CN116485503A

CN116485503A - Commodity combination recommendation method, device, equipment and medium thereof

Info

Publication number: CN116485503A
Application number: CN202310613899.3A
Authority: CN
Inventors: 钟媛媛
Original assignee: Guangzhou Shangyan Network Technology Co ltd
Current assignee: Guangzhou Shangyan Network Technology Co ltd
Priority date: 2023-05-26
Filing date: 2023-05-26
Publication date: 2023-07-25

Abstract

The application relates to a commodity combination recommendation method, a device, equipment and a medium thereof in the technical field of electronic commerce, wherein the method comprises the following steps: acquiring a total commodity combination set corresponding to a target commodity; determining a corresponding confidence interval upper limit value based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the full commodity combination set; generating sampling distribution based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the full commodity combination set, and determining sampling scores from the sampling distribution; and summarizing after matching the confidence interval upper limit value with the sampling score corresponding to the preset weight, obtaining the exploration score corresponding to each commodity combination in the total commodity combination set, and selecting the recommended commodity combination with the exploration score meeting the preset condition. The method and the system can be used for recommending the commodity combination based on the exploration scores obtained by fusing the quantitative values of the benefits generated after the multi-dimensional exploration commodity combination exposure.

Description

Commodity combination recommendation method, device, equipment and medium thereof

Technical Field

The present disclosure relates to the field of electronic commerce technologies, and in particular, to a method for recommending a combination of commodities, and a corresponding apparatus, computer device, and computer readable storage medium thereof.

Background

The commodity combination recommendation is a popularization mode of high-frequency use in an e-commerce platform, particularly in an e-commerce platform based on an independent station, new commodities or new users of each independent station can possibly not provide proper new commodity recommendation for the new users because the new commodities and the new users lack history data and need to realize cold start of recommendation service, a high-quality commodity combination is determined according to priori knowledge, a plurality of commodities with associated access, particularly associated purchasing relationship, form the high-quality commodity combination, and when a user accesses one commodity, other commodities are recommended for the user, so that the barriers faced by cold start can be overcome, and the aim of effective marketing and popularization is achieved.

In the conventional technology, the co-occurrence opportunity of each commodity purchased by the same user is predicted according to user behavior data of each commodity according to various commodity similarity algorithms, and two commodities with higher co-occurrence opportunity are identified as a basic commodity combination. The algorithm is characterized in that the probability that the identified commodity combination can generate benefits is difficult to determine because the algorithm is identified based on two commodities at a time, and the number of recommended commodities provided at a time is limited, so that good recommendation effect is difficult to obtain.

In view of the defects of the traditional technology, the applicant has long been engaged in research in the related field, and is in order to solve the problem in the field of electronic commerce, so a new way is developed.

Disclosure of Invention

It is a primary object of the present application to solve at least one of the above problems and provide a commodity combination recommendation method and corresponding apparatus, computer device, and computer readable storage medium.

In order to meet the purposes of the application, the application adopts the following technical scheme:

one of the purposes of the present application is to provide a commodity combination recommendation method, which comprises the following steps:

acquiring a total commodity combination set corresponding to a target commodity, wherein the total commodity combination set comprises a plurality of commodity combinations, and each commodity combination comprises the target commodity and other commodities;

determining a corresponding confidence interval upper limit value based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, wherein the confidence interval upper limit value represents the confidence degree of the corresponding commodity combination estimated purchased by the user;

generating sampling distribution based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, and determining sampling scores from the sampling distribution, wherein the sampling scores represent the possibility that the corresponding commodity combination is purchased by a user;

And summarizing after matching the confidence interval upper limit value with the sampling score corresponding to the preset weight, obtaining the exploration score corresponding to each commodity combination in the total commodity combination set, and selecting the recommended commodity combination with the exploration score meeting the preset condition.

In a further embodiment, before acquiring the total commodity combination set corresponding to the target commodity, the method includes the following steps:

acquiring historical access behavior data of each user of the same online store, and determining a plurality of commodities related to the historical access behaviors of the user from the historical access behavior data to form a commodity set of the user, wherein the historical access behaviors comprise adding the commodities to a shopping cart and/or purchasing the commodities;

aiming at the commodity set of each user, combining commodities in the commodity set according to any multiple commodities, and exhausting all possible commodity combinations in the commodity set to form the commodity set of the user;

the commodity combinations in the user combination set of all users are de-registered and constructed into a full combination set.

In a further embodiment, the determining the corresponding upper limit value of the confidence interval based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set includes the following steps:

Determining a corresponding profit average and a confidence coefficient according to the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, wherein the profit average represents the average purchase rate of the commodity combination, and the confidence coefficient represents the reliability degree of the profit average;

and calculating the upper limit value of the confidence coefficient interval according to the income average value and the confidence coefficient.

In a further embodiment, a sampling distribution is generated based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, and a sampling score is determined from the sampling distribution, which includes the following steps:

determining distribution parameters according to the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, and constructing corresponding sampling distribution;

sample scores are randomly generated from the sample distribution.

acquiring a historical commodity combination set, and calculating a corresponding confidence interval upper limit value and a sampling score according to the exposure frequency and the purchase frequency corresponding to each historical commodity combination;

determining the weights corresponding to the upper limit values and the sampling scores of a plurality of groups of confidence intervals by adopting a grid search algorithm, respectively matching the upper limit values and the sampling scores of the confidence intervals corresponding to each historical commodity combination with each group of weights, and summarizing to obtain the exploration scores corresponding to each historical commodity combination under each group of weights;

Determining a corresponding positive sample and a negative sample according to whether each historical commodity combination is clicked by a user after exposure;

and determining the performance evaluation value corresponding to each group of weights according to the exploration scores corresponding to each positive sample and each negative sample under each group of weights, and screening out the weight of the optimal group of the performance evaluation values.

In a further embodiment, selecting a recommended commodity combination with an exploration score meeting a preset condition includes the following steps:

determining an exploration probability and a complementary utilization probability from a preset probability space;

triggering and executing the commodity combination with the exploration probability, wherein the exploration scores in the total commodity combination set meet the preset conditions, and the commodity combination is selected as a recommended commodity combination;

and executing selection of a historical commodity combination with the average benefit meeting a preset condition in the historical commodity combination set by using probability triggering as a recommended commodity combination, wherein the historical commodity combination set comprises a plurality of historical exposed historical commodity combinations, each historical commodity combination comprises the target commodity and other commodities, and the average benefit represents the benefit obtained by the average historical commodity combination after exposure.

In a further embodiment, before determining the exploration probability and the utilization probability complementary to the exploration probability from the preset probability space, the method includes the following steps:

Acquiring exposure frequency, click frequency and conversion frequency corresponding to each historical commodity combination in the historical commodity combination set, wherein each historical commodity combination comprises the target commodity and other commodities;

and summarizing the click frequency and the conversion frequency after matching with corresponding preset weights, and obtaining average benefits compared with the exposure frequency.

On the other hand, the commodity combination recommending device provided in accordance with one of the purposes of the present application comprises a combination acquiring module, a confidence determining module, a sampling determining module and a recommendation selecting module 1400, wherein the combination acquiring module is used for acquiring a total commodity combination set corresponding to a target commodity, and comprises a plurality of commodity combinations, and each commodity combination comprises the target commodity and other commodities; the confidence determining module is used for determining a corresponding confidence interval upper limit value based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, wherein the confidence interval upper limit value represents the confidence degree of the corresponding commodity combination estimated purchased by the user; the sampling determining module is used for generating sampling distribution based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, and determining sampling scores from the sampling distribution, wherein the sampling scores represent the possibility that the corresponding commodity combination is purchased by a user; the recommendation selecting module 1400 is configured to match the confidence interval upper limit value with the sampling score to a preset weight, summarize the confidence interval upper limit value and the sampling score to obtain a search score corresponding to each commodity combination in the total commodity combination set, and select a recommended commodity combination whose search score meets a preset condition.

In a further embodiment, before the combination obtaining module, the method includes: the data acquisition sub-module is used for acquiring historical access behavior data of each user of the same online store, determining that a plurality of commodities related to the historical access behaviors of the user form a commodity set of the user from the historical access behavior data, wherein the historical access behaviors comprise adding the commodities to a shopping cart and/or purchasing the commodities; the commodity set forming sub-module is used for combining commodities in the commodity set according to any multiple commodities aiming at the commodity set of each user, and exhausting all possible commodity combinations in the commodity set to form the commodity set of the user; and constructing a full-quantity combination set sub-module for de-overlapping the commodity combinations in the user combination set of all the users and constructing a full-quantity combination set.

In a further embodiment, the confidence determination module includes: the profit average value and confidence calculation submodule is used for determining corresponding profit average values and confidence values according to the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, wherein the profit average values represent the average purchase rate of the commodity combinations, and the confidence values represent the reliability of the profit average values; and the confidence calculation sub-module is used for calculating the upper limit value of the confidence interval according to the profit average value and the confidence.

In a further embodiment, the sample determination module includes: the sampling distribution construction submodule is used for determining distribution parameters according to the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set and constructing corresponding sampling distribution; and the score generation sub-module is used for randomly generating sampling scores from the sampling distribution.

In a further embodiment, before combining the acquisition modules, the method comprises: the confidence interval upper limit value and sampling score calculation sub-module is used for acquiring a historical commodity combination set and calculating a corresponding confidence interval upper limit value and sampling score according to the exposure frequency and the purchase frequency corresponding to each historical commodity combination; the exploration score determining submodule is used for determining weights corresponding to a plurality of groups of confidence interval upper limit values and sampling scores by adopting a grid search algorithm, respectively matching the confidence interval upper limit values and the sampling scores corresponding to each historical commodity combination with each group of weights, and then summarizing to obtain exploration scores corresponding to each historical commodity combination under each group of weights; the positive and negative sample determining submodule is used for determining corresponding positive samples and negative samples according to whether each historical commodity combination is clicked by a user after exposure; and the weight group screening sub-module is used for determining the performance evaluation value corresponding to each group of weights according to the exploration scores corresponding to each positive sample and each negative sample under each group of weights, and screening out the weight of the optimal group of the performance evaluation values.

In a further embodiment, the recommendation selection module 1400 includes: the probability determination submodule is used for determining the exploration probability and the complementary utilization probability from a preset probability space; the exploration probability trigger execution sub-module is used for triggering and executing the commodity combination, the exploration scores of which meet the preset conditions, in the total commodity combination set to be selected as the recommended commodity combination according to the exploration probability; and the probability triggering execution sub-module is used for executing the historical commodity combination which is selected from the historical commodity combination set and has average profits meeting preset conditions by using the probability triggering as a recommended commodity combination, wherein the historical commodity combination set comprises a plurality of historical commodity combinations subjected to historical exposure, each historical commodity combination comprises the target commodity and other commodities, and the average profits represent the profits obtained by the average historical commodity combination after exposure.

In a further embodiment, before the confidence interval upper limit and the sampling score, the method includes: the numerical value acquisition unit is used for acquiring exposure frequency, click frequency and conversion frequency corresponding to each historical commodity combination in the historical commodity combination set, and each historical commodity combination comprises the target commodity and other commodities; and the average gain obtaining unit is used for summarizing the click frequency and the conversion frequency after matching with the corresponding preset weights, and obtaining average gain compared with the exposure frequency.

In yet another aspect, a computer device is provided, adapted for one of the objects of the present application, comprising a central processor and a memory, the central processor being adapted to invoke the steps of running a computer program stored in the memory to perform the merchandise combination recommendation method described herein.

In yet another aspect, a computer readable storage medium adapted to another object of the present application is provided, in which a computer program implemented according to the product combination recommendation method is stored in the form of computer readable instructions, and the computer program when executed by a computer is executed by invoking the computer program to perform the steps included in the method.

The technical solution of the present application has various advantages, including but not limited to the following aspects:

according to the method, the corresponding upper limit value of the confidence interval is determined based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set corresponding to the target commodity, the confidence interval upper limit value represents the confidence degree of the corresponding commodity combination estimated purchased by the user, sampling distribution is generated, sampling scores are determined from the sampling distribution, the sampling scores represent the possibility of the corresponding commodity combination purchased by the user, and then the confidence interval upper limit value and the sampling scores are summarized after being matched with the corresponding preset weights, so that the corresponding exploration scores are obtained, and the commodity combination with the better exploration scores is selected as the recommended commodity combination. On one hand, the search score obtained by fusing the upper limit value of the confidence interval and the quantized value of the score can accurately estimate the possibility that the commodity combination is purchased by the user, and the search score can adapt to the real-time exposure frequency and the purchase frequency for updating, namely, the search score can dynamically adapt to the purchase preference of the user, on the other hand, the lightweight operation is efficient, is suitable for online recommendation scenes with strong real-time performance, and the search score is used for rapidly recommending the estimated high-quality commodity combination purchased by the user, so that more commodity combinations can obtain reasonable exposure opportunities.

Drawings

The foregoing and/or additional aspects and advantages of the present application will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:

FIG. 1 is a flow chart of an exemplary embodiment of a product portfolio recommendation method of the present application;

FIG. 2 is a flow diagram of constructing a full-scale portfolio in an embodiment of the present application;

FIG. 3 is a flowchart illustrating a method for determining a confidence interval upper limit value according to an embodiment of the present application;

FIG. 4 is a schematic flow chart of determining a sampling score according to an embodiment of the present application;

FIG. 5 is a flow chart of screening out weights for evaluating an optimal set in an embodiment of the present application;

FIG. 6 is a flow chart of determining recommended commodity combinations using exploration and utilization balancing strategies in an embodiment of the present application;

FIG. 7 is a flow chart of determining average benefit in an embodiment of the present application;

FIG. 8 is a schematic block diagram of a product combination recommendation device of the present application;

fig. 9 is a schematic structural diagram of a computer device used in the present application.

Detailed Description

Embodiments of the present application are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary only for the purpose of illustrating the present application and are not to be construed as limiting the present application.

As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless expressly stated otherwise, as understood by those skilled in the art. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. The term "and/or" as used herein includes all or any element and all combination of one or more of the associated listed items.

It will be understood by those skilled in the art that all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs unless defined otherwise. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

As used herein, "client," "terminal device," and "terminal device" are understood by those skilled in the art to include both devices that include only wireless signal receivers without transmitting capabilities and devices that include receiving and transmitting hardware capable of two-way communication over a two-way communication link. Such a device may include: a cellular or other communication device such as a personal computer, tablet, or the like, having a single-line display or a multi-line display or a cellular or other communication device without a multi-line display; a PCS (Personal Communications Service, personal communication system) that may combine voice, data processing, facsimile and/or data communication capabilities; a PDA (Personal Digital Assistant ) that can include a radio frequency receiver, pager, internet/intranet access, web browser, notepad, calendar and/or GPS (Global Positioning System ) receiver; a conventional laptop and/or palmtop computer or other appliance that has and/or includes a radio frequency receiver. As used herein, "client," "terminal device" may be portable, transportable, installed in a vehicle (aeronautical, maritime, and/or land-based), or adapted and/or configured to operate locally and/or in a distributed fashion, at any other location(s) on earth and/or in space. As used herein, a "client," "terminal device," or "terminal device" may also be a communication terminal, an internet terminal, or a music/video playing terminal, for example, a PDA, a MID (MobileInternet Device ), and/or a mobile phone with music/video playing function, or may also be a device such as a smart tv, a set top box, or the like.

The hardware referred to by the names "server", "client", "service node" and the like in the present application is essentially an electronic device having the performance of a personal computer, and is a hardware device having necessary components disclosed by von neumann's principle, such as a central processing unit (including an arithmetic unit and a controller), a memory, an input device, and an output device, and a computer program is stored in the memory, and the central processing unit calls the program stored in the external memory to run in the memory, executes instructions in the program, and interacts with the input/output device, thereby completing a specific function.

It should be noted that the concept of "server" as referred to in this application is equally applicable to the case of a server farm. The servers should be logically partitioned, physically separate from each other but interface-callable, or integrated into a physical computer or group of computers, according to network deployment principles understood by those skilled in the art. Those skilled in the art will appreciate this variation and should not be construed as limiting the implementation of the network deployment approach of the present application.

One or several technical features of the present application, unless specified in the plain text, may be deployed either on a server to implement access by remotely invoking an online service interface provided by the acquisition server by a client, or directly deployed and run on the client to implement access.

The neural network model cited or possibly cited in the application can be deployed on a remote server and used for implementing remote call on a client, or can be deployed on a client with sufficient equipment capability for direct call unless specified in a clear text, and in some embodiments, when the neural network model runs on the client, the corresponding intelligence can be obtained through migration learning so as to reduce the requirement on the running resources of the hardware of the client and avoid excessively occupying the running resources of the hardware of the client.

The various data referred to in the present application, unless specified in the plain text, may be stored either remotely in a server or in a local terminal device, as long as it is suitable for being invoked by the technical solution of the present application.

Those skilled in the art will appreciate that: although the various methods of the present application are described based on the same concepts so as to be common to each other, the methods may be performed independently, unless otherwise indicated. Similarly, for each of the embodiments disclosed herein, the concepts presented are based on the same inventive concept, and thus, the concepts presented for the same description, and concepts that are merely convenient and appropriately altered although they are different, should be equally understood.

The various embodiments to be disclosed herein, unless the plain text indicates a mutually exclusive relationship with each other, the technical features related to the various embodiments may be cross-combined to flexibly construct a new embodiment, so long as such combination does not depart from the inventive spirit of the present application and can satisfy the needs in the art or solve the deficiencies in the prior art. This variant will be known to the person skilled in the art.

The commodity combination recommendation method can be programmed into a computer program product and can be deployed in a client or a server for operation, for example, in the exemplary application scenario of the application, the commodity combination recommendation method can be deployed in a server of an electronic commerce platform, and therefore the method can be executed by accessing an interface opened after the computer program product is operated and performing man-machine interaction with a process of the computer program product through a graphical user interface.

Referring to fig. 1, in an exemplary embodiment, the method for recommending commodity combinations of the present application includes the following steps:

step S1100, acquiring a full-quantity commodity combination set corresponding to a target commodity, wherein the full-quantity commodity combination set comprises a plurality of commodity combinations, and each commodity combination comprises the target commodity and other commodities;

In the application scenario of the demonstration program, a recommended commodity combination list required for carrying out commodity recommendation service according to target commodities is generated for an independent station of one deployment online shop, and a plurality of high-quality commodity combinations are recommended in the list, wherein each commodity combination comprises the target commodities and other commodities. Accordingly, the recommended product combination list can be constructed using access data generated by a user accessing the online store of the independent station as base data. The target commodity is a commodity related to the current access behavior of the user for accessing the online store, and the current access behavior of the user can be any one or more of adding the commodity into a shopping cart, purchasing the commodity, clicking the commodity, browsing the commodity and searching the commodity.

In one embodiment, the embedded point codes are set in the visual page displayed at the front end of the independent station, which is interacted with by the user, so that various access behaviors of the user can be obtained, and the access behaviors correspondingly generated are data, wherein the access data comprise accessed commodities, access time and access users.

The total commodity combination set is formed by acquiring a commodity set of each user, wherein the commodity set comprises a plurality of commodities related to the historical access behaviors of the user, combining any commodity number of the commodities in the commodity set, enumerating all commodity combinations to form the user combination set, and combining all the user combination sets after de-duplication.

For ease of understanding, the exemplary examples: for each commodity in the commodity set of each user, according to the number of any exhaustive commodity, the commodity number is more than or equal to 2, and a plurality of commodities with the corresponding commodity number are selected for combination to form a commodity combination. For example, when three commodities such as { A ] exist in a commodity set; b, a step of preparing a composite material; c, four commodity combinations, namely { AB }; BC; an AC; ABC }, and so on, by enumeration, various possible commodity combinations corresponding to each user are obtained. After all commodity combinations are enumerated for each commodity set, a user combination set is obtained, wherein all commodity combinations corresponding to the user are contained.

The historical access behavior includes any one or more of adding merchandise to a shopping cart, purchasing merchandise, and browsing multiple merchandise corresponding during the same session.

And the commodity set is formed by extracting all commodities visited by each user. The same commodity can be de-duplicated as needed, considering that the same commodity can be accessed by multiple users at the same time or multiple times by the same user.

It will be appreciated that there may be identical combinations of products in the user combinations of different users, and that for convenience in subsequent data processing, all user combinations may be combined into the same full combination set, and that for identical product combinations in different user combinations, deduplication may be performed such that in the full combination set, the same product combination occurs only once.

Step 1200, determining a corresponding confidence interval upper limit value based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, wherein the confidence interval upper limit value represents the confidence degree of the corresponding commodity combination estimated purchased by the user;

and acquiring all access behavior data corresponding to the exposed commodity combination due to searching, clicking, browsing, shopping cart adding and purchasing behavior of a user in a preset time before the current time in real time through the embedded point code, determining the frequency of all the access behavior data, namely the exposure frequency, wherein the exposure frequency can be 0, and indicating that the corresponding commodity combination is not exposed once when the exposure frequency is 0.

And acquiring all the access behavior data corresponding to the purchase behavior of the user in a preset time before the current time in real time through the embedded point code, determining the frequency of all the access behavior data, namely the purchase frequency, wherein the purchase frequency can be 0, and when the purchase frequency is 0, the corresponding commodity combination is not exposed at one time and then purchased by the user.

The preset duration can be any value of one week, one month, one quarter, half year and the like, and the access behavior data effectively reflects the latest dynamics of commodity combination through time constraint with timeliness.

And calculating a corresponding mean value as a profit mean value and a standard deviation of the mean value as a confidence coefficient based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set by adopting a UCB algorithm, and further adding to obtain the upper limit value of the confidence interval corresponding to each commodity combination. The profit average characterizes the average purchasing rate of the commodity combination, and the confidence represents the reliability degree of the profit average.

It will be appreciated that the larger the average value of the benefit, the smaller the confidence, and the larger the upper limit value of the confidence interval corresponding to the one of the good combinations selected as the good combination, the greater the chance of selecting the good combination, and the less frequently selected good combinations are selected by testing, i.e. exploring, the chance that each good combination can be reasonably selected.

Step S1300, generating a sampling distribution based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, and determining a sampling score from the sampling distribution, wherein the sampling score represents the possibility that the corresponding commodity combination is purchased by a user;

And adopting a Thompson sampling algorithm to maintain a beta distribution, namely the sampling distribution, based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the full commodity combination set as distribution parameters, and generating a random number from the beta distribution as a sampling score corresponding to each commodity combination.

It will be appreciated that the greater the sampling score on which a combination of good is selected as one of the good combinations, the greater the chance of selecting that combination of good, so that each combination of good can have a reasonable chance of being selected.

And step 1400, summarizing after matching the confidence interval upper limit value with the sampling score corresponding to the preset weight, obtaining the exploration score corresponding to each commodity combination in the total commodity combination set, and selecting the recommended commodity combination with the exploration score meeting the preset condition.

It can be understood that the upper limit value of the confidence interval and the sampling score can be adapted to real-time exposure frequency and purchase frequency for updating, namely, the method can be dynamically adapted to the purchase preference of a user, and quantization standards for generating benefits by commodity combination can be provided from different dimensions respectively, so that the quantization values of different dimensions can be fused, and the corresponding weights matched with the upper limit value of the confidence interval and the sampling score are multiplied and summed for measuring the acting force of the upper limit value of the confidence interval and the sampling score before fusion, so as to obtain the corresponding exploration score. The preset weights corresponding to the confidence interval upper limit and the sampling score can be an empirical threshold or an experimental threshold, and can be set by a person skilled in the art according to requirements.

It may be appreciated that the exploration score characterizes a likelihood that the corresponding commodity combination can obtain a benefit after exposure, and accordingly, in one embodiment, one or more commodity combinations with the total commodity combination centralized exploration score meeting a preset threshold are selected as recommended commodity combinations, and all the recommended commodity combinations are further summarized to construct a recommended commodity combination list to be pushed to a user accessing the target commodity. In another embodiment, according to the exploration score corresponding to each commodity combination in the total commodity combination set, all commodity combinations are ranked in order from high score to low score, one or more commodity combinations with the top ranking are selected as recommended commodity combinations, and further, all recommended commodity combinations are summarized to form a recommended commodity combination list and pushed to the user accessing the target commodity.

From the above embodiments, the present application achieves various technical advantages, including but not limited to:

Referring to fig. 2, in a further embodiment, before step S1100, the step of obtaining the total merchandise combination set corresponding to the target merchandise includes the following steps:

step S1000, acquiring historical access behavior data of each user of the same online store, and determining that a plurality of commodities related to the historical access behaviors of the user form a commodity set of the user from the historical access behavior data, wherein the historical access behaviors comprise adding the commodities to a shopping cart and/or purchasing the commodities;

in this embodiment, the acquisition of access behavior data of the user based on the same online store is mainly considered, so that the obtained recommended product combination list is more suitable for a product recommendation service serving an independent station where the online store is located. This is done in consideration of online stores based on independent stations, when the online stores sell goods with each other when the electronic commerce is independent stations, the difference of the goods can be large, and the action data of the correlation of the goods lacks necessary correlation, so that the collection of the access action data is limited to the same online store, and the obtained recommended goods combination list is more targeted.

Therefore, access behavior data corresponding to a certain access behavior event generated by the online store in a specific time range can be obtained from a user behavior database of an independent station where the online store is located, then the access behavior data of each user is analyzed, access time and commodities are extracted, all commodities visited by each user in history are constructed into the same commodity set by taking the user as a unit, and preferably, the commodities in the commodity set are ordered according to the access time.

The access behavior event, in the embodiment of the present application, recommends to use an event of adding a commodity to a shopping cart and/or an event of adding a commodity to a purchase order, the former mainly because of representing a strong desire of a user to purchase the corresponding commodity, and the latter mainly because of representing a fact action of the user to make a purchase, and both can be seen to represent the requirement of the user to purchase the corresponding commodity, thus having an information reference value required as basic data.

The historical access behavior data generally refer to various historical data generated by various access behaviors of corresponding users, the historical data can be obtained by setting a buried point code in a visual page displayed at the front end of the independent station, which is interacted with by the users, for example, for the ordering and purchasing behaviors of the users, the corresponding orders contain commodities selected and purchased by the users, so that mapping relation data between the users and the commodities purchased by the users is obtained; similarly, for the action of adding the commodity selected by the user to the shopping cart, the user browses the actions corresponding to the multiple commodities in the same session, and the like, the actions are triggered and generated at the front end layer of the independent station. For various behavior data, one or more of the behavior data may be selected as the basic data required to construct the recommended combination list.

In one embodiment, the time of the access behavior data to be acquired may be constrained, for example, the historical access data of a preset market in the near term may be acquired, the preset time may be any value of one week, one month, one quarter, half year, etc., and the time constraint with timeliness may enable the historical access behavior data to effectively reflect the latest dynamics of the commodity combination.

Step S1001, for each commodity set of the user, combining the commodities therein according to any plurality of commodities, and exhausting all possible commodity combinations in the commodity set to form the commodity set of the user;

the commodity set of each user can comprise massive commodity items, one commodity combination can be constructed by combining any more than two commodities with any commodity number, and according to the principle, all commodity combinations formed by all commodity combinations in the commodity set of each user can be determined by enumerating any commodity number commodity combinations of all commodity sets in the commodity set of each user, so that the user combination set of the corresponding user.

Step S1002, the commodity combinations in the user combination sets of all users are de-overlapped and constructed into a full combination set.

Although there is no duplicate product combination in each user combination set, there may be duplicate product combinations in different user combination sets, a full combination set may be constructed in consideration of the need to evaluate for each unique product combination, all product combinations in all user combination sets are added to the full combination set, and only one of them is reserved for the identical product combination, and deduplication is achieved, so that each product combination in the full combination set has uniqueness.

According to the embodiment, the deep data processing of the access behavior data of all users of the same online store is realized, and the user combination set and the full combination set are constructed, wherein the user combination set comprises enumeration of all possible commodity combinations in the historical behavior data, the number of commodities in the commodity combinations is not additionally limited, and the richness of the commodity combinations of the full combination set is ensured.

Referring to fig. 3, in a further embodiment, step S1200, determining a corresponding upper limit value of the confidence interval based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, includes the following steps:

Step S1210, determining a corresponding profit average and a confidence coefficient according to the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, wherein the profit average represents the average purchase rate of the commodity combination, and the confidence coefficient represents the reliability degree of the profit average;

according to the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, calculating a corresponding average value as a profit average value, and the exemplary formula is as follows:

wherein: income Average is the mean value of benefits, count _exposure For the exposure frequency of the commodity combination, i.e. the number of times the target commodity is exposed together with other commodities in the commodity combination, count _Buy The purchase frequency of the commodity combination is the number of times the target commodity is purchased by the user after being exposed with other commodities in the commodity combination.

Calculating the confidence of the commodity combination according to the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, wherein an exemplary formula is as follows:

wherein: confidence Level is Confidence, T is the sum of exposure frequency corresponding to each commodity combination in the total commodity combination set, namely the sum of exposure times of the target commodity corresponding to other commodities in all commodity combinations, and T is _t The exposure frequency of the commodity combination is the number of times the target commodity is exposed together with other commodities in the commodity combination.

Step S1220, calculating the upper limit value of the confidence interval according to the profit average and the confidence.

And determining the average value of the benefits and the confidence corresponding to each commodity combination in the full commodity combination set based on the above process, adding the average value of the benefits and the confidence, and calculating the upper limit value of the confidence interval.

In this embodiment, based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, a corresponding profit average value and a corresponding confidence coefficient are calculated and added to obtain a corresponding confidence interval upper limit value, so that the confidence coefficient estimated by the user for purchasing the commodity combination can be accurately represented, and the confidence interval upper limit value can be correspondingly dynamically updated due to the update of the real-time exposure frequency and the purchase frequency, so that the method has timeliness, and can be suitable for the purchase preference of mass users in the sales market.

Referring to fig. 4, in a further embodiment, step S1300, generating a sampling distribution based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, and determining a sampling score from the sampling distribution, includes the following steps:

Step S1310, determining distribution parameters according to the exposure frequency and the purchase frequency corresponding to each commodity combination in the full commodity combination set, and constructing corresponding sampling distribution;

in one embodiment, the exposure frequency corresponding to each commodity combination in the total commodity combination set, that is, the number of times that the target commodity is exposed together with other commodities in the commodity combination, is used as a distribution parameter α, and the corresponding purchase frequency, that is, the number of times that the target commodity is purchased by a user after being exposed together with other commodities in the commodity combination, is used as a distribution parameter β, and the α and β are regarded as obeying the bat distribution and respectively used as parameters thereof, so as to generate bat (α, β) as a sampling distribution corresponding to each commodity combination.

In another embodiment, the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set are subtracted, that is, the number of times that the target commodity is exposed together with other commodities in the commodity combination and not purchased by a user is used as a distribution parameter α, and the purchase frequency, that is, the number of times that the target commodity is purchased together with other commodities after being exposed in the commodity combination and then purchased by the user is used as a distribution parameter β, and the α and the β are regarded as obeying bat distribution and respectively used as parameters thereof, so as to generate bat (α, β) as a sampling distribution corresponding to each commodity combination.

Step S1320, randomly generating a sampling score from the sampling distribution.

Based on the sampling distribution corresponding to each commodity combination, a random number is randomly generated as the sampling score corresponding to each commodity combination.

In this embodiment, based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the full commodity combination set as distribution parameters, a sampling distribution is generated, and a sampling score is randomly generated from the sampling distribution, so that the probability that the commodity combination is purchased by a user can be accurately represented, and the sampling score is correspondingly dynamically updated due to the update of the real-time exposure frequency and the purchase frequency, so that the method has timeliness, and can adapt to the purchase preference of a large number of users in the sales market.

Referring to fig. 5, in a further embodiment, before step S1100, the step of obtaining the total merchandise combination set corresponding to the target merchandise includes the following steps:

step S1010, acquiring a historical commodity combination set, and calculating a corresponding confidence interval upper limit value and a sampling score according to the exposure frequency and the purchase frequency corresponding to each historical commodity combination;

the historical commodity combination set can be artificially constructed by operators of the electronic commerce platform and/or merchant users of the online shops based on the commodities of the online shops, or constructed according to the steps S1000-S1002, and it is pointed out that the historical commodity combinations are exposed, and corresponding exposure frequency and purchase frequency can be obtained according to the implementation of the step S1200.

The calculation of the confidence interval upper limit value and the sampling score can be correspondingly implemented according to steps S1210-S1220 and steps S1310-S1320.

Step S1020, determining the weights corresponding to the upper limit values and the sampling scores of a plurality of groups of confidence intervals by adopting a grid search algorithm, respectively matching the upper limit values and the sampling scores of the confidence intervals corresponding to each historical commodity combination with each group of weights, and summarizing to obtain the exploration scores corresponding to each historical commodity combination under each group of weights;

and for the weights corresponding to the exposure frequency and the purchase frequency of each historical commodity combination in the historical commodity combination set, confirming a plurality of groups of weight correspondence by adopting a grid search mode to calculate, and for each group of weights, multiplying the exposure frequency and the purchase frequency correspondence corresponding to each historical commodity combination by the weights in the group of weights respectively and then adding the weights to obtain the exploration scores corresponding to each commodity combination under each group of weights.

Step S1030, determining a corresponding positive sample and a negative sample according to whether each historical commodity combination is clicked by a user after exposure;

and taking the historical commodity combination after exposure as a positive sample, and taking the historical commodity combination after exposure as a negative sample without being clicked by the user.

Step S1040, determining performance evaluation values corresponding to each group of weights according to the exploration scores corresponding to each positive sample and each negative sample under each group of weights, and screening out the weights of the optimal group of the performance evaluation values.

According to the exploration scores corresponding to each positive sample and each negative sample under each group of weights, auc values corresponding to each group of weights are calculated to serve as performance evaluation values, and an exemplary formula is as follows:

wherein:m is the number of positive samples, N is the number of negative samples, P _{Positive sample} Search scoring for positive samples, P _{Negative sample} The negative samples were scored for exploration.

Accordingly, the weight of the corresponding group with the highest performance evaluation value can be screened out according to the performance evaluation value corresponding to each group of weights.

In this embodiment, the performance evaluation value corresponding to each set of weights is determined based on the exploration score corresponding to each historical commodity combination in the historical commodity combination set under each set of weights, so that the weight of the optimal set of performance evaluation values is screened out, and the performance of the corresponding weight can be accurately and quantitatively evaluated.

Referring to fig. 6, in a further embodiment, step S1400 of selecting a recommended product combination with an exploration score satisfying a preset condition includes the following steps:

step S1410, determining an exploration probability and a complementary utilization probability from a preset probability space;

The probability space is preset to be (0, 1) by adopting a search and utilization balance strategy, a smaller random number is randomly generated from the probability space to serve as search probability, and a difference value obtained by subtracting the search probability is taken as utilization probability.

Step S1420, triggering and executing the commodity combination with the exploration probability, wherein the exploration scores in the total commodity combination set meet the preset conditions, and the commodity combination is selected as a recommended commodity combination;

specifically, steps S1100-1400 are performed.

It will be appreciated that when the exploration probability hits, it is an attempt to push to the user a combination of recommended items predicted to be likely to be purchased by the user with a smaller probability exploration.

Step S1430, executing and selecting a historical commodity combination with the average benefit in the historical commodity combination set meeting the preset condition by using probability triggering as a recommended commodity combination, wherein the historical commodity combination set comprises a plurality of historical exposed historical commodity combinations, each historical commodity combination comprises the target commodity and other commodities, and the average benefit represents the benefit obtained by the average historical commodity combination after exposure.

Each historical commodity combination in the historical commodity combination set can be clicked and/or purchased by a user after exposure to generate benefits, so that the benefits can be properly averaged for each historical commodity combination to obtain corresponding average benefits.

In one embodiment, one or more historical commodity combinations with average gains meeting a preset threshold in the historical commodity combination set are selected as recommended commodity combinations, and all the recommended commodity combinations are further summarized to form a recommended commodity combination list to be pushed to the user accessing the target commodity. In another embodiment, according to the exploration score corresponding to each historical commodity combination in the historical commodity combination set, all the historical commodity combinations are ranked in order from high score to low score, one or more historical commodity combinations with the top ranking are screened out to be used as recommended commodity combinations, and further, all the recommended commodity combinations are summarized to construct a recommended commodity combination list and pushed to the user accessing the target commodity.

It will be appreciated that when the utilization probability hits, the recommended commodity combination with the better average value obtained after exposure is pushed to the user with larger probability utilization.

In this embodiment, by adopting the exploration and utilization balance policy, the exploration probability and the utilization probability are set to correspondingly select the corresponding recommended commodity combination, so that a new recommendation attempt can be explored simultaneously, and the known optimal can be utilized to conduct recommendation, so that the balance between the two can be balanced. To some extent, it can ensure that the existing most preferences are not sacrificed while exploring unknown possibilities, thereby making the algorithm more robust and practical.

Referring to fig. 7, in a further embodiment, before determining the exploration probability from the preset probability space and the utilization probability complementary thereto, step S1410 includes the following steps:

step S1401, acquiring exposure frequency, click frequency and conversion frequency corresponding to each historical commodity combination in a historical commodity combination set, wherein each historical commodity combination comprises the target commodity and other commodities;

and acquiring all access behavior data corresponding to the exposed historical commodity combination caused by searching, clicking, browsing, shopping cart adding and purchasing behavior of a user in the preset time before the current time through the embedded point code, and determining the frequency of all the access behavior data, namely the exposure frequency.

And acquiring all access behavior data corresponding to the clicking behaviors of the user within the preset time before the current time through the embedded point code, and determining the frequency of all access behavior data, namely the conversion frequency.

And acquiring all access behavior data corresponding to the purchasing and/or adding of the user to the shopping cart behavior within the preset time before the current time through the embedded point code, and determining the frequency of all the access behavior data, namely the conversion frequency.

Step S1402, summarizing the click frequency and the conversion frequency after matching with the corresponding preset weights, and comparing with the exposure frequency, obtaining average benefit.

The click frequency and the conversion frequency of different dimensions can be fused, and in order to measure the acting force of the click frequency and the conversion frequency before fusion, the corresponding weights matched with the click frequency and the conversion frequency are multiplied respectively and added, so that average benefit is obtained compared with the exposure frequency. The preset weights corresponding to the confidence interval upper limit value and the sampling score can be an empirical threshold or an experimental threshold, and can be set by a person skilled in the art as required, and the weights corresponding to the click frequency and the conversion frequency are exemplified as 0.4 and 0.6.

In this embodiment, based on the exposure frequency, conversion frequency, and click frequency of each historical commodity combination in the historical commodity combination set, a corresponding average benefit is reasonably calculated, so as to accurately reflect the effect of the historical commodity combination after exposure.

Referring to fig. 8, a commodity combination recommending apparatus provided in accordance with one of the purposes of the present application is a functional implementation of a commodity combination recommending method of the present application, and in another aspect, the commodity combination recommending apparatus provided in accordance with one of the purposes of the present application includes a combination acquiring module 1100, a confidence determining module 1200, a sampling determining module 1300, and a recommendation selecting module 1400, where the combination acquiring module 1100 is configured to acquire a total commodity combination set corresponding to a target commodity, and includes a plurality of commodity combinations, each commodity combination including the target commodity and other commodities; the confidence determining module 1200 is configured to determine a corresponding upper limit value of a confidence interval based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, where the upper limit value of the confidence interval characterizes a confidence that the corresponding commodity combination is estimated to be purchased by the user; the sampling determining module 1300 is configured to generate a sampling distribution based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, and determine a sampling score from the sampling distribution, where the sampling score characterizes a likelihood that the corresponding commodity combination is purchased by a user; the recommendation selecting module 1400 is configured to match the confidence interval upper limit value with the sampling score to a preset weight, summarize the confidence interval upper limit value and the sampling score to obtain a search score corresponding to each commodity combination in the total commodity combination set, and select a recommended commodity combination whose search score meets a preset condition.

In a further embodiment, before the combination obtaining module 1100, the method includes: the data acquisition sub-module is used for acquiring historical access behavior data of each user of the same online store, determining that a plurality of commodities related to the historical access behaviors of the user form a commodity set of the user from the historical access behavior data, wherein the historical access behaviors comprise adding the commodities to a shopping cart and/or purchasing the commodities; the commodity set forming sub-module is used for combining commodities in the commodity set according to any multiple commodities aiming at the commodity set of each user, and exhausting all possible commodity combinations in the commodity set to form the commodity set of the user; and constructing a full-quantity combination set sub-module for de-overlapping the commodity combinations in the user combination set of all the users and constructing a full-quantity combination set.

In a further embodiment, the confidence determination module 1200 includes: the profit average value and confidence calculation submodule is used for determining corresponding profit average values and confidence values according to the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, wherein the profit average values represent the average purchase rate of the commodity combinations, and the confidence values represent the reliability of the profit average values; and the confidence calculation sub-module is used for calculating the upper limit value of the confidence interval according to the profit average value and the confidence.

In a further embodiment, the sample determination module 1300 includes: the sampling distribution construction submodule is used for determining distribution parameters according to the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set and constructing corresponding sampling distribution; and the score generation sub-module is used for randomly generating sampling scores from the sampling distribution.

In a further embodiment, before combining the acquisition module 1100, it includes: the confidence interval upper limit value and sampling score calculation sub-module is used for acquiring a historical commodity combination set and calculating a corresponding confidence interval upper limit value and sampling score according to the exposure frequency and the purchase frequency corresponding to each historical commodity combination; the exploration score determining submodule is used for determining weights corresponding to a plurality of groups of confidence interval upper limit values and sampling scores by adopting a grid search algorithm, respectively matching the confidence interval upper limit values and the sampling scores corresponding to each historical commodity combination with each group of weights, and then summarizing to obtain exploration scores corresponding to each historical commodity combination under each group of weights; the positive and negative sample determining submodule is used for determining corresponding positive samples and negative samples according to whether each historical commodity combination is clicked by a user after exposure; and the weight group screening sub-module is used for determining the performance evaluation value corresponding to each group of weights according to the exploration scores corresponding to each positive sample and each negative sample under each group of weights, and screening out the weight of the optimal group of the performance evaluation values.

In order to solve the technical problems, the embodiment of the application also provides computer equipment. As shown in fig. 9, the internal structure of the computer device is schematically shown. The computer device includes a processor, a computer readable storage medium, a memory, and a network interface connected by a system bus. The computer readable storage medium of the computer device stores an operating system, a database and computer readable instructions, the database can store a control information sequence, and when the computer readable instructions are executed by a processor, the processor can realize a commodity combination recommendation method. The processor of the computer device is used to provide computing and control capabilities, supporting the operation of the entire computer device. The memory of the computer device may have stored therein computer readable instructions that, when executed by the processor, may cause the processor to perform the merchandise combination recommendation method of the present application. The network interface of the computer device is for communicating with a terminal connection. It will be appreciated by those skilled in the art that the structure shown in fig. 9 is merely a block diagram of a portion of the structure associated with the present application and is not limiting of the computer device to which the present application applies, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.

The processor in this embodiment is configured to execute specific functions of each module and its sub-module in fig. 8, and the memory stores program codes and various data required for executing the above modules or sub-modules. The network interface is used for data transmission between the user terminal or the server. The memory in the present embodiment stores program codes and data required for executing all modules/sub-modules in the commodity combination recommendation device of the present application, and the server can call the program codes and data of the server to execute the functions of all sub-modules.

The present application also provides a storage medium storing computer readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the merchandise combination recommendation method of any of the embodiments of the present application.

Those skilled in the art will appreciate that implementing all or part of the above-described methods of embodiments of the present application may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed, may comprise the steps of embodiments of the methods described above. The storage medium may be a computer readable storage medium such as a magnetic disk, an optical disk, a Read-Only Memory (ROM), or a random access Memory (Random Access Memory, RAM).

In summary, the method and the device can implement commodity combination recommendation based on the exploration scores obtained by fusing the quantitative values of the benefits generated after multi-dimensional exploration commodity combination exposure, so that the constructed rich commodity combination can obtain reasonable exposure opportunities.

Those of skill in the art will appreciate that the various operations, methods, steps in the flow, actions, schemes, and alternatives discussed in the present application may be alternated, altered, combined, or eliminated. Further, other steps, means, or steps in a process having various operations, methods, or procedures discussed in this application may be alternated, altered, rearranged, split, combined, or eliminated. Further, steps, measures, schemes in the prior art with various operations, methods, flows disclosed in the present application may also be alternated, altered, rearranged, decomposed, combined, or deleted.

The foregoing is only a partial embodiment of the present application, and it should be noted that, for a person skilled in the art, several improvements and modifications can be made without departing from the principle of the present application, and these improvements and modifications should also be considered as the protection scope of the present application.

Claims

1. The commodity combination recommending method is characterized by comprising the following steps of:

2. The commodity combination recommendation method according to claim 1, wherein before obtaining the total commodity combination set corresponding to the target commodity, comprising the steps of:

3. The commodity combination recommendation method according to claim 1, wherein the respective confidence interval upper limit value is determined based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, comprising the steps of:

4. The commodity combination recommendation method according to claim 1, wherein a sampling distribution is generated based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, and a sampling score is determined from the sampling distribution, comprising the steps of:

sample scores are randomly generated from the sample distribution.

5. The commodity combination recommendation method according to claim 1, wherein before obtaining the total commodity combination set corresponding to the target commodity, comprising the steps of:

6. The commodity combination recommendation method according to claim 1, wherein selecting a recommended commodity combination whose search score satisfies a preset condition comprises the steps of:

7. The merchandise combination recommendation method according to claim 6, wherein before determining the exploration probability from the preset probability space and the utilization probability complementary thereto, comprising the steps of:

8. A commodity combination recommendation device, comprising:

the combination acquisition module is used for acquiring a full-quantity commodity combination set corresponding to a target commodity, wherein the full-quantity commodity combination set comprises a plurality of commodity combinations, and each commodity combination comprises the target commodity and other commodities;

the confidence determining module is used for determining a corresponding confidence interval upper limit value based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, wherein the confidence interval upper limit value represents the confidence degree of the corresponding commodity combination estimated purchased by the user;

the sampling determining module is used for generating sampling distribution based on the exposure frequency and the purchase frequency corresponding to each commodity combination in the total commodity combination set, and determining sampling scores from the sampling distribution, wherein the sampling scores represent the possibility that the corresponding commodity combination is purchased by a user;

And the recommendation selection module is used for summarizing after matching the confidence interval upper limit value with the sampling score corresponding to the preset weight to obtain the exploration score corresponding to each commodity combination in the total commodity combination set, and selecting the recommended commodity combination with the exploration score meeting the preset condition.

9. A computer device comprising a central processor and a memory, characterized in that the central processor is arranged to invoke a computer program stored in the memory for performing the steps of the method according to any of claims 1 to 7.

10. A computer-readable storage medium, characterized in that it stores in the form of computer-readable instructions a computer program implemented according to the method of any one of claims 1 to 7, which, when invoked by a computer, performs the steps comprised by the corresponding method.