CN110335123B - Commodity recommendation method, system, computer readable medium and device based on social e-commerce platform - Google Patents
Commodity recommendation method, system, computer readable medium and device based on social e-commerce platform Download PDFInfo
- Publication number
- CN110335123B CN110335123B CN201910626799.8A CN201910626799A CN110335123B CN 110335123 B CN110335123 B CN 110335123B CN 201910626799 A CN201910626799 A CN 201910626799A CN 110335123 B CN110335123 B CN 110335123B
- Authority
- CN
- China
- Prior art keywords
- data set
- commodity
- user
- commodity data
- commodities
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Abstract
The invention discloses a commodity recommendation method, a commodity recommendation system, a computer readable medium and a commodity recommendation device based on a social e-commerce platform, which comprise the following steps of S1: obtaining commodities historically purchased by users in social relations associated with the master user based on the social relation information of the master user to form a candidate commodity data set; and step S2: obtaining a union set according to the candidate commodity data set and the E-commerce recommended commodity data set obtained in the step S1, and recommending commodities to the master user from the union set by using a Bandit algorithm, wherein the recommended commodities are closer to the range in which the master user is interested; therefore, the quality of the commodity data set is high, and the pertinence is relatively strong.
Description
[ technical field ] A method for producing a semiconductor device
The invention relates to the technical field of prediction and calculation, in particular to a commodity recommendation method, a commodity recommendation system, a computer readable medium and a commodity recommendation device based on a social e-commerce platform.
[ background of the invention ]
As the e-commerce network increasingly enters the living field of people, people contact with explosion type massive commodity data every day, and users spend a great deal of time when acquiring data needed by themselves, and when facing huge data materials, the users are difficult to independently screen uninteresting data, so that the information utilization rate of a plurality of platforms is very low;
the personalized recommendation system is generated, in order to improve the accuracy of information, the system carries out recommendation by comprehensively utilizing the interest preference and attribute of the user, the attribute, the content, the classification and the like of the commodity, and a plurality of social electronic commerce platforms begin to use the personalized recommendation system to seek for the maximization of benefits; however, in social e-commerce, when a user does not have any historical behavior, how to recommend the user is a so-called cold start problem, and the problem is also a difficult point faced by various large social e-commerce platforms nowadays.
In the prior art, hot recommendation and similar commodity recommendation through user historical shopping behaviors are generally adopted for solving the cold start problem of the e-commerce platform, but recommendation is only carried out by means of the hot recommendation, and the advantages of the social e-commerce platform are not utilized, so that how to accurately establish a commodity data set on the social e-commerce platform and enable the commodity data set to continuously approach an area really interested by a user is a problem to be solved in the prior art.
[ summary of the invention ]
In order to overcome the problems, the invention provides a commodity recommendation method, a commodity recommendation system, a computer readable medium and a commodity recommendation device based on a social e-commerce platform.
In order to solve the technical problem, the invention provides a commodity recommendation method based on a social e-commerce platform, which comprises the following steps: step S1: obtaining historical purchased commodities of a user with a proxy behavior by a master user to form a commodity data set A; if the commodity data set A is smaller than a preset threshold value, obtaining a commodity data set B formed by historical shopping commodities of a user, of which the main user has information exchange, and supplementing the commodity data set B to the commodity data set A; if the union of the commodity data set A and the commodity data set B is still smaller than a preset threshold value, expanding the union of the commodity data set A and the commodity data set B to obtain a total candidate commodity data set; the device information in purchasing according to the main user and the social relationship can be expanded; or if the device data is not authorized to be read and the device information cannot be acquired, the device data can be expanded through attribute clustering according to the background attribute of the master user or the commodity; or recommending and expanding based on the association rule according to the background attribute of the master user or the commodity; and step S2: step S2: obtaining a union set according to the candidate commodity data set and the E-commerce recommended commodity data set obtained in the step S1, and recommending commodities to a master user from the union set by using a Bandit algorithm; the social relationship users comprise users who have proxy purchasing behavior or only carry out information communication with the master user.
Preferably, in the step S1, obtaining the commodities historically purchased by the users of the social relationship associated with the master user based on the social relationship information of the master user to form the candidate commodity data set, the method includes the following steps: step S101: obtaining historical purchased commodities of a user with a proxy behavior by a master user to form a commodity data set A; step S102: if the commodity data set A is smaller than a preset threshold value, obtaining a commodity data set B formed by historical shopping commodities of a user, of which the main user has information exchange, and supplementing the commodity data set B to the commodity data set A; step S103: and if the union of the commodity data set A and the commodity data set B is still smaller than the preset threshold value, expanding the union of the commodity data set A and the commodity data set B to obtain a total candidate commodity data set.
Preferably, in step S103, the expanding is performed on the basis of the union of the product data set a and the product data set B, and includes: and expanding according to the background attributes of the main user and the commodities in the candidate commodity data set or through background attribute clustering or background attribute association.
Preferably, in the step S2, a union is obtained according to the candidate commodity data set and the e-commerce recommended commodity data set obtained in the step S1, and a Bandit algorithm is used to recommend commodities from the union to the master user, including the following steps: step S201: selecting any number between one (0,1) as epsilon; step S202: randomly recommending the commodities to the user in the union set according to the epsilon probability and acquiring the return rate of the commodities; step S203: and confirming that the commodities with the return rate higher than the preset return rate value are the commodities with the high return rate, and selecting the commodities with the similarity higher than the preset similarity value from the union set according to the probability of 1-epsilon to continuously recommend the commodities to the user.
Preferably, after the step S2, the method further includes the following steps: step S3: and recommending commodities in the commodity data set by the epsilon probability for each round of the master user, and recommending similar commodities with return rate higher than a preset return rate value by the 1-epsilon probability until commodity recommendation is cut off.
Preferably, the step S3, up to the recommendation cutoff, includes the following steps: step S301, performing dominant behavior scoring and invisible behavior scoring on behaviors of recommended commodities according to a master user to obtain a total score value; step S302, setting a threshold value, and comparing the total score value of the behavior of the master user with the threshold value; and step S303, judging whether the commodity recommendation is cut off according to the comparison result of the total score value and the cut-off threshold value.
In order to solve the above technical problems, the present invention provides another technical solution as follows: a commodity recommendation system based on a social e-commerce platform comprises the following modules: an acquisition module: obtaining social relationship information of a master user, and obtaining commodities historically purchased by the user in the social relationship associated with the master user to form a candidate commodity data set; obtaining historical purchased commodities of a user with a proxy behavior by a master user to form a commodity data set A; if the commodity data set A is smaller than a preset threshold value, obtaining a commodity data set B formed by historical shopping commodities of a user, of which the main user has information exchange, and supplementing the commodity data set B to the commodity data set A; if the union of the commodity data set A and the commodity data set B is still smaller than a preset threshold value, expanding the union of the commodity data set A and the commodity data set B to obtain a total candidate commodity data set; the device information in purchasing according to the main user and the social relationship can be expanded; or if the device data is not authorized to be read and the device information cannot be acquired, the device data can be expanded through attribute clustering according to the background attribute of the master user or the commodity; or recommending and expanding based on the association rule according to the background attribute of the master user or the commodity; a recommendation module: obtaining a union set of the candidate commodity data set and the E-commerce recommended commodity data set, and recommending commodities to a master user from the union set by using a Bandit algorithm; a judging module: and judging whether the recommended commodity is cut off.
Preferably, further comprising: the user terminal is used for authorizing the system to access the social network account of the master user; and the social network server is used for storing the information of the social network account of the master user.
In order to solve the above technical problems, the present invention provides another technical solution as follows: a computer-readable medium, in which a computer program is stored, wherein the computer program is arranged to execute the social provider platform based item recommendation method as described in the preceding item when executed.
In order to solve the above technical problems, the present invention provides another technical solution as follows: the commodity recommending device based on the social e-commerce platform can comprise a commodity recommending system based on the social e-commerce platform.
Compared with the prior art, the commodity recommendation method, the commodity recommendation system, the computer readable medium and the commodity recommendation device based on the social e-commerce platform are provided by the invention. The method has the following beneficial effects:
1. compared with the existing commodity recommendation system, the Multi-arm slot machine (Multi-arm bandit) problem is applied to individual recommendation by adopting hot recommendation and recommendation according to historical shopping to solve the cold start problem, and the social relationship information of the master user is utilized when a commodity data set is selected, and extends it into a candidate commodity data set, and at the same time, combines the commodities of the e-commerce hot platform for developing the freshness of the users, compared with the existing hot recommendations, in a mode of collecting commodity aggregation types in a large range, the commodity data set is constructed from the historical commodities purchased by the user in the social relationship of the master user, and because the master user and other users have social behaviors on the social e-commerce platform, therefore, the advantages of the social e-commerce platform can be fully utilized, and recommended commodities are closer to the range interested by the master user; compared with the recommended commodities, the recommended commodities are relatively more targeted and can more easily approach the real requirements of cold-start users.
Meanwhile, a Bandit algorithm is applied: when the commodities and similar commodities which are interesting to the master user are continuously recommended, the new interest of the user is explored at the same time, and the probability that the master user indicates that the commodity type which is not interesting is recommended is lower and lower for many times along with the increase of the number of recommendation rounds; after repeated execution for multiple times, the potential characteristics of the master user are accurately fitted, and the cold start problem is effectively solved;
2. considering that a reasonable commodity data set cannot be constructed after the agent behavior relation of the master user and the information exchange commodity data set of the master user are obtained, and the agent behaviors of some master users or the users only exchanging information are not wide due to the difference of different users, the historical shopping commodities in the social relation of the users are continuously added if the historical shopping commodities in the superior-subordinate relation of the users are not enough after the agent behaviors of some master users or the users only exchanging information are obtained at first, and the commodity data set is expanded if the commodity data set is still small; in such a manner, the commodity composition in the commodity data set mostly takes the historical shopping commodities of the user superior-subordinate relationship and the social relationship as the main, so the total commodity data set is closer to the requirements of the master user.
3. The expansion mode can be expansion through master user equipment information, expansion can be carried out through attribute clustering according to the background attributes of the master user and the commodities in the commodity data set, and the expansion can be carried out according to the background attributes of the master user or the commodities and based on association rule recommendation, so that the expanded commodities of the data commodity set are also the commodities which are similar to the range possibly interested by the master user.
4. By utilizing an Epsilon-Greedy algorithm in the Bandit algorithm, a master user can easily select fatigue only when recommending commodities which are known to be interested by the user, so that some fresh commodities need to be scientifically recommended, the interest range of the user is continuously expanded, and the maximum information recommending efficiency is achieved; therefore, the user commodities are recommended by the Bandit algorithm combined by the known (explicit) algorithm and the unknown (explicit) algorithm, so that the main user can continuously obtain interested commodities while taking freshness into account;
5. and judging which actions of the main user are taken as interesting actions, and scoring the dominant action and the invisible action of the main user in a numerical value conversion mode, so that the multi-round circulation of the recommended commodities is cut off more intelligently.
6. The system based on the method for solving the cold start of the social e-commerce can comprise a main user terminal and a social network server, so that the system can directly acquire the social network account number of the main user and store the information of the social network account number of the main user, and the system can be applied to a social e-commerce platform, so that the social platform can recommend commodities to the main user.
[ description of the drawings ]
Fig. 1 is a flowchart illustrating a commodity recommendation method based on a social provider platform according to a first embodiment of the present invention;
fig. 2 is a detailed flowchart of step S1 in fig. 2 according to the first embodiment of the present invention;
fig. 3 is a specific flowchart illustrating that "expansion" is performed on the basis of the union of the commodity data set a and the commodity data set B in step S103 in fig. 2 according to the first embodiment of the present invention;
fig. 4 is a detailed flowchart showing step S2 in fig. 1 according to the first embodiment of the present invention;
FIG. 5 is a flow chart illustrating a variation of the first embodiment of the present invention;
fig. 6 is a flowchart showing the implementation of the determination "until commodity recommendation ends" in step S3 in fig. 5 according to the first embodiment of the present invention;
fig. 7 is a block diagram illustrating a solution to a product recommendation system based on a social e-commerce platform according to a second embodiment of the present invention.
Description of the drawings: 1. a user terminal; 2. a social network server; 3. an acquisition module; 4. a judgment module; 5. and a recommendation module.
[ detailed description ] embodiments
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The commodity recommendation method based on the social e-commerce platform is used for intelligently recommending commodities to users.
Fig. 1 is a flowchart of a method for personalized recommendation for a user according to an embodiment of the present invention, where the method includes the following steps:
step S1: obtaining commodities historically purchased by users in social relations associated with the master user based on the social relation information of the master user to form a candidate commodity data set; and
step S2: obtaining a union set according to the candidate commodity data set and the E-commerce recommended commodity data set obtained in the step S1, and recommending commodities to a master user from the union set by using a Bandit algorithm;
the social relationship users comprise users who have agent behaviors with the master user or only carry out information communication.
The "social relationship user" refers to a relationship user on a platform that performs information exchange, sharing, evaluation, discussion, propagation, and the like with a master user, and examples thereof include: microblog, twitter, blog, human web, internet of fun, forum, community, WeChat, and the like.
The term "agency action" is to be interpreted as agency purchase, and in the present invention, not only the purchase service is mainly used, but also package service, distribution service, convenience service, and the like, which represent the contract and sale contract of the product or service signed by the buyer and the seller.
"merchandise" characterizes a social term that represents an entity that can conduct business, can trade, and thus can include actual products, as well as service labor.
Therefore, in step S1, the product data set may be a product with specific tangible properties, such as: clothing, drinks, books, etc., products that may also include service attributes, such as, for example: travel link, maintenance service, freight service, learning and training.
It is to be noted that: in the step S1, the commodity historically purchased by the user in the social relationship associated with the master user is obtained based on the social relationship information of the master user to form a candidate commodity data set, and the commodity may be obtained on the same platform or an authorized third-party platform; the method is as follows schematically: the method comprises the steps that a user A and a user B under the same shopping platform, the user B purchases a commodity of the user A, the user A purchases a certain commodity as an active user, then the commodity enters a commodity data set recommended by the user B at the moment, the commodity is in an upper and lower level purchasing relationship, or account numbers of a third party such as QQ and WeChat are allowed to be adopted by a plurality of platforms at present, the advantage is that some basic user information can be obtained from the third party, and therefore the commodity can be classified as purchasing with a social relationship no matter the third party has a social relationship under the same platform.
The overall flow and some features of the method of the present invention are explained above, and the detailed flow of the method of the present invention is explained below with reference to examples.
In a first embodiment provided by the present invention, a method for recommending a commodity based on a social e-commerce platform is all processing manners in the method shown in fig. 1, wherein the first embodiment is only one specific implementation manner for step S1, and in step S1, steps S101 to S103 are specifically included; please refer to fig. 2:
step S101: obtaining historical purchased commodities of a user with a proxy behavior by a master user to form a commodity data set A;
step S102: if the commodity data set A is smaller than a preset threshold value, obtaining a commodity data set B formed by historical shopping commodities of a user, of which the main user has information exchange, and supplementing the commodity data set B to the commodity data set A;
step S103: and if the union of the commodity data set A and the commodity data set B is still smaller than the preset threshold value, expanding the union of the commodity data set A and the commodity data set B to obtain a total candidate commodity data set.
"expansion" refers to the supplementation of the commodity data set to avoid the less scope of selection that it may give to the primary user, as explained in detail below:
it can be understood that, when the "commodity data set" cannot satisfy the preset range selected by the master user, that is, the number, the types and the interval range of the so-called commodity data set are not enough, the ground step that the master user can continuously screen is not reached, that is, the preset threshold value is not exceeded, then the expansion filling needs to be performed, and the mode of specifically expanding the commodity data set may be: and performing extension based on association rule recommendation, utility recommendation or combined recommendation.
Further, the "preset threshold" may be a numerical value or a model of one to many items in the number, category, interval range,
a second embodiment of the present invention is a method of "expanding on the basis of the commodity data set a and the commodity data set B" in S103 of the first embodiment, please refer to fig. 4:
the filling may be performed by:
step S1031, expanding according to the main user and equipment information in purchasing of the social relationship; or
Step S1032, if the equipment data is not authorized to be read and the equipment information cannot be acquired, expansion can be performed through attribute clustering according to the background attribute of the master user or the commodity; or
And step S1033, recommending and expanding based on the association rule according to the background attribute of the master user or the commodity.
For convenience of understanding, step S1031 is supplemented, and examples are as follows: when the union of the commodity data set A and the commodity data set B in the step S1 is not enough, if the mobile phone of the master user is provided with the software of 'the car' the mobile phone, the commodities such as car accessories and car services are supplemented to enter the total commodity data set;
for convenience of understanding, step S1032 is supplemented, and schematically: if the commodity data set already contains the commodity of 'refrigerator', the 'refrigerator' can be clustered into the class of 'household appliances', and the 'washing machine' of the same kind of commodity is supplemented into the commodity data set;
for convenience of understanding, step S1033 is supplemented, and schematically: if the commodity data set already contains the commodity of the chopsticks, the bowls associated with the chopsticks are supplemented into the commodity data set.
The specific sequence of steps S1031 to S1033 is not limited, and even the necessity is not limited, and only some optional schemes are shown, and whether the implementation is required depends on specific conditions.
In a third embodiment provided by the present invention, a method for recommending a commodity based on a social e-commerce platform is all processing modes in the method shown in fig. 1, the third embodiment is only a specific implementation mode for step S2, and in step S2, steps S201 to S203 are specifically included; please refer to fig. 2:
step S201: selecting any number between one (0,1) as epsilon;
step S202: randomly recommending the commodities to the user in the union set according to the epsilon probability and acquiring the return rate of the commodities;
step S203: and confirming that the commodities with the return rate higher than the preset return rate value are the commodities with the high return rate, and selecting the commodities with the similarity higher than the preset similarity value from the union set according to the probability of 1-epsilon to continuously recommend the commodities to the user.
The commodity with the "rate of return" can be further understood as the behavior of the primary user on the interested commodity, wherein the behavior for judging the primary user to be interested can include clicking, collecting and consulting by the primary user, so as to deduce whether the primary user is interested in a certain commodity through the behavior of the primary user.
The "preset return rate value" and the "preset similarity value" are different threshold values preset according to specific situations, the "preset return rate value" is a threshold value for judging the commodity return rate, that is, an upper limit value when a user clicks, collects, and consults behaviors, after the user performs a series of behaviors, an upper limit value is set for judging that the user is interested in, how to make statistics on the user behaviors specifically can be obtained according to the dirichlet algorithm, and the "preset similarity value" is another threshold value for judging the commodity similarity, which mainly represents the similarity between commodities, as an example: the similarity between the mobile phone and the mobile phone cover may be 80%, the similarity between the mobile phone and the computer may be 60%, the preset similarity value is set at 70%, the mobile phone cover is a commodity with the commodity similarity greater than the preset similarity value, the specific similarity calculation algorithm may be a collaborative filtering (collaborative filtering) algorithm, and the method for acquiring the set of similar commodities may refer to the steps described in embodiment two.
It can be understood that, if epsilon is taken to be 0.1, the union set obtained in S1 is used, and in each time step, commodities are selected from the union set of the commodity data set and the popular commodities and recommended to the master user at a probability that epsilon is 0.1 (utilization) each time, it is determined which commodities can obtain a higher return rate, that is, which commodities are interested by the master user, which commodities the master user clicks, and then commodities similar to the commodities are randomly selected from the commodity data set at a probability that q is 1-epsilon is 0.9; illustratively, assuming epsilon is 0.1, some items are selected with a probability of 10% for the first time in the union of the item data set and the popular items, for example, the master user clicks the link of the corresponding item "one-piece dress", and then at this time, the items similar to the "one-piece dress" are continuously found with a probability of 90% from the union of the item data set and the popular items, such as: and recommending the long skirt, the wrapped skirt and the like to a master user.
The reason that the candidate commodity data set needs to be merged with the hot commodity set is that in the recommendation system, if only commodities which are known to be interested by the master user are recommended, the master user can easily select fatigue, so that some fresh commodities need to be scientifically recommended, the interest range of the master user is continuously expanded, and the recommendation information efficiency is maximized; therefore, the Bandit algorithm combining "explicit known (using)" and "explicit unknown (exploring)" is used to recommend the primary user commodity;
the fourth embodiment of the present invention is a method for recommending a commodity based on a social e-commerce platform for a user, which is all processing modes in the method shown in fig. 1, wherein the method is only a supplement to step S2 based on the third embodiment, and further includes step S3; please refer to fig. 6:
step S3: and recommending commodities in the commodity data set by the epsilon probability for each round of the master user, and recommending similar commodities with return rate higher than a preset return rate value by the 1-epsilon probability until commodity recommendation is cut off.
As will be readily understood by those skilled in the art, the algorithm is the Epsilon-Greedy algorithm in the Bandit algorithm, where the value of Epsilon controls the degree of preference for explore (exploite: utilization) and explore (explore: exploration), with each decision to explore a new good with probability Epsilon, 1-Epsilon probability to utilize good with high rate of return; it can be understood that the repetition of steps S202 to S203, i.e. the product in which the user is interested, is recommended more often, and new products are developed with less probability, so as to realize multiple recommendation cycles.
The fifth embodiment of the present invention is directed to a method for recommending a commodity based on a social provider platform for a user, which is all processing manners in the method shown in fig. 1, wherein the fifth embodiment is only one method for determining until the recommendation is ended in step S3, and please refer to fig. 5:
step S301: according to the behavior of the recommended commodity, dominant behavior scoring and invisible behavior scoring are carried out by the master user, and a total score value is obtained;
step S302: setting a cut-off threshold value, and comparing the total score value of the behavior of the master user with the cut-off threshold value;
step S303: and judging whether the commodity recommendation is cut off according to the comparison result of the total score value and the cut-off threshold value.
It can be understood that: the dominant scoring and the invisible scoring are the forms of converting the behaviors of the master users into mathematics, namely, the commodities are scored according to the number of times of attempts and the feedback of the master users, when the total score value exceeds a threshold value, the master users are determined to have interesting behaviors, and vice versa, repeated recommendation is continued; alternatively, stealth scoring may utilize a dirichlet algorithm, such as, for example: calculating an explicit score: setting the click rate as a Boolean parameter, setting the click rate as 1, and setting the click rate as 0 if not, similarly, setting corresponding score values for collection, comment, consultation and purchase, wherein if the score value of a collection behavior is i, the score value of a comment is j, and the invisible score can be used for calculating other elements of the main user, such as example score; assuming that a master user stays for a long time on an interface of a recommended commodity and has continuous actions of browsing the interface, such as screen sliding and multiple clicks, the master user is judged to have an interested invisible emotion on the commodity, so that the commodity is scored, or for example, a hot commodity in a commodity data set has a commodity which is contrary to the preference of the master user, but the master user is satisfied with the commodity, and the actions of clicking, collecting and the like are performed on the commodity, so that the master user is judged to have surprise on the commodity, and the scoring can be performed. Finally, the total value of the dominant behavior score and the invisible behavior is compared with a threshold value;
it is understood that the "cutoff threshold" is a value used to define primary user behavior, and above.
Referring to the same 7, a sixth embodiment of the present invention provides a system for solving the problem of commodity recommendation based on a social e-commerce platform, which can be applied to multiple devices such as a server, a platform, a computer, a camera, a mobile phone, a tablet computer, and the like, and the technology can be implemented by using hardware and software, wherein when the software is implemented and used, any one or a combination of the first embodiment to the fifth embodiment is required; it may include the following modules:
an acquisition module 3: obtaining social relationship information of a master user, and obtaining commodities historically purchased by the user in the social relationship associated with the master user to form a candidate commodity data set;
and a judging module 4: judging whether the recommended commodity is cut off; and
the recommending module 5: obtaining a union set of the candidate commodity data set and the E-commerce recommended commodity data set, and recommending commodities to a master user from the union set by using a Bandit algorithm;
it can be understood that, when the determining module 4 is implemented, the determining module is the steps S301 to S303 of the above-mentioned commodity recommending method based on the social e-commerce platform;
further, in some variations of the sixth embodiment, the solution is further provided with the user terminal 1 and the social network server 2 in the product recommendation system based on the social e-commerce platform:
the user terminal 1 is used for authorizing access to a social network account of a master user;
the social network server 2 is used for storing information of a social network account of the master user, and the obtaining module can obtain social relationship information in the social network server.
It can be understood that, in the acquisition module 3, many information data are required to acquire authorization across a third-party platform in order to acquire a commodity historically purchased by a user having social relationship information and acquiring a social relationship associated with a master user, so that it is particularly important to access the social network account of the user and information of the account; through a user terminal, a direct authorization platform can access a social network account of a master user; exemplary are as follows: when a plurality of APPs (APP: application programs of mobile phones/computers) log in, whether an address list or qq and WeChat can be accessed or not is inquired; meanwhile, the social network server 2 may store information of the social network account of the master user and give the acquisition module 3 to utilize it.
A seventh embodiment of the present invention provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processing device, the computer program controlling the storage medium or a processor to execute needs to execute any one of the above methods for recommending commodities based on a social provider platform, that is, the methods as described above in the first to fifth embodiments as a whole, or perform any one or a combination of the above steps.
An eighth embodiment of the present invention provides a commodity recommendation device based on a social e-commerce platform, which may include the computer-readable storage medium of the seventh embodiment or the system of the sixth embodiment.
It will be appreciated that the apparatus may be an unmanned vending apparatus, such as an unmanned container.
It will also be understood by those skilled in the art that if all or part of the method, system, readable storage medium, embodiments one to seven of the present invention are combined or modified to replace the corresponding embodiments in the system of the present invention, the method, system, readable storage medium and embodiments one to seven also fall within the protection scope of the present invention.
Compared with the prior art, the commodity recommendation method based on the social e-commerce platform has the following beneficial effects:
1. compared with the existing social e-commerce platform, the cold start problem is solved by adopting hot recommendation and recommendation according to historical shopping, the Multi-arm slot machine (Multi-arm basis) problem is applied to individual recommendation, when a commodity data set is selected, the social relationship information of a master user is utilized and expanded into a candidate commodity data set, meanwhile, the commodity of the e-commerce hot platform is combined for developing the freshness of the user, and compared with the existing hot recommendation, a commodity aggregation type collection mode is carried out in a large range, the commodity data set is constructed from the historical commodity purchased by the user of the social relationship of the master user, and the advantages of the social e-commerce platform can be fully utilized due to the fact that the master user and other users have social behaviors on the social e-commerce platform, so that the recommended commodity is closer to the range in which the master user is interested; compared with the recommended commodities, the recommended commodities are relatively more targeted and can more easily approach the real requirements of cold-start users.
Meanwhile, a Bandit algorithm is applied: when the commodities and similar commodities which are interesting to the master user are continuously recommended, the new interest of the user is explored at the same time, and the probability that the master user indicates that the commodity type which is not interesting is recommended is lower and lower for many times along with the increase of the number of recommendation rounds; after repeated execution for multiple times, the potential characteristics of the master user are accurately fitted, and the cold start problem is effectively solved;
2. considering that a reasonable commodity data set cannot be constructed after the agent behavior relation of the master user and the information exchange commodity data set of the master user are obtained, and the agent behaviors of some master users or the users only exchanging information are not wide due to the difference of different users, the historical shopping commodities in the social relation of the users are continuously added if the historical shopping commodities in the superior-subordinate relation of the users are not enough after the agent behaviors of some master users or the users only exchanging information are obtained at first, and the commodity data set is expanded if the commodity data set is still small; in such a manner, the commodity composition in the commodity data set mostly takes the historical shopping commodities of the user superior-subordinate relationship and the social relationship as the main, so the total commodity data set is closer to the requirements of the master user.
3. The expansion mode can be expansion through master user equipment information, expansion can be carried out through attribute clustering according to the background attributes of the master user and the commodities in the commodity data set, and the expansion can be carried out according to the background attributes of the master user or the commodities and based on association rule recommendation, so that the expanded commodities of the data commodity set are also the commodities which are similar to the range possibly interested by the master user.
4. By utilizing an Epsilon-Greedy algorithm in the Bandit algorithm, a master user can easily select fatigue only when recommending commodities which are known to be interested by the user, so that some fresh commodities need to be scientifically recommended, the interest range of the user is continuously expanded, and the maximum information recommending efficiency is achieved; therefore, the user commodities are recommended by the Bandit algorithm combined by the known (explicit) algorithm and the unknown (explicit) algorithm, so that the main user can continuously obtain interested commodities while taking freshness into account;
5. and judging which actions of the main user are taken as interesting actions, and scoring the dominant action and the invisible action of the main user in a numerical value conversion mode, so that the multi-round circulation of the recommended commodities is cut off more intelligently.
6. The system based on the method for solving the cold start of the social e-commerce can comprise a main user terminal and a social network server, so that the system can directly acquire the social network account number of the main user and store the information of the social network account number of the main user, and the system can be applied to a social e-commerce platform, so that the social platform can recommend commodities to the main user.
The present invention is not limited to the above preferred embodiments, and any modifications, equivalent alterations and improvements made within the spirit of the present invention should be included in the scope of the present invention.
Claims (9)
1. A commodity recommendation method based on a social e-commerce platform is characterized by comprising the following steps:
step S1: obtaining historical purchased commodities of a user with a proxy behavior by a master user to form a commodity data set A; if the commodity data set A is smaller than a preset threshold value, obtaining a commodity data set B formed by historical shopping commodities of a user, of which the main user has information exchange, and supplementing the commodity data set B to the commodity data set A; if the union of the commodity data set A and the commodity data set B is still smaller than a preset threshold value, expanding the union of the commodity data set A and the commodity data set B to obtain a total candidate commodity data set; the method comprises the following steps that expansion is carried out according to the main user and equipment information in purchasing of the social relationship; or if the equipment data is not authorized to be read and the equipment information cannot be acquired, expanding through attribute clustering according to the background attribute of the master user or the commodity; or recommending and expanding based on the association rule according to the background attribute of the master user or the commodity; and
step S2: obtaining a union set according to the candidate commodity data set and the E-commerce recommended commodity data set obtained in the step S1, and recommending commodities to a master user from the union set by using a Bandit algorithm;
the social relationship comprises a relationship with a master user through proxy purchasing behavior or information exchange.
2. The social merchant platform-based commodity recommendation method according to claim 1, wherein the expansion is performed on the basis of a union of the commodity data set a and the commodity data set B, and the expansion comprises: and expanding through background attribute clustering or background attribute association according to the background attributes of the main user and the commodities in the candidate commodity data set.
3. The social merchant platform-based commodity recommendation method according to claim 1, wherein in the step S2, a union is obtained according to the candidate commodity data set and the e-commerce recommended commodity data set obtained in the step S1, and a Bandit algorithm is used to recommend commodities from the union to the primary user, comprising the steps of:
step S201: selecting any number between one (0,1) as epsilon;
step S202: randomly recommending the commodities to the user in the union set according to the epsilon probability and acquiring the return rate of the commodities; and
step S203: and confirming that the commodities with the return rate higher than the preset return rate value are the commodities with the high return rate, and selecting the commodities with the similarity higher than the preset similarity value from the union set according to the probability of 1-epsilon to continuously recommend the commodities to the user.
4. The social merchant platform-based commodity recommendation method according to claim 1, further comprising, after the step S2, the steps of:
step S3: recommending commodities in a centralized manner according to the union obtained by the candidate commodity data set and the E-commerce recommended commodity data set obtained in the step S1 by the epsilon probability for each round of the master user, recommending similar commodities with the return rate being larger than the preset return rate value according to the 1-epsilon probability until commodity recommendation is cut off.
5. The social merchant platform-based commodity recommendation method according to claim 4, wherein the step S3, up to the recommendation cutoff, comprises the steps of:
step S301, performing dominant behavior scoring and invisible behavior scoring on behaviors of recommended commodities according to a master user to obtain a total score value;
step S302, setting a cut-off threshold value, and comparing the total fraction value of the behavior of the main user with the cut-off threshold value;
and step S303, judging whether the commodity recommendation is cut off according to the comparison result of the total score value and the cut-off threshold value.
6. A commodity recommendation system based on a social e-commerce platform is characterized by comprising the following modules:
an acquisition module: obtaining commodities historically purchased by users in social relations associated with the master user based on the social relation information of the master user to form a candidate commodity data set; specifically, historical commodity purchasing of a user with a proxy behavior by a master user is obtained to form a commodity data set A; if the commodity data set A is smaller than a preset threshold value, obtaining a commodity data set B formed by historical shopping commodities of a user, of which the main user has information exchange, and supplementing the commodity data set B to the commodity data set A; if the union of the commodity data set A and the commodity data set B is still smaller than a preset threshold value, expanding the union of the commodity data set A and the commodity data set B to obtain a total candidate commodity data set; the method comprises the following steps that expansion is carried out according to the main user and equipment information in purchasing of the social relationship; or if the equipment data is not authorized to be read and the equipment information cannot be acquired, expanding through attribute clustering according to the background attribute of the master user or the commodity; or recommending and expanding based on the association rule according to the background attribute of the master user or the commodity;
a recommendation module: obtaining a union set of the candidate commodity data set and the E-commerce recommended commodity data set, and recommending commodities to a master user from the union set by using a Bandit algorithm; and
a judging module: and judging whether the recommended commodity is cut off.
7. The social merchant platform-based merchandise recommendation system of claim 6, further comprising:
the user terminal is used for authorizing the system to access the social network account of the master user;
the social network server is used for storing the information of the social network account of the master user, and the obtaining module can obtain the social relationship information in the social network server.
8. A computer-readable medium, in which a computer program is stored, wherein the computer program is configured to execute the social e-commerce platform based item recommendation method of any one of claims 1 to 5 when executed.
9. A commodity recommendation device based on the social e-commerce platform, wherein the commodity recommendation device based on the e-commerce platform comprises the commodity recommendation system based on the social e-commerce platform as claimed in claim 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910626799.8A CN110335123B (en) | 2019-07-11 | 2019-07-11 | Commodity recommendation method, system, computer readable medium and device based on social e-commerce platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910626799.8A CN110335123B (en) | 2019-07-11 | 2019-07-11 | Commodity recommendation method, system, computer readable medium and device based on social e-commerce platform |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110335123A CN110335123A (en) | 2019-10-15 |
| CN110335123B true CN110335123B (en) | 2021-12-07 |
Family
ID=68146475
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910626799.8A Active CN110335123B (en) | 2019-07-11 | 2019-07-11 | Commodity recommendation method, system, computer readable medium and device based on social e-commerce platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110335123B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111475737B (en) * | 2020-03-23 | 2023-08-18 | 平安直通咨询有限公司 | User recommendation method, device, electronic equipment and storage medium |
| CN111476643A (en) * | 2020-04-15 | 2020-07-31 | 创新奇智(重庆)科技有限公司 | Interested commodity prediction method and device, electronic equipment and computer storage medium |
| CN111652648A (en) * | 2020-06-03 | 2020-09-11 | 陈包容 | Method for intelligently generating personalized combined promotion scheme and system with same |
| CN111652631B (en) * | 2020-06-09 | 2023-06-06 | 创新奇智(成都)科技有限公司 | Commodity recall method, commodity recall device, electronic equipment and computer readable storage medium |
| CN111681085A (en) * | 2020-06-10 | 2020-09-18 | 创新奇智(成都)科技有限公司 | Commodity pushing method and device, server and readable storage medium |
| CN111754300A (en) * | 2020-06-22 | 2020-10-09 | 深圳市分期乐网络科技有限公司 | Commodity recommendation method, commodity recommendation device, commodity recommendation equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103995866A (en) * | 2014-05-19 | 2014-08-20 | 北京邮电大学 | Commodity information pushing method and device based on link forecasting |
| CN107103499A (en) * | 2017-04-26 | 2017-08-29 | 中国人民大学 | Method is recommended in a kind of cross-site cold start-up based on deep learning |
| CN107563867A (en) * | 2017-09-06 | 2018-01-09 | 天津大学 | A kind of commending system cold start-up method based on multi-arm fruit machine confidence upper limit |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2816511A1 (en) * | 2013-06-21 | 2014-12-24 | Thomson Licensing | Method for cold start of a multi-armed bandit in a recommender system |
| KR102012676B1 (en) * | 2016-10-19 | 2019-08-21 | 삼성에스디에스 주식회사 | Method, Apparatus and System for Recommending Contents |
| CN106897912B (en) * | 2017-01-18 | 2020-08-28 | 太仓中科信息技术研究院 | Cold start method of recommendation system based on epsilon feedback algorithm |
| CN109063003A (en) * | 2018-07-09 | 2018-12-21 | 维沃移动通信有限公司 | A kind of content recommendation method and mobile terminal |
| CN109684507A (en) * | 2018-11-23 | 2019-04-26 | 北京达佳互联信息技术有限公司 | Video recommendation method, video recommendations device and computer readable storage medium |
-
2019
- 2019-07-11 CN CN201910626799.8A patent/CN110335123B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103995866A (en) * | 2014-05-19 | 2014-08-20 | 北京邮电大学 | Commodity information pushing method and device based on link forecasting |
| CN107103499A (en) * | 2017-04-26 | 2017-08-29 | 中国人民大学 | Method is recommended in a kind of cross-site cold start-up based on deep learning |
| CN107563867A (en) * | 2017-09-06 | 2018-01-09 | 天津大学 | A kind of commending system cold start-up method based on multi-arm fruit machine confidence upper limit |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110335123A (en) | 2019-10-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110335123B (en) | Commodity recommendation method, system, computer readable medium and device based on social e-commerce platform | |
| KR101939554B1 (en) | Determining a temporary transaction limit | |
| US20140081800A1 (en) | Recommending Product Information | |
| CN106251168A (en) | Information-pushing method and system | |
| Cho et al. | Implementation of Personalized recommendation System using k-means Clustering of Item Category based on RFM | |
| CN108805598A (en) | Similarity information determines method, server and computer readable storage medium | |
| JP2019508756A (en) | Method and apparatus for selecting and recommending objects on an electronic delivery platform | |
| CN107808346B (en) | Evaluation method and evaluation device for potential target object | |
| CN110134783A (en) | Method, apparatus, equipment and the medium of personalized recommendation | |
| CN111626767B (en) | Resource data issuing method, device and equipment | |
| Liu et al. | Context similarity metric for multidimensional service recommendation | |
| CN108182268B (en) | Collaborative filtering recommendation method and system based on social network | |
| CN111475744B (en) | Personalized position recommendation method based on ensemble learning | |
| Wei et al. | A study of customer behaviour through web mining | |
| CN113327154B (en) | E-commerce user message pushing method and system based on big data | |
| ur Rehman et al. | Frequency-based similarity measure for multimedia recommender systems | |
| Pinto et al. | Hybrid recommendation system based on collaborative filtering and fuzzy numbers | |
| CN115456656A (en) | Method and device for predicting purchase intention of consumer, electronic equipment and storage medium | |
| CN113822691A (en) | User account identification method, device, system and medium | |
| Zhu | Research on multi‐source mobile commerce service recommendation model of data fusion based on tree network | |
| Osial et al. | Smartphone recommendation system using web data integration techniques | |
| Liu et al. | Using contextual information for service recommendation | |
| Ronca et al. | A flexible framework for context-aware recommendations in the Social Commerce domain | |
| Gao | Design of online retail commodity recommendation system based on Item-CF algorithm | |
| Xu et al. | Research on User Recommendation Algorithm Based on Cross-border E-commerce Platform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information |
Inventor after: Zhou Wei Inventor after: Zhang Faen Inventor after: Chen Binbin Inventor after: Zhou Pengcheng Inventor before: Zhang Faen Inventor before: Chen Binbin Inventor before: Zhou Pengcheng |
|
| CB03 | Change of inventor or designer information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |