present application first collects training samples for the preset reinforcement learning model, constructs a training sample set, and finally, based on the training sample set, generates the preset reinforcement learning model for the preset reinforcement learning model. Perform training to obtain a repurchase rate prediction model. The preset reinforcement learning model may be a commonly used model in related technologies. The repurchase rate prediction model may be used to execute some or all of the steps in the above method proposed in the embodiments of the present application.

Referring to FIG. 1 , FIG. 1 is a training flow chart of a repurchase rate prediction model proposed by an embodiment of the present application. As shown in Figure 1, the training process includes the following steps:

Step S11: For a historical user who has placed an order, obtain a sequence of business browsing records of the historical user within a preset time period from the time of placing the order.

In this embodiment, the sequence of business browsing records of a historical user within a preset time period from the time of placing an order refers to: starting from an order placing behavior of the historical user, within the preset time period, the historical user has continuously After entering the home page of the server multiple times, the business information pushed by the server to the historical user, and/or a business information sequence composed of business information actively searched by the historical user.

In addition to the merchant information, each merchant browsing record in the merchant browsing record sequence also includes: a timestamp when the merchant information was browsed or pushed, and a mark indicating whether the user placed an order for the merchant corresponding to the merchant information.

For example, it is assumed that a historical user placed an order on April 16, with April 16 as the starting point and 15 days as the preset time period, referring to Table 1, which schematically shows the history A sequence of user's business browsing records.

Table 1 List of business browsing records of historical users

serial number time Business information Whether to place an order 1 April 17 at 11:23 c Merchant No order 2 April 17 at 11:25 b business place an order 3 April 21 at 11:49 c Merchant No order 4 April 22 at 18:33 d Merchant No order 5 April 26 at 11:19 e-merchants No order 6 April 26 at 11:24 d Merchant place an order 7 April 29 at 12:07 b business place an order 8 May 1st at 18:51 f Merchant No order 9 May 1st at 19:15 f Merchant place an order

Each row in Table 1 represents a merchant browsing record. Taking the first row in Table 1 as an example, it means that the historical user browsed or was pushed the c-merchants information at 11:23 on April 17, and the historical user did not Place an order at c merchant.

Step S12: For each merchant browsing record in the merchant browsing record sequence, establish a training sample corresponding to the merchant browsing record, and the training sample includes: merchant characteristics, user characteristics and user repetitions corresponding to the merchant browsing record. The reward value corresponding to the purchase situation, the merchant characteristics and user characteristics corresponding to the next merchant browsing record.

Referring to FIG. 2 , FIG. 2 is a schematic diagram of a training sample proposed by an embodiment of the present application. As shown in Figure 2, where s represents user characteristics, a represents merchant characteristics, and r represents reward value. Take the first training sample in Figure 2 as an example for illustration: where s1 represents the user feature corresponding to the first merchant browsing record, a1 represents the merchant feature corresponding to the first merchant browsing record, and r1 represents the first merchant browsing record The reward value corresponding to the user's repurchase situation, s2 represents the user feature corresponding to the second merchant's browsing record, and a2 represents the merchant's feature corresponding to the second merchant's browsing record.

Following Table 1 above, take the second merchant browsing record in Table 1 as an example to illustrate: the training samples corresponding to this merchant browsing record include: merchant characteristics of merchant b at 11:25 on April 17, The user characteristics of the historical user at 11:25 on the 17th, the reward value corresponding to the order behavior of the historical user, the merchant characteristics of the merchant at 11:49 on April 21, and the merchant characteristics at 11:49 on April 21 User characteristics of the historical user at 49:00.

In this embodiment, the merchant characteristics may include at least one of the following: merchant ID, category ID, delivery time, sales volume, full discount limit, and customer unit price. Specifically, the merchant feature may be a vectorized representation of the above, such as word vectors.

In this embodiment, the user characteristics may include at least one of the following: the user's consumption preference, the user's geographic location, the user's portrait, and behavior characteristics. Specifically, the user feature may be a vectorized representation of the above, such as word vectors.

Among them, the user's consumption preference refers to the user's preference for commodity categories. For example, the takeaway business may be the user's taste preference. User portrait refers to user attribute information extracted from massive user data. The user attribute information can include one or more of the following information: gender, occupation, age group, income level, marital status, education level, etc. Basic attributes, behavioral attributes such as APP usage frequency and order probability, as well as interest attributes such as takeaway preferences, movie preferences, and commodity preferences. The behavioral characteristics of the user may refer to the order rate when the user browses the business information multiple times, the frequency of the user using the client, the probability of the user using coupons, and the like.

In this embodiment, when a training sample corresponding to the merchant browsing record is established for each merchant browsing record, the merchant characteristics and user characteristics of the time corresponding to the merchant browsing record can be obtained from the cached log, and the corresponding time of the merchant browsing record can be obtained from the cached log. The business characteristics and user characteristics of the time corresponding to the next business browsing record are obtained from the log.

Or, after the historical user has placed an order on April 16, when step S11 is executed, for each merchant browsing of the historical user, the current merchant characteristics and user characteristics are recorded, and the recorded merchant characteristics and user characteristics are recorded. User characteristics are part of the information in the business browsing record. In this way, when establishing a training sample corresponding to the merchant browsing record for each merchant browsing record, the merchant characteristics and user characteristics are directly read from the merchant browsing record, and the merchant characteristics are read from the next merchant browsing record. and user characteristics.

In this embodiment, the reward value corresponding to the user's repurchase situation is used to represent whether the historical user places an order for the merchant in the merchant's browsing record. During the implementation of this application, if a historical user has placed an order for a merchant in the merchant's browsing record, the corresponding reward value is a larger value. If the historical user does not place an order for the merchant in the merchant's browsing record, the corresponding reward value is a small value.

In this embodiment, when establishing a training sample for each merchant browsing record in the merchant browsing record sequence, in order to determine the reward value in the training sample, a feasible implementation is: for the merchant browsing record Each merchant browsing record in the sequence: when the historical user places an order for the merchant corresponding to the merchant browsing record, it is determined that the reward value corresponding to the user's repurchase situation is a value greater than zero; In the case that the merchant corresponding to the merchant browsing record has not placed an order, it is determined that the reward value corresponding to the user's repurchase situation is a value not greater than zero.

Exemplarily, following Table 1 above, the reward value in the training samples corresponding to the merchant browsing records with the serial numbers 1, 3, 4, 5 or 8 is not greater than 0, and the merchant browsing records with the serial numbers 2, 6, 7 or 9 correspond to The reward value in the training samples is greater than 0.

Among them, when a historical user places an order for a merchant corresponding to a certain merchant browsing record, the specific method of determining the reward value can refer to FIG. As shown in Figure 3, the method for determining the reward value includes the following steps:

Step S12-1: In the case that the historical user places an order for the merchant corresponding to the current merchant browsing record, determine the time difference between the historical user's current ordering time and the previous ordering time;

Step S12-2: Determine a reward value corresponding to the user's repurchase situation according to the order time difference and a preset reward value function, and the reward value is negatively correlated with the order time difference.

For example, following Table 1 above, the ordering behavior of historical users is recorded in the merchant browsing record with serial number 2. The ordering time of the ordering behavior (April 17) and the last order time (April 16) are recorded. day) time difference is 1 day. The ordering behavior of historical users is recorded in the merchant browsing record with the serial number 6. The time difference between the ordering time (April 26) and the last ordering time (April 17) of the ordering behavior is 9 days. After the above step S12-2, the reward value in the training sample corresponding to the merchant browsing record with the serial number 2 is greater than the reward value in the training sample corresponding to the merchant browsing record with the serial number 6.

The reward value function can be in the form of: r=C/(T+1), where r represents the reward value corresponding to the user's repurchase situation, C represents the weight adjustment coefficient, and T represents the time difference for placing the order. The weighting coefficient C is a positive number greater than 0, and the weighting coefficient can be manually changed according to training needs. The unit of time difference for placing an order may be minutes, hours, days, etc., which is not limited in this application.

By determining the reward value in the manner of the above steps S12-1 and S12-2, the obtained reward value not only represents whether the historical user repurchases (ie, places an order), but also represents the repurchase interval. Using the training samples including the reward value to train the reinforcement learning model can make the repurchase rate prediction model obtained from the training not only recommend business information for the user's long-term needs, but also improve the user's repurchase probability and help users reduce active searches and browsing. time. The repurchase rate prediction model can also distinguish factors of the length of the repurchase interval, and can recommend merchants that make the user repurchase shorter, increase the user's order frequency, further improve the recommendation accuracy, and further improve the user's operation efficiency.

Step S13: Construct a training sample set according to the training samples corresponding to the multiple browsing records of the merchants.

In this embodiment, the training sample set includes training samples corresponding to multiple merchant browsing records, and the multiple training samples are sorted according to their corresponding merchant browsing record times. As shown in FIG. 2 , training sample 1 to training sample 4 form a training sample set. The training samples in the training sample set are arranged in chronological order: training sample 1 , training sample 2 , training sample 3 and training sample 4 .

Step S14: Based on the training sample set, a preset reinforcement learning model is trained to obtain the repurchase rate prediction model.

In this embodiment, based on the training sample set, a preset reinforcement learning model is trained to obtain a repurchase rate prediction model. The preset reinforcement learning model may be a DQN (Deep Q Network; Deep Q Network) model, a DDPG (Deep Deterministic Policy Gradient; Deep Deterministic Policy Gradient) model, or a DDQN (Double DQN; Double Deep Q Network) model, or the like. The repurchase rate prediction model can be used as an optional means to execute some or all of the steps in the above method proposed in the embodiment of the present application.

Referring to FIG. 4 , FIG. 4 is a flowchart of a method for recommending a merchant proposed by an embodiment of the present application. As shown in Figure 4, the method includes the following steps:

Step S41: Obtain the user characteristics of the target user and the respective merchant characteristics of the multiple candidate merchants.

The user characteristics include at least one of the following: the user's consumption preference, the user's geographic location, the user's portrait, and behavior characteristics. User features can be abstract representations of at least one of the above, such as word vectors. The merchant characteristics include at least one of the following: merchant ID, category ID, delivery time, sales volume, full discount limit, and customer unit price. The merchant feature can be an abstract representation of at least one of the above, such as a word vector.

The target user may refer to: a user who is opening the client, or a user who is entering the homepage of the server through a browser. Take a takeaway server as an example, the server is oriented to a large number of users at the same time, some users have a takeaway client installed on their terminal devices, and some users have a browser installed on their terminal devices. Due to the large number of concurrent users accessing the food delivery server, at a certain moment, there may be multiple target users on the food delivery server at the same time, some of which are opening the food delivery client, and some target users are browsing The server enters the home page of the takeaway server. The takeaway server performs the above step S41 , the following step S42 and the following step S43 respectively for each target user.

The candidate merchants may refer to: multiple merchants selected by the existing screening methods for the target user. For example, multiple merchants are selected based on delivery distance, user taste, sales volume, historical evaluation, or advertising auction.

Step S42: Input the user characteristics and the multiple merchant characteristics into a repurchase rate prediction model to obtain the repurchase rate of the target user for each of the multiple candidate merchants.

Wherein, the repurchase rate prediction model may be the repurchase rate prediction model trained by the above steps S11 to S14, or may be the repurchase rate prediction model trained by other methods. In this step, the repurchase rate prediction model is The source and training method are not limited.

In this embodiment, the user feature and multiple merchant features can be combined one-to-one to form multiple feature combinations, and each feature combination includes the user feature and the merchant feature of a candidate merchant. Then, multiple feature combinations are input into the repurchase rate prediction model in turn, and the repurchase rate output by the repurchase rate prediction model for each feature combination is obtained. The higher the repurchase rate, the greater the long-term demand of the target user for the candidate merchant, and the higher the probability of the target user placing an order for the candidate merchant.

Step S43: Recommend a target merchant to the target user according to the repurchase rate of the target user for each candidate merchant among the multiple candidate merchants, where the target merchant is at least one of the multiple candidate merchants.

For example, the candidate merchant with the largest repurchase rate among the multiple candidate merchants may be directly used as the target merchant and recommended to the target user.

Alternatively, in order to increase the exploratory nature of the recommendation, the recommendation probability of each candidate merchant can be determined according to the repurchase rate of multiple candidate merchants, and then the candidate merchant can be recommended to the target user based on the recommendation probability of each candidate merchant. For example, the repurchase rates corresponding to candidate merchants A, B, and C are: 0.2, 0.6, and 0.4, respectively, then the recommendation probability of merchant A is equal to 0.2/(0.2+0.6+0.4)=0.17, and the recommendation probability of merchant B is equal to 0.6 /(0.2+0.6+0.4)=0.50, the recommendation probability of merchant C is equal to 0.4/(0.2+0.6+0.4)=0.33. In this way, when recommending candidate merchants to the target user, merchant A is recommended with a probability of 0.17, merchant B is recommended with a probability of 0.5, and merchant C is recommended with a probability of 0.33. In other words, each candidate merchant has a possibility of being recommended, and the possibility of being recommended for each candidate merchant is sorted as follows: merchant B, merchant C, and merchant A.

Alternatively, after determining the repurchase rate of each candidate merchant, other considerations for each candidate merchant, such as delivery distance, user taste, sales, historical evaluation, ad bidding, click-through rate (CTR), or conversion rate may be used (CVR), etc., comprehensively determine the target merchants.

By executing the above-mentioned method for recommending merchants including steps S41 to S43, first obtain the user characteristics of the target user and the respective merchant characteristics of multiple candidate merchants, and then input these characteristics into the pre-trained repurchase rate prediction model to obtain the target user's target The repurchase rate of each candidate merchant among multiple candidate merchants. Among them, the repurchase rate of each candidate merchant represents the long-term demand of the target user for the candidate merchant. The higher the repurchase rate, the greater the long-term demand of the target user for the candidate merchant, and the higher the probability of the target user placing an order for the candidate merchant. Finally, according to the repurchase rates of multiple candidate merchants, the target merchant is recommended to the target user, so as to meet the user's long-term needs, improve the long-term revenue of the server, and improve the user's long-term operation efficiency on the server homepage.

Referring to FIG. 5 , FIG. 5 is a schematic diagram of interaction between a server and a client according to an embodiment of the present application. The server shown in FIG. 5 is used to implement the method for recommending a merchant in any of the above method embodiments. As shown in Figure 5:

The server mainly includes a data storage module and a merchant recommendation module. The data storage module stores user order logs, user click logs, and merchant exposure logs, as well as merchant features and user features. The merchant recommendation module includes a repurchase rate prediction model and a delivery service interface. The repurchase rate prediction model is used to receive the user characteristics of the target user and the merchant characteristics of the candidate merchant, and output the repurchase rate of the target user for the candidate merchant. The delivery service interface outputs the target merchant information in the candidate merchants to the client, and receives the client's actions such as clicking and placing orders for the target merchant.

Specifically, as shown in Figure 5, the data storage module sequentially sends log information such as user order logs, user click logs, and business exposure logs into KV (key-value ) module, the data storage module sends user features and merchant features into the KV module through the data warehouse tool HIVE and the distributed computing model MapReduce in turn. In this way, there are multiple key-value pairs composed of user characteristics and merchant characteristics in the KV module. When applied online, the repurchase rate prediction model reads the user characteristics of the target user and the merchant characteristics of the candidate merchants from the KV module, thereby outputting the repurchase rate of the candidate merchants.

In addition, as shown in Figure 5, the client also sends the user's click behavior, ordering behavior and other information to the data storage module, so that the data storage module generates the user's order log and user click log based on the click behavior, ordering behavior and other information . The server should build a training sample set based on log information such as user order logs, user click logs, and merchant exposure logs, which are used to train and update the repurchase rate prediction model.

Based on the same inventive concept, an embodiment of the present application provides a device for recommending merchants. Referring to FIG. 6 , FIG. 6 is a schematic diagram of an apparatus for recommending a merchant provided by an embodiment of the present application. As shown in Figure 6, the device includes:

The feature obtaining module 61 is used to obtain the user feature of the target user and the respective merchant features of a plurality of candidate merchants;

A repurchase rate obtaining module 62, configured to input the user characteristics and the multiple merchant characteristics into a repurchase rate prediction model to obtain the repurchase rate of the target user for each candidate merchant in the multiple candidate merchants;

A target merchant recommendation module 63, configured to recommend a target merchant to the target user according to the repurchase rate of the target user for each candidate merchant in the plurality of candidate merchants, where the target merchant is the plurality of candidate merchants at least one of the.

In a possible implementation, the device further includes:

The model building module is used to build a repurchase rate prediction model.

In a possible implementation, the model building module includes:

The merchant browsing record sequence sub-module is used to obtain the merchant browsing record sequence of the historical user within the preset time period from the time of placing the order for the historical user who has placed the order;

A training sample establishment sub-module is used to establish a training sample corresponding to the merchant browsing record for each merchant browsing record in the merchant browsing record sequence, and the training sample includes: the merchant feature corresponding to the merchant browsing record, User characteristics and reward value corresponding to the user's repurchase situation, merchant characteristics and user characteristics corresponding to the next merchant browsing record;

The training sample set construction sub-module is used to construct the training sample set according to the corresponding training samples of multiple merchant browsing records;

The model training sub-module is used for training a preset reinforcement learning model based on the training sample set to obtain the repurchase rate prediction model.

In a possible implementation, the model building module further includes:

The reward value determination sub-module is used for each merchant browsing record in the merchant browsing record sequence, in the case that the historical user places an order for the merchant corresponding to the merchant browsing record, determine the user's repurchase situation corresponding to the situation. The reward value is a value greater than zero; in the case that the historical user has not placed an order for the merchant corresponding to the merchant's browsing record, it is determined that the reward value corresponding to the user's repurchase is a value not greater than zero.

In a possible implementation, the reward value determination submodule includes:

an order time difference determining unit, configured to determine the order time difference between the time the historical user places an order this time and the time when the previous order is placed when the historical user places an order for the merchant corresponding to the merchant browsing record;

The reward value determination unit is configured to determine the reward value corresponding to the user's repurchase situation according to the order time difference and a preset reward value function, and the reward value is negatively correlated with the order time difference.

In a possible implementation, the reward value determination unit includes:

The reward value determination subunit is used to determine the reward value corresponding to the user's repurchase situation according to the following formula:

Among them, r represents the reward value corresponding to the user's repurchase situation, C represents the weight adjustment coefficient, and T represents the time difference between the order placement.

In a possible implementation manner, the user characteristics include at least one of the following: the user's consumption preference, the user's geographic location, the user's portrait, and behavior characteristics.

In a possible implementation manner, the merchant characteristics include at least one of the following: merchant ID, category ID, delivery time, sales volume, full discount limit, and customer unit price.

Based on the same inventive concept, another embodiment of the present application provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the steps in the method described in any of the foregoing embodiments of the present application .

Based on the same inventive concept, another embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, the processor implements any of the above-mentioned applications when executed Steps in the methods described in the Examples.

As for the apparatus embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and reference may be made to the partial description of the method embodiment for related parts.

The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the various embodiments may be referred to each other.

Those skilled in the art should understand that the embodiments of the embodiments of the present application may be provided as methods, apparatuses, or computer program products. Accordingly, the embodiments of the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The embodiments of the present application are described with reference to the flowcharts and/or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing terminal equipment to produce a machine that causes the instructions to be executed by the processor of the computer or other programmable data processing terminal equipment Means are created for implementing the functions specified in the flow or flows of the flowcharts and/or the blocks or blocks of the block diagrams.

These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing terminal equipment to operate in a particular manner, such that the instructions stored in the computer readable memory result in an article of manufacture comprising instruction means, the The instruction means implement the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing terminal equipment, so that a series of operational steps are performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby executing on the computer or other programmable terminal equipment The instructions executed on the above provide steps for implementing the functions specified in the flowchart or blocks and/or the block or blocks of the block diagrams.

Although the preferred embodiments of the embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once the basic inventive concepts are known. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications that fall within the scope of the embodiments of the present application.

Finally, it should also be noted that in this document, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply these entities or that there is any such actual relationship or sequence between operations. Moreover, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article or terminal device comprising a list of elements includes not only those elements, but also a non-exclusive list of elements. other elements, or also include elements inherent to such a process, method, article or terminal equipment. Without further limitation, an element defined by the phrase "comprises a..." does not preclude the presence of additional identical elements in the process, method, article or terminal device comprising said element.

A method, device, electronic device, and readable storage medium for recommending a merchant provided by the present application have been described in detail above. The principles and implementations of the present application are described with specific examples. The description is only used to help understand the method of the present application and its core idea; meanwhile, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific embodiments and application scope. In summary, the above , the contents of this specification should not be construed as limiting the application.

Claims (11)

1. A method of recommending merchants, comprising:

obtaining user characteristics of a target user and merchant characteristics of a plurality of candidate merchants respectively;

inputting the user characteristics and the merchant characteristics into a repurchase rate prediction model to obtain the repurchase rate of the target user for each candidate merchant in the candidate merchants;

recommending a target merchant to the target user according to the repurchase rate of the target user for each candidate merchant in the candidate merchants, wherein the target merchant is at least one of the candidate merchants.

2. The method of claim 1, further comprising constructing a repurchase rate prediction model.

3. The method of claim 2, wherein the constructing a repurchase rate prediction model comprises:

aiming at the historical users who have placed an order, acquiring a merchant browsing record sequence of the historical users in a preset time period from the order placing time;

aiming at each merchant browsing record in the merchant browsing record sequence, establishing a training sample corresponding to the merchant browsing record, wherein the training sample comprises: the merchant characteristics and the user characteristics corresponding to the merchant browsing records, the reward value corresponding to the user re-purchase condition and the merchant characteristics and the user characteristics corresponding to the next merchant browsing records are recorded;

constructing a training sample set according to training samples corresponding to multiple merchant browsing records;

and training a preset reinforcement learning model based on the training sample set to obtain the repurchase rate prediction model.

4. The method of claim 3, further comprising:

for each merchant browsing record in the sequence of merchant browsing records:

under the condition that the historical user places an order for the merchant corresponding to the browsing record of the merchant, determining that the reward value corresponding to the user re-purchasing condition is a value larger than zero;

and under the condition that the historical user does not place an order for the merchant corresponding to the browsing record of the merchant, determining that the reward value corresponding to the user re-purchasing condition is a value which is not more than zero.

5. The method according to claim 4, wherein in the case that the historical user places an order for the merchant corresponding to the merchant browsing record, determining the reward value corresponding to the user who purchases again comprises:

under the condition that the historical user places an order for a merchant corresponding to the merchant browsing record, determining an order placing time difference between the order placing time of the historical user at this time and the order placing time of the historical user at the last time;

and determining an incentive value corresponding to the user re-purchasing condition according to the order time difference and a preset incentive value function, wherein the incentive value is negatively related to the order time difference.

6. The method according to claim 5, wherein determining the reward value corresponding to the user re-purchase condition according to the ordering time difference and a preset reward value function comprises:

determining the reward value corresponding to the user re-purchase condition according to the following formula:

wherein r represents an award value corresponding to the user re-purchase condition, C represents a weight adjusting coefficient, and T represents the ordering time difference.

7. The method of any of claims 1-6, wherein the user characteristics include at least one of: a consumption preference of the user, a geographic location of the user, a user representation, and a behavioral characteristic.

8. The method of any of claims 1-6, wherein the merchant characteristics include at least one of: merchant ID, item class ID, delivery duration, sales volume, full reduction, and customer unit price.

9. An apparatus for recommending merchants, the apparatus comprising:

the characteristic obtaining module is used for obtaining the user characteristic of the target user and the merchant characteristics of the candidate merchants;

the purchase-resuming rate obtaining module is used for inputting the user characteristics and the merchant characteristics into a purchase-resuming rate prediction model to obtain the purchase-resuming rate of the target user for each candidate merchant in the candidate merchants;

and the target merchant recommending module is used for recommending a target merchant to the target user according to the repurchase rate of the target user for each candidate merchant in the candidate merchants, wherein the target merchant is at least one of the candidate merchants.

10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 8.

11. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executed implements the steps of the method according to any of claims 1 to 8.

Images