CN110827112B - Deep learning commodity recommendation method and device, computer equipment and storage medium - Google Patents
Deep learning commodity recommendation method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN110827112B CN110827112B CN201910871299.0A CN201910871299A CN110827112B CN 110827112 B CN110827112 B CN 110827112B CN 201910871299 A CN201910871299 A CN 201910871299A CN 110827112 B CN110827112 B CN 110827112B
- Authority
- CN
- China
- Prior art keywords
- commodity
- vector
- information
- product
- exposure
- 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
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Business, Economics & Management (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Accounting & Taxation (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Finance (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Evolutionary Biology (AREA)
- Strategic Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a commodity recommendation method, a commodity recommendation device, computer equipment and a storage medium for deep learning, and relates to the field of artificial intelligence.
Description
Technical Field
The present application relates to the field of artificial intelligence, and in particular, to a method and an apparatus for recommending commodities for deep learning, a computer device, and a storage medium.
Background
With the rapid development of e-commerce websites, a large amount of user data including user basic information, user purchasing behavior data, user browsing and collecting data, article information and the like are accumulated in e-commerce websites. How to analyze and mine the accumulated mass user data and build a user purchase model to recommend articles to the user is a hot problem of an e-commerce website optimization model. However, the recommended goods output by the traditional model are bias goods, and the recommendation mode has the repeated information recommendation problem, namely the recommendation content of one face of thousands of people is low in prediction accuracy of information recommendation.
Disclosure of Invention
The embodiment of the application aims to provide a deep learning commodity recommendation method to solve the problem that the accuracy of recommended commodity information is low in a traditional commodity recommendation mode.
In order to solve the above technical problem, an embodiment of the present application provides a deep learning commodity recommendation method, including the following steps:
acquiring a commodity list of a user, wherein the commodity list comprises commodity information of exposure products and sequence products;
vectorizing the commodity information to obtain a mean vector corresponding to the commodity information;
inputting the mean vector into a preset recommendation model for commodity classification, and outputting commodity recommendation information corresponding to the mean vector;
recommending the commodity recommendation information to the user.
Further, the deep learning commodity recommendation method further comprises the following steps:
if the time that the user browses the target commodity in the commodity interface is detected to exceed a preset time value, marking the target commodity as an exposure product, and otherwise, marking the target commodity as a sequence product;
respectively recording commodity information of the exposure product and the sequence product in a commodity list;
and acquiring the commodity information of the preset quantity in the same preset time period from the commodity list.
Further, the deep learning commodity recommendation method further comprises the following steps:
respectively extracting commodity attribute information of the exposure product and the sequence product;
inputting the commodity attribute information into a preset attribute identification model to respectively obtain an exposure vector corresponding to the commodity attribute information of the exposure product and a sequence vector corresponding to the commodity attribute information of the sequence product;
and carrying out mean value processing on the sequence vector and the exposure vector to obtain a mean value vector corresponding to the same attribute type, wherein the mean value vector comprises a first vector corresponding to the sequence vector and a second vector corresponding to the exposure vector.
Further, the deep learning commodity recommendation method further comprises the following steps:
and respectively identifying first commodity attribute information corresponding to the sequence product and second commodity attribute information corresponding to the exposure product from the commodity information through a preset convolutional neural network model.
Further, the deep learning commodity recommendation method further comprises the following steps:
calculating similarity values of a first vector and a second vector in the mean vector;
determining a style model corresponding to the commodity information according to the similarity value;
and outputting the commodity recommendation information matched with the style model.
Further, the deep learning commodity recommendation method further comprises the following steps:
and converting the text format of the commodity attribute information into a numerical vector through the attribute identification model.
Further, the deep learning commodity recommendation method further comprises the following steps:
determining the final attribute type of the commodity information according to the similarity value;
and inputting the final attribute type into a recommendation model, and acquiring a style model corresponding to the commodity information.
In order to solve the above technical problem, an embodiment of the present application further provides a deep learning commodity recommendation device, where the deep learning commodity recommendation device includes:
the system comprises an acquisition module, a display module and a display module, wherein the acquisition module is used for acquiring a commodity list of a user, and the commodity list comprises commodity information of an exposure product and a sequence product;
the processing module is used for vectorizing the commodity information to obtain a mean vector corresponding to the commodity information;
the classification module is used for inputting the mean vector into a preset recommendation model for commodity classification and outputting commodity recommendation information corresponding to the mean vector;
and the execution module is used for recommending the commodity recommendation information to the user.
Further, the obtaining module further includes:
the detection submodule is used for marking the target commodity as an exposure product if the time for the user to browse the target commodity in the commodity interface is detected to exceed a preset time value, and otherwise, marking the target commodity as a sequence product;
the recording sub-module is used for respectively recording the commodity information of the exposure product and the sequence product in a commodity list;
and the acquisition sub-module is used for acquiring the commodity information of the preset quantity in the same preset time period from the commodity list.
Further, the processing module further comprises:
the extraction submodule is used for respectively extracting the commodity attribute information of the exposure product and the sequence product;
the vector submodule is used for inputting the commodity attribute information into a preset attribute identification model to respectively obtain an exposure vector corresponding to the commodity attribute information of the exposure product and a sequence vector corresponding to the commodity attribute information of the sequence product;
and the mean value submodule is used for carrying out mean value processing on the sequence vector and the exposure vector to obtain a mean value vector corresponding to the same attribute type, wherein the mean value vector comprises a first vector corresponding to the sequence vector and a second vector corresponding to the exposure vector.
Further, the extracting sub-module further includes:
and the identification unit is used for respectively identifying first commodity attribute information corresponding to the sequence product and second commodity attribute information corresponding to the exposure product from the commodity information through a preset convolutional neural network model.
Further, the execution module further comprises:
the calculation submodule is used for calculating the similarity value of a first vector and a second vector in the mean value vector;
the style sub-module is used for determining a style model corresponding to the commodity information according to the similarity value;
and the recommending submodule is used for outputting the commodity recommending information matched with the style model.
Further, the vector submodule further includes:
and the conversion unit is used for converting the text format of the commodity attribute information into a numerical vector through the attribute identification model.
Further, the style sub-module further comprises:
the attribute subunit is used for determining the final attribute type of the commodity information according to the similarity value;
and the style recommending subunit is used for inputting the final attribute type into a recommending model and acquiring a style model corresponding to the commodity information.
In order to solve the above technical problem, an embodiment of the present application further provides a computer device, which includes a memory and a processor, where the memory stores a computer program, and the processor implements the steps of the deep learning commodity recommendation method when executing the computer program.
In order to solve the above technical problem, an embodiment of the present application further 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 processor, the steps of the deep learning commodity recommendation method are implemented.
Compared with the prior art, the embodiment of the application mainly has the following beneficial effects:
according to the method, commodity information is subjected to mean vector processing by introducing commodity information such as exposure products and sequence products, and after the commodity information is subjected to mean vector processing, similarity factors between the exposure products and the sequence products are further obtained, so that the influence of the similarity on the result of the recommendation model is enhanced, the convergence rate of the recommendation model is improved, the prediction accuracy of information recommendation is improved, and meanwhile, the recommended commodity information of the user is predicted according to the commodity list of the user, so that the recommendation has personalized characteristics, and the problem that one-to-one-thousand recommendation modes exist in the recommendation mode in the prior art is solved.
Drawings
In order to more clearly illustrate the solution of the present application, the drawings needed for describing the embodiments of the present application will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present application, and that other drawings can be obtained by those skilled in the art without inventive effort.
FIG. 1 is an exemplary system architecture diagram to which the present application may be applied;
FIG. 2 is a flow diagram of one embodiment of a deep-learning merchandise recommendation method of the present application;
FIG. 3 is a schematic diagram of a network architecture of a recommendation model of the present application;
FIG. 4 is a schematic block diagram of one embodiment of a deep learning merchandise recommendation device according to the present application;
FIG. 5 is a schematic block diagram of one embodiment of a computer device according to the present application.
Detailed Description
Unless defined otherwise, all 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; the terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "including" and "having," and any variations thereof in the description and claims of this application and the description of the figures above, are intended to cover non-exclusive inclusions. The terms "first," "second," and the like in the description and claims of this application or in the above-described drawings are used for distinguishing between different objects and not for describing a particular order.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.
As shown in fig. 1, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 serves as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.
The user may use the terminal devices 101, 102, 103 to interact with the server 105 via the network 104 to receive or send messages or the like. Various communication client applications, such as a web browser application, a shopping application, a search application, an instant messaging tool, a mailbox client, social platform software, etc., may be installed on the terminal devices 101, 102, 103.
The terminal devices 101, 102, 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III, motion Picture Experts compression standard Audio Layer 3), MP4 players (Moving Picture Experts Group Audio Layer IV, motion Picture Experts compression standard Audio Layer 4), laptop portable computers, desktop computers, and the like.
The server 105 may be a server providing various services, such as a background server providing support for pages displayed on the terminal devices 101, 102, 103.
It should be noted that the deep learning commodity recommendation method provided in the embodiments of the present application is generally executed by a server/terminal device, and accordingly, the deep learning commodity recommendation apparatus is generally disposed in the server/terminal device.
It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
With continued reference to FIG. 2, a flow diagram of one embodiment of a method of deep-learned merchandise recommendation in accordance with the present application is shown. The deep learning commodity recommendation method comprises the following steps:
s201: the method comprises the steps of obtaining a commodity list of a user, wherein the commodity list comprises commodity information of an exposure product and a sequence product.
Specifically, a user obtains commodity information on a commodity interface when the user terminal browses the commodity interface, wherein the user terminal may include a smart phone, a tablet, a computer, and the like, and the commodity list refers to the commodity information browsed by the user in the commodity interface, that is, the commodity attribute information of the exposure product and the commodity attribute information of the sequence product.
S202: and vectorizing the commodity information to obtain a mean vector corresponding to the commodity information.
In this embodiment, the commodity information vectorization processing may be performed by respectively obtaining an exposure vector corresponding to the commodity attribute information of the exposure product and a sequence vector corresponding to the commodity attribute information of the sequence product in a preset attribute identification model, and performing an average processing on the sequence vector and the exposure vector to obtain an average vector, where the average vector includes a first vector corresponding to the sequence vector and a second vector corresponding to the exposure vector.
S203: and inputting the mean vector into a preset recommendation model for commodity classification, and outputting commodity recommendation information corresponding to the mean vector.
Specifically, a specific process of obtaining the commodity recommendation information is shown in fig. 3, fig. 3 is a schematic diagram of a network structure of a recommendation model, the recommendation model is a neural network model trained to be convergent and used for commodity classification, a mean vector is input into the recommendation model, a similarity value, namely a similarity factor, of a first vector (item embedding, sequence vector) and a second vector (exposure vector) in the mean vector is calculated, the first vector, the second vector and the similarity value after being processed by a triple activation function (Relu) are matched with a commodity type corresponding to the mean vector through an activation function (sigmoid), corresponding commodity recommendation information is obtained according to the commodity type, and the commodity recommendation information is sequentially output according to a preset arrangement mode, and the preset arrangement mode can be sorted according to the heat degree of the obtained commodity recommendation information.
S204: recommending the commodity recommendation information to the user.
The recommendation mode can be a webpage popup mode, a short message link mode, an email mode or the like.
Specifically, according to basic information for identifying the user, for example, a mobile phone number or a mailbox address of the user, the obtained commodity recommendation information is converted into a short link and sent to a short message of the user or a mailbox of the user.
In this embodiment, the recommendation manner may be that when it is detected that the user browses the commodity information on the shopping website, the commodity recommendation information ranked in the top may be presented on the browsing interface of the user in a webpage popup mode.
The method comprises the steps of obtaining a commodity list of a user, vectorizing commodity information, obtaining a mean vector corresponding to the commodity information, inputting the mean vector into a preset recommendation model for commodity classification, outputting commodity recommendation information corresponding to the mean vector, and recommending the commodity recommendation information to the user.
In some optional implementation manners of this embodiment, in step S201, that is, acquiring a commodity list of a user, the electronic device may further perform the following steps:
if the time that the user browses the target commodity in the commodity interface is detected to exceed a preset time value, marking the target commodity as an exposure product, and otherwise, marking the target commodity as a sequence product;
respectively recording commodity information of the exposure product and the sequence product in a commodity list;
and acquiring the commodity information of the preset quantity in the same preset time period from the commodity list.
Specifically, the process of browsing the target product by the user includes the target product which is not viewed in detail by the user in a product interface and the target product which is viewed in detail by the user through a detail viewing operation by the user, wherein the detail viewing may be performed by clicking or touching a screen or the like to enter a detail page of the target product, when the browsing time of the user on the detail page of the target product exceeds a preset time value, the target product is marked as an exposure product, the preset time value may be set to 10 seconds, 15 seconds or other time periods, and if the target product which is not viewed in detail by the user or the target product whose browsing time on the detail page of the target product is smaller than the preset time value is detected, the target product is marked as a sequence product, and the marking may be in a text label form to distinguish the sequence product from the exposure product.
Further, the commodity information may be, but not limited to, commodity attribute information of the serial product and the exposure product, and a time for browsing the serial product and the exposure product, the commodity attribute information may be a commodity model, a commodity usage, commodity review information, a commodity price, and the like, and the commodity list is used for recording the commodity information.
The same preset time period may be 10 minutes, the preset number may be an integer of 10, 20, 30, and the like, and the preset number of commodity information is commodity information of the preset number of sequence products and commodity information of the exposure products. For example, if a user continuously browses 10 items during a certain period of time while logging in a shopping website, the 10 items may be regarded as a short sequence, and the short sequence includes a user-ordered product and an exposure product, i.e., the item information of the short sequence may be obtained from an item list recording the 10 items.
Through detecting the process that a user browses a target commodity in a commodity interface, the target commodity is divided into marked exposure products and sequence products according to different conditions of browsing the target commodity, commodity information of the exposure products and the sequence products is recorded in a commodity list, and a preset amount of commodity information in the same preset time period is obtained from the commodity list, so that targeted useful data information can be obtained according to browsing habits of the user, and a data basis is provided for subsequently and accurately predicting commodity recommendation of the user.
In some optional implementation manners of this embodiment, in step S201, the vectorizing the commodity information, and obtaining the mean vector corresponding to the commodity information specifically include:
respectively extracting commodity attribute information of the exposure product and the sequence product;
respectively acquiring an exposure vector corresponding to commodity attribute information of an exposure product and a sequence vector corresponding to commodity attribute information of a sequence product in a preset attribute identification model;
and carrying out mean value processing on the sequence vector and the exposure vector to obtain a mean value vector corresponding to the same attribute type, wherein the mean value vector comprises a first vector corresponding to the sequence vector and a second vector corresponding to the exposure vector.
Specifically, since the extracted commodity information includes a large amount of commodity attribute information, it is necessary to extract some useful commodity attribute information from a large amount of commodity attribute information, and the extraction condition may be a commodity type, a commodity price, a commodity style, or the like.
The preset attribute identification model is used for acquiring a sequence vector corresponding to the commodity attribute information of the sequence product and an exposure vector corresponding to the commodity attribute information of the exposure product. The attribute identification model adopts a deep neural network to respectively convert the commodity attribute information of the sequence product and the commodity attribute information of the exposure product into embedded (embedding) feature vectors, namely the sequence vector and the exposure vector.
Specifically, embedded feature vectors corresponding to the commodity attribute information are searched in the deep neural network, namely weights corresponding to all dimensions corresponding to the commodity attribute information are searched in the deep neural network, the searched weights of all dimensions form the embedded feature vectors corresponding to all commodity attribute information, and the weights are obtained by training in the deep neural network in advance.
Specifically, since there are a plurality of commodity attribute information belonging to the same attribute type in different sequence products or exposure products, that is, there are a plurality of sequence vectors or exposure vectors belonging to the same attribute type, it is necessary to perform an averaging process on the sequence vectors and the exposure vectors to obtain an average value vector, that is, after each sequence vector or each element of the exposure vector of the same attribute type with the same dimension is accumulated, the average value vector is obtained by dividing the accumulated value by the number of sequence vectors or exposure vectors of the same attribute type. For example, two sequence vectors in the product attribute information of the sequence product are both of the financing type, and are assumed to be (1, 3,5, 7) and (2, 4,6, 8), respectively, and the first vector obtained after the averaging processing is (1.5, 3.5,5.5, 7.5).
The method comprises the steps of respectively extracting commodity attribute information of exposure products and sequence products, respectively obtaining exposure vectors corresponding to the commodity attribute information of the exposure products and sequence vectors corresponding to the commodity attribute information of the sequence products in a preset attribute identification model, and carrying out mean processing on the sequence vectors and the exposure vectors to obtain mean vectors, so that the interference of repeated exposure vectors or sequence vectors is reduced, excessive irrelevant data is prevented from being input into a subsequent recommendation model, and the prediction accuracy of the recommendation model is improved.
In some optional implementations of this embodiment, the step of extracting the commodity attribute information of the exposure product and the sequence product respectively specifically includes:
and respectively identifying first commodity attribute information corresponding to the sequence product and second commodity attribute information corresponding to the exposure product from the commodity information through a preset convolutional neural network model.
In this embodiment, the preset convolutional neural network model may extract a keyword that better represents the commodity attribute information based on the compound keyword processing and named entity recognition of the morpheme analysis result. Here, if there is a keyword pair in which inter-Point Mutual Information (PMI) has a preset PMI value among a plurality of keywords, the keywords in the pair are combined into a composite keyword so as to be treated as a single keyword. For example, it may be assumed that the keyword B, which is the brand name of the serial product a, is extracted from the analysis result of morphemes in the item information on the serial product a. Then, when a keyword C having a statistical significance with respect to the keyword B is extracted using the word, a composite keyword in which the keyword B and the keyword C are combined may be regarded as a single keyword with respect to the serial product a. Generally, a plurality of same attribute information exists in the commodity information of the same sequence product or exposure commodity, so that the method can be used for extracting the commodity attribute information. Therefore, the first commodity attribute information corresponding to the sequence product and the second commodity attribute information corresponding to the exposure product can be respectively identified from the commodity information according to the convolutional neural network model.
The first commodity attribute information corresponding to the sequence product and the second commodity attribute information corresponding to the exposure product are respectively identified from the commodity information through the preset convolutional neural network model, and the commodity attribute information of the exposure product and the sequence product is accurately extracted.
In some embodiments, in the step S203, the average vector is input into a preset recommendation model for product classification, and product recommendation information corresponding to the average vector is output, and the electronic device may perform the following steps:
calculating similarity values of a first vector and a second vector in the mean vector;
determining a style model corresponding to the commodity information according to the similarity value;
and outputting the commodity recommendation information matched with the style model.
The style model refers to a style, a price interval, a brand of a commodity, and the like of which a user prefers a certain commodity type, for example, the style of the short skirt favored by the user may be a pleated skirt, a hip-wrapped skirt, or an irregular skirt, or may be various types of short skirts corresponding to the price interval predicted to be favored by the user.
Because the weight factors respectively set for different exposure commodities and sequence commodities are different, namely when the mean value vector is calculated, the first vector corresponding to the sequence commodity is different from the second vector corresponding to the exposure commodity. The similarity value is an index for measuring the similarity of the commodities, and the similarity value of the first vector and the second vector can be calculated by adopting a cosine similarity calculation mode. The purpose of adopting this mode is to improve the prediction accuracy of the attribute type of the user's favorite commodity.
Further, in this embodiment, the final attribute type of the commodity information is determined according to the similarity value; and inputting the final attribute type into a recommendation model to determine a style model corresponding to the commodity information.
Specifically, the first commodity attribute information and the second commodity attribute information which are preliminarily screened out are provided with a plurality of similarity values, the obtained similarity values are at least 2, each final attribute type corresponds to a similarity interval, for example, the similarity interval of the life insurance type is (0.4, 0.6), a final model corresponding to each similarity value is matched in a final attribute type table, each final attribute type is graded in a recommendation model to obtain the total grade of the final type of a user, a style model is determined according to the total grade, the recommendation model is preset with the weight of each style model and the weight of each final attribute type, each style model carries a unique label, and commodity recommendation information is matched in a style database according to the label, wherein the style database stores the recommendation commodity information in advance, and the recommendation commodity information can be stored in a text link mode.
By calculating the similarity values of the first vector and the second vector in the mean vector, determining the style model corresponding to the commodity information according to the similarity values, and outputting the commodity recommendation information matched with the style model, the output commodity recommendation information is more accurate, and the problem that the output commodity recommendation information is one-to-one-for-one is avoided
In this embodiment, respectively obtaining an exposure vector corresponding to the commodity attribute information of an exposure product and a sequence vector corresponding to the commodity attribute information of a sequence product in a preset attribute identification model includes:
and converting the text format of the commodity attribute information into a numerical vector through the attribute identification model.
Further, the manner of extracting the sequence vector corresponding to the commodity attribute information of the sequence product and the exposure vector corresponding to the commodity attribute information of the exposure product may also be:
because the extracted commodity attribute information is in a text format, and the input requirement of the recommendation model is a standard numerical vector, the text format of the commodity attribute information needs to be converted into each numerical vector through a preset attribute identification model, wherein the attribute identification model is used for converting the commodity attribute information into the input vector which accords with the recommendation model, and the text format is converted into the numerical vector in two ways, one way represents a word in a one-hot mode, and the other way represents a word embedding mode (word embedding).
Specifically, the one-hot manner represents one word by a one-hot matrix, which is a matrix having 1 as one element per line and 0 as the other elements. For each word in the dictionary, a number is allocated, and when a certain text is coded, each word in the dictionary is converted into a one-hot matrix with the position of 1 corresponding to the word number in the dictionary. The problem with this approach is that it ignores the correlation between words and the vector length of each word is the length of the dictionary.
Preferably, the word embedding method assigns a vector representation with a fixed length to each word through the embedding matrix, the length can be set by itself, such as 300, which is actually much smaller than the dictionary length (such as 10000), and the included angle value between two word vectors can be used as a measure of the relationship between them. For example, an embedding matrix consisting of embedded vectors for each word is
By simple cosine function, the similarity of two words can be calculated
Wherein, A is an embedded vector of a certain word, and B is a word vector pre-stored in a corpus database; and when the similarity is greater than or equal to a preset similarity threshold, determining the numerical vector of the word by using the current embedded vector. Essentially, the attribute recognition model maps or embeds (embedding) a certain text word in the text space) To another numeric vector space to improve the efficiency of fast text conversion into numeric vectors.
Specifically, the corpus database sets a weight size for each word in advance, the weight corresponding to each commodity attribute information is different, and the sum of scalar products of a plurality of word vectors for the weight of each word is applied, thereby creating a word vector of commodity attribute information. The numeric vector may correspond to one or more dimensions, the number of which may be preset when vectorized, e.g., when the product attribute information is "eighteen years" in the application age category may match the corresponding numeric vector to (2, 1).
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the computer program is executed. The storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disk, a Read-Only Memory (ROM), or a Random Access Memory (RAM).
It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and may be performed in other orders unless explicitly stated herein. Moreover, at least a portion of the steps in the flow chart of the figure may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of execution is not necessarily sequential, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
With further reference to fig. 4, as an implementation of the method shown in fig. 2, the present application provides an embodiment of a deep learning commodity recommendation device, where the embodiment of the device corresponds to the embodiment of the method shown in fig. 2, and the device may be specifically applied to various electronic devices.
As shown in fig. 4, the deep learning merchandise recommendation device according to the present embodiment includes: an acquisition module 401, a processing module 402, a classification module 403, and an execution module 404. Wherein:
an obtaining module 401, configured to obtain a commodity list of a user, where the commodity list includes commodity information of an exposure product and a sequence product;
a processing module 402, configured to perform vectorization processing on the commodity information to obtain a mean vector corresponding to the commodity information;
the classification module 403 is configured to input the mean vector into a preset recommendation model to perform commodity classification, and output commodity recommendation information corresponding to the mean vector;
and the execution module 404 is configured to recommend the commodity recommendation information to the user.
Further, the obtaining module further includes:
the detection submodule is used for marking the target commodity as an exposure product if the time for the user to browse the target commodity in the commodity interface is detected to exceed a preset time value, and otherwise, marking the target commodity as a sequence product;
the recording submodule is used for respectively recording the commodity information of the exposure product and the sequence product in a commodity list;
and the acquisition submodule is used for acquiring the commodity information of the preset quantity in the same preset time period from the commodity list.
Further, the processing module further comprises:
the extraction submodule is used for respectively extracting the commodity attribute information of the exposure product and the sequence product;
the vector submodule is used for inputting the commodity attribute information into a preset attribute identification model to respectively obtain an exposure vector corresponding to the commodity attribute information of the exposure product and a sequence vector corresponding to the commodity attribute information of the sequence product;
and the mean value submodule is used for carrying out mean value processing on the sequence vector and the exposure vector to obtain a mean value vector corresponding to the same attribute type, wherein the mean value vector comprises a first vector corresponding to the sequence vector and a second vector corresponding to the exposure vector.
Further, the extracting sub-module further includes:
and the identification unit is used for respectively identifying first commodity attribute information corresponding to the sequence product and second commodity attribute information corresponding to the exposure product from the commodity information through a preset convolutional neural network model.
Further, the execution module further includes:
the calculation submodule is used for calculating the similarity value of a first vector and a second vector in the mean value vector;
the style sub-module is used for determining a style model corresponding to the commodity information according to the similarity value;
and the recommending submodule is used for outputting the commodity recommending information matched with the style model.
Further, the vector submodule further includes:
and the conversion unit is used for converting the text format of the commodity attribute information into a numerical vector through the attribute identification model.
Further, the style sub-module further comprises:
the attribute subunit is used for determining the final attribute type of the commodity information according to the similarity value;
and the style recommending subunit is used for inputting the final attribute type into a recommending model and acquiring a style model corresponding to the commodity information.
With regard to the deep learning merchandise recommendation device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment related to the method, and will not be elaborated herein.
In order to solve the technical problem, an embodiment of the present application further provides a computer device. Referring to fig. 5, fig. 5 is a block diagram of a basic structure of a computer device according to the present embodiment.
The computer device 5 comprises a memory 51, a processor 52, a network interface 53, which are communicatively connected to each other via a system bus. It is noted that only a computer device 5 having components 51-53 is shown, but it is understood that not all of the shown components are required to be implemented, and that more or fewer components may be implemented instead. As will be understood by those skilled in the art, the computer device is a device capable of automatically performing numerical calculation and/or information processing according to instructions set or stored in advance, and the hardware thereof includes but is not limited to a microprocessor, an Application Specific Integrated Circuit (ASIC), a Programmable Gate Array (FPGA), a Digital Signal Processor (DSP), an embedded device, and the like.
The computer device can be a desktop computer, a notebook, a palm computer, a cloud server and other computing devices. The computer equipment can carry out man-machine interaction with a user through a keyboard, a mouse, a remote controller, a touch panel or voice control equipment and the like.
The memory 51 includes at least one type of readable storage medium including a flash memory, a hard disk, a multimedia card, a card type memory (e.g., SD or D deep learning commodity recommendation memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a Read Only Memory (ROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), a Programmable Read Only Memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 51 may be an internal storage unit of the computer device 5, such as a hard disk or a memory of the computer device 5. In other embodiments, the memory 51 may also be an external storage device of the computer device 5, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like, which are provided on the computer device 5. Of course, the memory 51 may also comprise both an internal storage unit of the computer device 5 and an external storage device thereof. In this embodiment, the memory 51 is generally used for storing an operating system installed in the computer device 5 and various application software, such as program codes of a deep learning commodity recommendation method. Further, the memory 51 may also be used to temporarily store various types of data that have been output or are to be output.
The processor 52 may be a Central Processing Unit (CPU), controller, microcontroller, microprocessor, or other data Processing chip in some embodiments. The processor 52 is typically used to control the overall operation of the computer device 5. In this embodiment, the processor 52 is configured to run a program code stored in the memory 51 or process data, for example, a program code of the deep learning merchandise recommendation method.
The network interface 53 may comprise a wireless network interface or a wired network interface, and the network interface 53 is generally used for establishing a communication connection between the computer device 5 and other electronic devices.
The present application further provides another embodiment, which is to provide a computer-readable storage medium storing a deep-learning item recommendation program, which is executable by at least one processor to cause the at least one processor to perform the steps of the deep-learning item recommendation method as described above.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present application may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present application.
It should be understood that the above-described embodiments are merely exemplary of some, and not all, embodiments of the present application, and that the drawings illustrate preferred embodiments of the present application without limiting the scope of the claims appended hereto. This application is capable of embodiments in many different forms and is provided for the purpose of enabling a thorough understanding of the disclosure of the application. Although the present application has been described in detail with reference to the foregoing embodiments, it will be apparent to one skilled in the art that the present application may be practiced without modification or with equivalents of some of the features described in the foregoing embodiments. All equivalent structures made by using the contents of the specification and the drawings of the present application are directly or indirectly applied to other related technical fields and are within the protection scope of the present application.
Claims (7)
1. A deep-learning merchandise recommendation method, the method comprising:
if the time that a user browses a target commodity in a commodity interface is detected to exceed a preset time value, marking the target commodity as an exposure product, and otherwise, marking the target commodity as a sequence product; respectively recording commodity information of the exposure product and the sequence product in a commodity list; acquiring commodity information of a preset number in the same preset time period from a commodity list;
respectively extracting commodity attribute information of the exposure product and the sequence product; inputting the commodity attribute information into a preset attribute identification model, and respectively acquiring an exposure vector corresponding to the commodity attribute information of the exposure product and a sequence vector corresponding to the commodity attribute information of the sequence product; performing mean processing on the sequence vector and the exposure vector to obtain a mean vector corresponding to the same attribute type, wherein the mean vector comprises a first vector corresponding to the sequence vector and a second vector corresponding to the exposure vector;
calculating similarity values of a first vector and a second vector in the mean value vector; determining a style model corresponding to the commodity information according to the similarity value; outputting commodity recommendation information matched with the style model;
recommending the commodity recommendation information to the user.
2. The deep-learning commodity recommendation method according to claim 1, wherein the extracting commodity attribute information of the exposure product and the sequence product, respectively, comprises:
and respectively identifying first commodity attribute information corresponding to the sequence product and second commodity attribute information corresponding to the exposure product from the commodity information through a preset convolutional neural network model.
3. The deep learning commodity recommendation method according to claim 1, wherein the inputting the commodity attribute information into a preset attribute recognition model, and the obtaining the exposure vector corresponding to the commodity attribute information of the exposure product and the sequence vector corresponding to the commodity attribute information of the sequence product respectively comprises:
and converting the text format of the commodity attribute information into a numerical vector format through the attribute identification model, wherein the numerical vector comprises an exposure vector and a sequence vector.
4. The deep-learning commodity recommendation method according to claim 1, wherein the determining the style model corresponding to the commodity information according to the similarity value comprises:
determining the final attribute type of the commodity information according to the similarity value;
and inputting the final attribute type into a recommendation model, and acquiring a style model corresponding to the commodity information.
5. A deep learning merchandise recommendation device, comprising:
the system comprises an acquisition module, a display module and a display module, wherein the acquisition module is used for acquiring a commodity list of a user, and the commodity list comprises commodity information of an exposure product and a sequence product;
the processing module is used for vectorizing the commodity information to obtain a mean vector corresponding to the commodity information;
the classification module is used for inputting the mean vector into a preset recommendation model for commodity classification and outputting commodity recommendation information corresponding to the mean vector;
the execution module is used for recommending the commodity recommendation information to the user;
the acquisition module includes: the detection submodule is used for marking the target commodity as an exposure product if the time for the user to browse the target commodity in the commodity interface is detected to exceed a preset time value, and otherwise, marking the target commodity as a sequence product; the recording submodule is used for respectively recording the commodity information of the exposure product and the sequence product in a commodity list;
the processing module comprises: the extraction sub-module is used for respectively extracting commodity attribute information of the exposure product and the sequence product; the vector sub-module is used for inputting the commodity attribute information into a preset attribute identification model to respectively obtain an exposure vector corresponding to the commodity attribute information of the exposure product and a sequence vector corresponding to the commodity attribute information of the sequence product; the mean value submodule is used for carrying out mean value processing on the sequence vector and the exposure vector to obtain a mean value vector corresponding to the same attribute type, wherein the mean value vector comprises a first vector corresponding to the sequence vector and a second vector corresponding to the exposure vector;
the execution module comprises: the calculation submodule is used for calculating the similarity value of a first vector and a second vector in the mean value vector; the style submodule is used for determining a style model corresponding to the commodity information according to the similarity value; and the recommending submodule is used for outputting the commodity recommending information matched with the style model.
6. A computer device comprising a memory having stored therein a computer program and a processor which when executed implements the steps of the deep-learning item recommendation method of any one of claims 1 to 4.
7. A computer-readable storage medium, characterized in that a computer program is stored thereon, which computer program, when being executed by a processor, carries out the steps of the deep-learning item recommendation method according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910871299.0A CN110827112B (en) | 2019-09-16 | 2019-09-16 | Deep learning commodity recommendation method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910871299.0A CN110827112B (en) | 2019-09-16 | 2019-09-16 | Deep learning commodity recommendation method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110827112A CN110827112A (en) | 2020-02-21 |
| CN110827112B true CN110827112B (en) | 2023-02-03 |
Family
ID=69548120
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910871299.0A Active CN110827112B (en) | 2019-09-16 | 2019-09-16 | Deep learning commodity recommendation method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110827112B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111439268B (en) * | 2020-03-31 | 2023-03-14 | 重庆长安汽车股份有限公司 | Method and device for actively providing personalized service, cloud server and automobile |
| CN111639989B (en) * | 2020-04-28 | 2024-02-02 | 上海秒针网络科技有限公司 | Commodity recommendation method and readable storage medium |
| CN113971471B (en) * | 2020-07-22 | 2025-03-25 | 上海顺如丰来技术有限公司 | Product information prediction model construction method and product information prediction method |
| CN112200636B (en) * | 2020-10-24 | 2024-10-22 | 内蒙古草都草牧业股份有限公司 | Intelligent shopping recommendation method based on big data |
| CN113822735A (en) * | 2021-02-24 | 2021-12-21 | 北京沃东天骏信息技术有限公司 | Goods recommendation method and device, storage medium and electronic equipment |
| CN113392325A (en) * | 2021-06-21 | 2021-09-14 | 电子科技大学 | Deep learning-based information recommendation method |
| CN113850638A (en) * | 2021-08-04 | 2021-12-28 | 欧冶工业品股份有限公司 | Purchasing strategy method and system based on industrial product digital matching |
| CN114282976B (en) * | 2021-12-27 | 2024-08-23 | 赛尔网络有限公司 | Vendor recommendation method and device, electronic equipment and medium |
| CN114896318A (en) * | 2022-05-10 | 2022-08-12 | 飞书深诺数字科技(上海)股份有限公司 | Method, device and electronic device for data synchronization |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106959966A (en) * | 2016-01-12 | 2017-07-18 | 腾讯科技（深圳）有限公司 | A kind of information recommendation method and system |
| CN107451894A (en) * | 2017-08-03 | 2017-12-08 | 北京京东尚科信息技术有限公司 | Data processing method, device and computer-readable recording medium |
| CN108665329A (en) * | 2017-03-29 | 2018-10-16 | 北京京东尚科信息技术有限公司 | A kind of Method of Commodity Recommendation based on user browsing behavior |
| CN108776914A (en) * | 2018-05-22 | 2018-11-09 | 广州品唯软件有限公司 | A kind of user individual goods browse method and device based on neural network |
| WO2019153518A1 (en) * | 2018-02-08 | 2019-08-15 | 平安科技（深圳）有限公司 | Information pushing method and device, computer device and storage medium |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105469263A (en) * | 2014-09-24 | 2016-04-06 | 阿里巴巴集团控股有限公司 | Commodity recommendation method and device |
| US20180374138A1 (en) * | 2017-06-23 | 2018-12-27 | Vufind Inc. | Leveraging delayed and partial reward in deep reinforcement learning artificial intelligence systems to provide purchase recommendations |
-
2019
- 2019-09-16 CN CN201910871299.0A patent/CN110827112B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106959966A (en) * | 2016-01-12 | 2017-07-18 | 腾讯科技（深圳）有限公司 | A kind of information recommendation method and system |
| CN108665329A (en) * | 2017-03-29 | 2018-10-16 | 北京京东尚科信息技术有限公司 | A kind of Method of Commodity Recommendation based on user browsing behavior |
| CN107451894A (en) * | 2017-08-03 | 2017-12-08 | 北京京东尚科信息技术有限公司 | Data processing method, device and computer-readable recording medium |
| WO2019153518A1 (en) * | 2018-02-08 | 2019-08-15 | 平安科技（深圳）有限公司 | Information pushing method and device, computer device and storage medium |
| CN108776914A (en) * | 2018-05-22 | 2018-11-09 | 广州品唯软件有限公司 | A kind of user individual goods browse method and device based on neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110827112A (en) | 2020-02-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110827112B (en) | Deep learning commodity recommendation method and device, computer equipment and storage medium | |
| CN110825957B (en) | Deep learning-based information recommendation method, device, equipment and storage medium | |
| CN110866181B (en) | Resource recommendation method, device and storage medium | |
| US10664888B2 (en) | Method and system for attribute extraction from product titles using sequence labeling algorithms | |
| CN109299994B (en) | Recommendation method, device, equipment and readable storage medium | |
| CN113449187B (en) | Product recommendation method, device, equipment and storage medium based on double images | |
| CN111523976A (en) | Commodity recommendation method and device, electronic equipment and storage medium | |
| CN111125495A (en) | Information recommendation method, equipment and storage medium | |
| CN113643103A (en) | Product recommendation method, device, equipment and storage medium based on user similarity | |
| CN105718184A (en) | Data processing method and apparatus | |
| CN104216881A (en) | Method and device for recommending individual labels | |
| CN111598596B (en) | Data processing method, device, electronic device and storage medium | |
| CN118628214A (en) | A personalized clothing recommendation method and system for e-commerce platforms based on artificial intelligence | |
| WO2019072098A1 (en) | Method and system for identifying core product terms | |
| CN113886708A (en) | Product recommendation method, device, equipment and storage medium based on user information | |
| CN117951386A (en) | Keyword recommendation method and device, electronic equipment and readable storage medium | |
| CN116703515A (en) | Recommendation method and device based on artificial intelligence, computer equipment and storage medium | |
| CN113379482B (en) | Article recommendation method, computing device and storage medium | |
| CN113821596A (en) | Information recommendation method and device, computer equipment and storage medium | |
| CN113327145A (en) | Article recommendation method and device | |
| CN110110199B (en) | Information output method and device | |
| CN113781180B (en) | Article recommendation method and device, electronic equipment and storage medium | |
| CN110837596B (en) | Intelligent recommendation method and device, computer equipment and storage medium | |
| CN112288512B (en) | Information processing method, information processing device, electronic equipment and readable storage medium | |
| CN116776932A (en) | E-commerce behavior recognition method and device for user |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |