CN110909258B - Information recommendation method, device, equipment and storage medium - Google Patents
Information recommendation method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN110909258B CN110909258B CN201911155587.2A CN201911155587A CN110909258B CN 110909258 B CN110909258 B CN 110909258B CN 201911155587 A CN201911155587 A CN 201911155587A CN 110909258 B CN110909258 B CN 110909258B
- Authority
- CN
- China
- Prior art keywords
- information
- model
- target user
- positive sample
- training
- 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
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
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses an information recommendation method, an information recommendation device, information recommendation equipment and a storage medium. Wherein the method comprises the following steps: the identification of the target user and the historical access record of the target user are obtained from the database to determine the to-be-recommended information, wherein the to-be-recommended information comprises any information in the database, and the to-be-recommended information is recommended to the target user, so that personalized recommendation of any information in the database to the user is realized, and the defect that in the prior art, the latest information updated in the database is difficult to recommend to the user through an item_CF algorithm is avoided.
Description
Technical Field
The embodiment of the invention relates to an information processing technology, in particular to an information recommending method, an information recommending device, information recommending equipment and a storage medium.
Background
In an information recommendation system, information is generally recommended to a user in various forms, wherein a representation method of the information directly affects the recommendation effect of the information. In the conventional information representation method, a one-hot (one-hot) process is usually performed on the identifier corresponding to the information, but this method can generate a high-dimensional sparse vector and restrict the subsequent use of the information identifier. Moreover, when the correlation between information is measured using the Item-based collaborative filtering (item_cf, item Collaboration Filter) algorithm to calculate the jaccard coefficient between information, the latest information updated by the platform has not been accessed by the user. Therefore, the jaccard coefficient between the latest information and other information cannot be obtained, which results in difficulty in recommending the latest information updated on the platform to the user.
Disclosure of Invention
In order to solve at least one of the above technical problems, the embodiments of the present invention provide the following solutions.
In a first aspect, an embodiment of the present invention provides an information recommendation method, where the method includes:
acquiring the identification of the target user and the historical access record of the target user from a database;
determining to-be-recommended information according to the identification of the target user and the historical access record of the target user;
the information to be recommended comprises any information in a database;
and recommending the to-be-recommended information to the target user.
In a second aspect, an embodiment of the present invention further provides an information recommendation apparatus, where the apparatus includes:
the acquisition module is used for acquiring the identification of the target user and the historical access record of the target user from the database;
the determining module is used for determining the to-be-recommended information according to the identification of the target user and the historical access record of the target user;
the information to be recommended comprises any information in a database;
and the recommending module is used for recommending the to-be-recommended information to the target user.
In a third aspect, an embodiment of the present invention further provides an information recommendation apparatus, including:
the information recommendation method according to the first aspect of the present invention is implemented when the processor executes the computer program.
In a fourth aspect, an embodiment of the present invention further provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements an information recommendation method as provided in the first aspect of the present invention.
The embodiment of the invention provides an information recommending method, device, equipment and storage medium, which can determine to-be-recommended information by acquiring the identification of a target user and the historical access record of the target user from a database, wherein the to-be-recommended information comprises any information in the database and recommends the to-be-recommended information to the target user, so that personalized recommendation of any information in the database to the user is realized, and the defect that in the prior art, the latest information which is difficult to update in the database is recommended to the user by an item_CF algorithm is avoided.
Drawings
FIG. 1 is a flow chart of an information recommendation method in an embodiment of the invention;
FIG. 2 is a flowchart of another information recommendation method in an embodiment of the present invention;
FIG. 3 is a schematic diagram of a training model in an embodiment of the invention;
FIG. 4 is a flowchart of another information recommendation method according to an embodiment of the present invention
FIG. 5 is a schematic diagram of an information recommendation device according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an information recommendation apparatus in an embodiment of the present invention.
Detailed Description
The invention is described in further detail below with reference to the drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting thereof. It should be further noted that, for convenience of description, only some, but not all of the structures related to the present invention are shown in the drawings.
In addition, in the embodiments of the present invention, words such as "optionally" or "exemplary" are used to mean serving as examples, illustrations, or descriptions. Any embodiment or design described herein as "optional" or "exemplary" is not to be construed as preferred or advantageous over other embodiments or designs. Rather, the use of the words "optionally" or "illustratively" and the like is intended to present the relevant concepts in a concrete manner.
Fig. 1 is a schematic diagram of an information recommendation method according to an embodiment of the present invention, where the method may be applied to each network platform to recommend various types of information to a user of the network platform, and as shown in fig. 1, the method may include:
s101, acquiring the identification of the target user and the historical access record of the target user from a database.
The information recommending method provided by the embodiment can be applied to recommending various information to the user in each network platform, so that the database in the step can be understood as a background database of each network platform. The target user can be any user browsing and accessing the network platform, and the identification of the target user can be various identifications such as a user name registered by the user on the network platform, a terminal equipment number accessing the network platform and the like.
Since the user browses information on the network platform, such as watching a movie, listening to a broadcast, music, etc., the database of the network platform automatically records the user's identification and access tracks, the user's identification and history access records can be obtained from the database.
S102, determining the recommendation information according to the identification of the target user and the historical access record of the target user.
And taking any user accessing the network platform as a target user, and determining information to be recommended to the user by the network platform according to the user identification and the history access record after the identification and the history access record of the user are acquired.
Illustratively, the information that the network platform is to recommend to the user may be determined from the recommendation model by the identification of the target user and the historical access record of the target user. The recommendation model may be trained based on a historical access record of the user, and the information to be recommended to the user determined from the recommendation model may include any information in a database, for example, information that the user has never browsed to access on a network platform.
S103, recommending the to-be-recommended information to the target user.
After the information to be recommended to the user by the network platform is obtained through the step S102, the information to be recommended can be recommended to the target user, so that any information on the network platform is recommended to the user in a personalized mode, and the defect that the latest information which is difficult to update in the user recommendation database by the item_cf algorithm in the prior art is avoided.
The embodiment provides an information recommendation method, which can determine to-be-recommended information by acquiring an identification of a target user and a history access record of the target user from a database, wherein the to-be-recommended information comprises any information in the database, and recommends the to-be-recommended information to the target user, so that personalized recommendation of any information in the database to the user is realized, and the defect that in the prior art, the latest information updated in the database is difficult to recommend to the user through an item_CF algorithm is avoided.
Fig. 2 is a flowchart of an information recommendation method provided in this embodiment, where the determining process of the recommendation model in step S102 is described in further detail, as shown in fig. 2, and the method includes:
s201, constructing a training model.
Alternatively, in the present embodiment, the training model may include an input layer, an embedded layer, an average layer, and a multi-classification layer. Of course, a person skilled in the art may also use a model with other structures, but the information recommendation method provided by the embodiment of the present invention is implemented by using a designed model, which belongs to the protection scope of the present invention.
Illustratively, the training model structure constructed as described above is shown in FIG. 3.
S202, acquiring a historical access sequence of the user from a database.
The historical access record of the user can be obtained from the database for a certain period of time, for example, if the user browses a certain music website, the playing record of the last two weeks of the user can be obtained from the database of the website, and the historical playing record is ordered according to the order of the playing time, so that the historical access sequence of the user is obtained.
Illustratively, when the user has played album 1, album 2, album 3, and album 4 on the music website in the last two weeks, the resulting historical access sequence for the user may be [ album 1, album 2, album 3, and album 4 ].
S203, training data is generated according to the historical access sequence.
In this embodiment, the training data may be generated from the historical access sequence by a data processing rule, for example, the training data may be generated from the historical access sequence according to a skip-gram rule. The skip-gram rule is used for predicting front and back information related to current information, the generated training data can comprise information characteristics, a first label and a second label of each sample information in the database, the form of the training data can be [ information characteristics, the first label and the second label ], wherein the first label is used for representing the prediction information of the current sample information, and the second label is used for representing whether the current sample information is a positive sample or not.
For example, if the window length of skip-gram is 1, the training data based on the history access sequence generated in step S202 may be the information feature of [ album 1 ], the first label=album 2, the second label=1 ], the information feature of [ album 2 ], the first label=album 1, the second label=1 ], the information feature of [ album 2 ], the first label=album 3, the second label=1 ], the information feature of [ album 3, the first label=2, the second label=1 ], the information feature of [ album 3, the first label=album 4, the second label=1 ], the information feature of [ album a, the first label=k, the second label=0 ], the information feature of [ album d, the first label=album p, and the second label=0 ].
Assuming that in the training data, the second label is 1 to indicate that the sample is a positive sample (i.e., information visited by the user), and the second label is 0 to indicate that the sample is a negative sample (i.e., information not visited by the user), it can be seen that, because the positive sample is information visited by the user, the information represented by the first label in the positive sample generated based on the history access sequence is information related to the current sample before and after the current sample information, and the negative sample is a sample generated through negative sampling, and the sample is not browsed by the user, so that the information represented by the first label in the negative sample is in a random relation with the current information.
Further, in the training data, the information features may be information identification, primary classification of information, secondary classification of information, information tag, and other features.
S204, inputting training data into a training model for training to obtain a recommended model.
In this embodiment, assuming that the information features in the training data include four features, that is, an information identifier, a first-stage classification of information, a second-stage classification of information, and an information tag, where the four features are respectively marked as a first feature, a second feature, a third feature, and a fourth feature, the four information features of the positive sample in the training data may be input into an embedding layer of the training model, as shown in fig. 3, and four feature vectors may be obtained through the embedding layer of the training model, where the four feature vectors are feature vector sequences of the positive sample.
Further, a first correction value L is calculated and generated based on the feature vector sequence of the positive sample and the auxiliary loss function i1 。
The auxiliary loss function may be, for example
Wherein, the liquid crystal display device comprises a liquid crystal display device,transpose of the first eigenvector in the sequence of positive sample eigenvectors, < >>Is the second eigenvector in the sequence of positive sample eigenvectors,>a third eigenvector, which is a sequence of positive sample eigenvectors,>a fourth eigenvector that is a sequence of positive sample eigenvectors.
The similarity between the first feature, the second feature, the third feature and the fourth feature of the information can be constrained based on the auxiliary loss function, namely, the conversion rule between different information is transferred to other features through the information identification. In this way, even if the user does not access a certain information, a corresponding information vector can be generated based on the characteristics thereof.
Generating a second correction value L according to the positive sample, the negative sample corresponding to the positive sample and the model loss function in the training data i2 Wherein the model loss function is
Wherein, the liquid crystal display device comprises a liquid crystal display device,for transposition of the average value of the feature vectors in the sequence of positive sample feature vectors, σ is a nonlinear function, e.g. sigmoid function, W can be chosen j First as positive sampleParameter vector of multi-classification layer of label corresponding prediction information, W z For the parameter vector of the negative sample first label corresponding to the positive sample, Z is the number of negative samples.
For example, the four features of the information identification, the first-stage classification, the second-stage classification and the information label of the positive sample in the training data are input into the embedded layer of the training model to obtain four feature vectors, the average layer of the training model is used for averaging the four feature vectors to obtain the average vector of the feature vectors, the average vector is the information vector, and then the transpose of the average vector isAnd further, calculating the correlation between each information vector and the first label in the multi-classification layer through the model loss function, namely measuring the incorrect degree of the transfer of the training model fitting current information to the information corresponding to the first label.
Then, a model correction value L is generated based on the calculated first correction value, second correction value, and integrated loss function.
Wherein the comprehensive loss function can be
Wherein, the value range of alpha is 0-0.1, and m is the number of positive samples.
It will be appreciated by those skilled in the art that different samples of the training data are input in the training model, and that the first correction value, the second correction value, and the model correction value may be generated differently.
Therefore, parameters in the training model can be continuously trained and optimized through the auxiliary loss function, the model loss function, the comprehensive loss function and the model correction value, so that a recommended model is obtained.
Fig. 4 is a schematic diagram of an information recommendation method provided in this embodiment, where the process of determining the information to be recommended in the step S102 is further described, as shown in fig. 4, and the method includes:
s401, information vectors of all information in the database are obtained from the recommendation model.
For example, after the recommendation model is obtained, information features of all information in the database may be input into the recommendation model, for example, assuming that the information features are four features of information identification, primary classification of information, secondary classification of information and information label, the four features of all information are input into the recommendation model, and information vectors of all information can be obtained based on an average layer of the recommendation model.
Of course, if updated latest information exists in the database, the first-order classification, the second-order classification, and the information label of the latest information may be input to the embedding layer of the recommendation model to obtain vectors corresponding to the respective features, and then each vector may be input to the averaging layer of the recommendation model, and the average value of the vectors corresponding to the three features of the latest information may be calculated in the averaging layer to obtain the information vector of the latest information.
S402, calculating cosine values among information vectors of all the information.
Since the information vectors of all the information in the data are obtained through the above-described step S401, the information vector between the information pairs can be calculated to obtain the cosine value between all the information vectors.
S403, determining the quasi-recommended information according to the identification of the target user, the historical access record of the target user and the cosine value.
After the history access record of the target user is obtained, the cosine value between the information vector of a certain information and the cosine value between the information vectors of other information in the history access record can be further selected from the cosine values between the information vectors of all the information in step S402, and the first plurality of other information with larger cosine values are used as related information of the certain information, so that the determined related information is used as to-be-recommended information recommended to the user.
For example, assuming that the identifier of the user accessing a certain network platform is xmly, determining the historical access record of the user from the database according to the user identifier, and assuming that the information accessed by the user last time is music a, based on the cosine values between the information vectors of all the information calculated in step S402, the cosine values between the music a and other information in the database can be obtained, and the information corresponding to the first 20 cosine values with larger cosine values is selected as the relevant information of the music a, and further, the relevant information obtained by screening can be determined as the recommendation-like information to be recommended to the user.
By the method, when the latest information is updated in the database and is not accessed by the user, the latest information can be individually recommended to the user, and the defect that the latest information updated in the database is difficult to recommend to the user through the item_CF algorithm in the prior art is overcome.
Fig. 5 provides an information recommendation apparatus, as shown in fig. 5, including: an acquisition module 501, a determination module 502 and a recommendation module 503;
the acquisition module is used for acquiring the identification of the target user and the historical access record of the target user from the database;
the determining module is used for determining the to-be-recommended information according to the identification of the target user and the historical access record of the target user;
the information to be recommended comprises any information in a database;
and the recommending module is used for recommending the to-be-recommended information to the target user.
Further, the recommendation module is specifically configured to determine the to-be-recommended information from the recommendation model according to the identification of the target user and the historical access record of the target user;
the recommendation model is trained based on historical access records of users.
In one example, the recommendation module is specifically configured to obtain information vectors of all information in the database from the recommendation model, calculate cosine values between the information vectors of all information, and determine the information to be recommended according to the identification of the target user, the historical access record of the target user, and the cosine values.
Further, the information recommendation device further includes: the system comprises a construction module, a generation module and a training module;
the building module is used for building training data;
the acquisition module is used for acquiring a historical access sequence of the user from the database;
the generation module is used for generating training data according to the historical access sequence;
the training data comprises information characteristics of each sample information in the database, a first label and a second label, wherein the first label is used for representing prediction information of current sample information, and the second label is used for representing whether the current sample information is a positive sample or not.
And the training module is used for inputting training data into the training model to train so as to obtain a recommended model.
Further, the training module is specifically configured to input information features of positive samples in training data into an embedding layer of the training model to obtain a positive sample feature vector sequence; generating a first correction value according to the positive sample characteristic vector sequence and the auxiliary loss function; generating a second correction value according to the positive sample, the negative sample corresponding to the positive sample and the model loss function in the training data; generating a model correction value according to the first correction value, the second correction value and the comprehensive loss function; and training the training model according to the model correction value to obtain a recommended model.
Wherein the auxiliary loss function isWherein (1)>Transpose of the first eigenvector in the sequence of positive sample eigenvectors, < >>Is the second eigenvector in the sequence of positive sample eigenvectors,>a third eigenvector, which is a sequence of positive sample eigenvectors,>a fourth eigenvector, L, being a sequence of positive sample eigenvectors i1 Calculating a first correction value generated for passing through the auxiliary loss function;
model loss function ofWherein (1)>Is the transpose of the average vector of the feature vectors in the sequence of positive sample feature vectors, σ is the nonlinear function, W j A parameter vector of a multi-classification layer corresponding to the prediction information for the first label of the positive sample, W z The parameter vector of the first label of the negative sample corresponding to the positive sample, Z is the number of the negative samples, L i2 Calculating a generated second correction value for the model loss function;
the comprehensive loss function isWherein, the value range of alpha is 0-0.1, m is the number of positive samples, and L is the model correction value calculated and generated by the comprehensive loss function.
The information recommending device provided by the embodiment of the invention can execute the information recommending method provided by the first embodiment of the invention, and has the corresponding functional modules and beneficial effects of the executing method.
Fig. 6 is a schematic structural diagram of an information recommendation device according to an embodiment of the present invention, as shown in fig. 6, the device includes a processor 601, a memory 602, an input device 603, and an output device 604; the number of processors 601 in the device may be one or more, one processor 601 being taken as an example in fig. 6; the processor 601, memory 602, input means 603 and output means 604 in the device may be connected by a bus or other means, in fig. 6 by way of example.
The memory 602 is used as a computer readable storage medium for storing a software program, a computer executable program, and modules, such as program instructions/modules corresponding to the information recommendation method in fig. 1-4 (e.g., the acquisition module 501, the determination module 502, and the recommendation module 503 in the information recommendation device) according to the embodiment of the present invention. The processor 601 executes various functional applications of the device and data processing by running software programs, instructions and modules stored in the memory 602, i.e., implements the information recommendation method described above.
The memory 602 may include primarily a program storage area and a data storage area, wherein the program storage area may store an operating system, at least one application program required for functionality; the storage data area may store data created according to the use of the terminal, etc. In addition, the memory 602 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some examples, the memory 602 may further include memory remotely located relative to the processor 601, which may be connected to the device/terminal/server via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input means 603 may be used to receive input numeric or character information and to generate key signal inputs related to user settings and function control of the device. The output device 304 may include a display device such as a display screen.
Embodiments of the present invention also provide a storage medium containing computer-executable instructions, which when executed by a computer processor, are for performing an information recommendation method, the method comprising:
acquiring the identification of the target user and the historical access record of the target user from a database;
determining to-be-recommended information according to the identification of the target user and the historical access record of the target user;
the information to be recommended comprises any information in a database;
and recommending the to-be-recommended information to the target user.
Of course, the storage medium containing the computer executable instructions provided in the embodiments of the present invention is not limited to the method operations described above, and may also perform the related operations in the information recommendation method provided in any embodiment of the present invention.
From the above description of embodiments, it will be clear to a person skilled in the art that the present invention may be implemented by means of software and necessary general purpose hardware, but of course also by means of hardware, although in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a FLASH Memory (FLASH), a hard disk or an optical disk of a computer, etc., and include several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the embodiments of the present invention.
It should be noted that, in the embodiment of the information recommending apparatus, each unit and module included are only divided according to the functional logic, but not limited to the above-mentioned division, so long as the corresponding functions can be implemented; in addition, the specific names of the functional units are also only for distinguishing from each other, and are not used to limit the protection scope of the present invention.
Note that the above is only a preferred embodiment of the present invention and the technical principle applied. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, while the invention has been described in connection with the above embodiments, the invention is not limited to the embodiments, but may be embodied in many other equivalent forms without departing from the spirit or scope of the invention, which is set forth in the following claims.
Claims (5)
1. An information recommendation method, comprising:
acquiring the identification of a target user and a historical access record of the target user from a database;
determining to-be-recommended information according to the identification of the target user and the historical access record of the target user, wherein the to-be-recommended information comprises any information in the database;
recommending the to-be-recommended information to the target user;
the determining the quasi-recommended information according to the identification of the target user and the historical access record of the target user comprises the following steps:
determining the quasi-recommended information from a recommendation model according to the identification of the target user and the historical access record of the target user; the recommendation model is trained based on historical access records of users;
the recommendation model is determined by the following steps:
constructing a training model;
acquiring a historical access sequence of a user from the database;
generating training data according to the historical access sequence;
inputting the training data into the training model for training to obtain a recommended model;
the training data comprises information characteristics, a first label and a second label of each sample information in the database;
the first label is used for representing prediction information of current sample information, and the second label is used for representing whether the current sample information is a positive sample or not;
the step of inputting the training data into the training model for training to obtain a recommended model comprises the following steps:
inputting the information characteristics of the positive sample in the training data into an embedded layer of the training model to obtain a positive sample characteristic vector sequence;
generating a first correction value according to the positive sample characteristic vector sequence and the auxiliary loss function;
generating a second correction value according to the positive sample in the training data, the negative sample corresponding to the positive sample and the model loss function;
generating a model correction value according to the first correction value, the second correction value and the comprehensive loss function;
training the training model according to the model correction value to obtain a recommended model;
the auxiliary loss function is
Wherein->Transpose of the first eigenvector in the sequence of positive sample eigenvectors, < >>Is the second eigenvector in the sequence of positive sample eigenvectors,>a third eigenvector, which is a sequence of positive sample eigenvectors,>fourth eigenvector, which is a sequence of positive sample eigenvectors,>calculating a first correction value generated for passing through the auxiliary loss function;
the model loss function isWherein->For each feature in the sequence of positive sample feature vectorsTranspose of the average value of the vectors, ">As a nonlinear function of +.>The first label, which is a positive sample, corresponds to a parameter vector of the multi-classification layer of the prediction information, ++>The parameter vector of the first label being the negative sample corresponding to the positive sample, Z is the number of negative samples, +.>Calculating a generated second correction value for the model loss function;
the comprehensive loss function isWherein->The value range of (2) is 0-0.1, m is the number of positive samples, and L is a model correction value calculated and generated through the comprehensive loss function.
2. The method of claim 1, wherein determining the to-be-recommended information from a recommendation model based on the identification of the target user and the historical access record of the target user comprises:
obtaining information vectors of all information in the database from the recommendation model;
calculating cosine values among information vectors of all the information;
and determining the quasi-recommended information according to the identification of the target user, the historical access record of the target user and the cosine value.
3. An information recommendation device, characterized by comprising:
the acquisition module is used for acquiring the identification of the target user and the historical access record of the target user from the database;
the determining module is used for determining the quasi-recommended information according to the identification of the target user and the historical access record of the target user;
wherein the to-be-recommended information comprises any information in the database;
the recommending module is used for recommending the to-be-recommended information to the target user;
the determining module is specifically configured to determine, according to the identification of the target user and the historical access record of the target user, to-be-recommended information from a recommendation model; the recommendation model is trained based on historical access records of users;
the recommendation model is determined by the following steps:
constructing a training model;
acquiring a historical access sequence of a user from the database;
generating training data according to the historical access sequence;
inputting the training data into the training model for training to obtain a recommended model;
the training data comprises information characteristics, a first label and a second label of each sample information in the database; the first label is used for representing prediction information of current sample information, and the second label is used for representing whether the current sample information is a positive sample or not;
the step of inputting the training data into the training model for training to obtain a recommended model comprises the following steps:
inputting the information characteristics of the positive sample in the training data into an embedded layer of the training model to obtain a positive sample characteristic vector sequence;
generating a first correction value according to the positive sample characteristic vector sequence and the auxiliary loss function;
generating a second correction value according to the positive sample in the training data, the negative sample corresponding to the positive sample and the model loss function;
generating a model correction value according to the first correction value, the second correction value and the comprehensive loss function;
training the training model according to the model correction value to obtain a recommended model;
the auxiliary loss function is
Wherein->Transpose of the first eigenvector in the sequence of positive sample eigenvectors, < >>Is the second eigenvector in the sequence of positive sample eigenvectors,>a third eigenvector, which is a sequence of positive sample eigenvectors,>fourth eigenvector, which is a sequence of positive sample eigenvectors,>calculating a first correction value generated for passing through the auxiliary loss function;
the model loss function isWherein->Transpose the mean value of the feature vectors in the sequence of positive sample feature vectors +.>As a nonlinear function of +.>The first label, which is a positive sample, corresponds to a parameter vector of the multi-classification layer of the prediction information, ++>The parameter vector of the first label being the negative sample corresponding to the positive sample, Z is the number of negative samples, +.>Calculating a generated second correction value for the model loss function;
the comprehensive loss function isWherein->The value range of (2) is 0-0.1, m is the number of positive samples, and L is a model correction value calculated and generated through the comprehensive loss function.
4. An information recommendation device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the information recommendation method according to claim 1 or 2 when executing the computer program.
5. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the information recommendation method according to claim 1 or 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911155587.2A CN110909258B (en) | 2019-11-22 | 2019-11-22 | Information recommendation method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911155587.2A CN110909258B (en) | 2019-11-22 | 2019-11-22 | Information recommendation method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110909258A CN110909258A (en) | 2020-03-24 |
| CN110909258B true CN110909258B (en) | 2023-09-29 |
Family
ID=69818738
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911155587.2A Active CN110909258B (en) | 2019-11-22 | 2019-11-22 | Information recommendation method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110909258B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111506820B (en) * | 2020-04-30 | 2023-05-09 | 喜大（上海）网络科技有限公司 | Recommendation model, recommendation method, recommendation device, recommendation equipment and recommendation storage medium |
| CN113676505B (en) * | 2020-05-15 | 2023-11-28 | 财付通支付科技有限公司 | Information pushing method, device, computer equipment and storage medium |
| CN112559868A (en) * | 2020-12-17 | 2021-03-26 | 广州博冠信息科技有限公司 | Information recall method and device, storage medium and electronic equipment |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107154043A (en) * | 2017-06-05 | 2017-09-12 | 杭州健培科技有限公司 | A kind of Lung neoplasm false positive sample suppressing method based on 3DCNN |
| CN107391512A (en) * | 2016-05-17 | 2017-11-24 | 北京邮电大学 | The method and apparatus of knowledge mapping prediction |
| CN108021933A (en) * | 2017-11-23 | 2018-05-11 | 深圳市华尊科技股份有限公司 | Neural network recognization model and recognition methods |
| CN108804633A (en) * | 2018-06-01 | 2018-11-13 | 腾讯科技（深圳）有限公司 | The content recommendation method of Behavior-based control Semantic knowledge network |
| CN109344257A (en) * | 2018-10-24 | 2019-02-15 | 平安科技（深圳）有限公司 | Text emotion recognition methods and device, electronic equipment, storage medium |
| CN109426858A (en) * | 2017-08-29 | 2019-03-05 | 京东方科技集团股份有限公司 | Neural network, training method, image processing method and image processing apparatus |
| CN109783687A (en) * | 2018-11-22 | 2019-05-21 | 广州市易杰数码科技有限公司 | A kind of recommended method based on graph structure, device, equipment and storage medium |
| CN109902705A (en) * | 2018-10-30 | 2019-06-18 | 华为技术有限公司 | A kind of object detection model to disturbance rejection generation method and device |
| CN109961032A (en) * | 2019-03-18 | 2019-07-02 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating disaggregated model |
| CN110321361A (en) * | 2019-06-15 | 2019-10-11 | 河南大学 | Examination question based on improved LSTM neural network model recommends determination method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10423972B2 (en) * | 2015-03-03 | 2019-09-24 | Go Daddy Operating Company, LLC | System and method for domain name community network |
-
2019
- 2019-11-22 CN CN201911155587.2A patent/CN110909258B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107391512A (en) * | 2016-05-17 | 2017-11-24 | 北京邮电大学 | The method and apparatus of knowledge mapping prediction |
| CN107154043A (en) * | 2017-06-05 | 2017-09-12 | 杭州健培科技有限公司 | A kind of Lung neoplasm false positive sample suppressing method based on 3DCNN |
| CN109426858A (en) * | 2017-08-29 | 2019-03-05 | 京东方科技集团股份有限公司 | Neural network, training method, image processing method and image processing apparatus |
| CN108021933A (en) * | 2017-11-23 | 2018-05-11 | 深圳市华尊科技股份有限公司 | Neural network recognization model and recognition methods |
| CN108804633A (en) * | 2018-06-01 | 2018-11-13 | 腾讯科技（深圳）有限公司 | The content recommendation method of Behavior-based control Semantic knowledge network |
| CN109344257A (en) * | 2018-10-24 | 2019-02-15 | 平安科技（深圳）有限公司 | Text emotion recognition methods and device, electronic equipment, storage medium |
| CN109902705A (en) * | 2018-10-30 | 2019-06-18 | 华为技术有限公司 | A kind of object detection model to disturbance rejection generation method and device |
| CN109783687A (en) * | 2018-11-22 | 2019-05-21 | 广州市易杰数码科技有限公司 | A kind of recommended method based on graph structure, device, equipment and storage medium |
| CN109961032A (en) * | 2019-03-18 | 2019-07-02 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating disaggregated model |
| CN110321361A (en) * | 2019-06-15 | 2019-10-11 | 河南大学 | Examination question based on improved LSTM neural network model recommends determination method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110909258A (en) | 2020-03-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110909258B (en) | Information recommendation method, device, equipment and storage medium | |
| JP5454357B2 (en) | Information processing apparatus and method, and program | |
| CN111061946A (en) | Scenario content recommendation method and device, electronic equipment and storage medium | |
| CN111639988B (en) | Broker recommendation method, device, electronic equipment and storage medium | |
| CN110837577A (en) | Video recommendation method, device, equipment and storage medium | |
| CN111125521A (en) | Information recommendation method, device, equipment and storage medium | |
| CN110990624B (en) | Video recommendation method, device, equipment and storage medium | |
| CN109117442B (en) | Application recommendation method and device | |
| KR101450453B1 (en) | Method and apparatus for recommending contents | |
| CN111159563A (en) | Method, device and equipment for determining user interest point information and storage medium | |
| CN110598109A (en) | Information recommendation method, device, equipment and storage medium | |
| CN111104599A (en) | Method and apparatus for outputting information | |
| Rubinstein et al. | How to deal with the curse of dimensionality of likelihood ratios in Monte Carlo simulation | |
| US20150379112A1 (en) | Creating an on-line job function ontology | |
| CN111523035A (en) | Recommendation method, device, server and medium for APP browsing content | |
| CN111368195B (en) | Model training method, device, equipment and storage medium | |
| CN110827078B (en) | Information recommendation method, device, equipment and storage medium | |
| CN116610872B (en) | Training method and device for news recommendation model | |
| CN116628346A (en) | Training method and device for search word recommendation model | |
| CN113836388A (en) | Information recommendation method and device, server and storage medium | |
| JP6896594B2 (en) | Music selection device, music selection method and program | |
| CN111507471B (en) | Model training method, device, equipment and storage medium | |
| US10853820B2 (en) | Method and apparatus for recommending topic-cohesive and interactive implicit communities in social customer relationship management | |
| CN113822684B (en) | Black-birth user identification model training method and device, electronic equipment and storage medium | |
| CN112200602A (en) | Neural network model training method and device for advertisement recommendation |
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 |