CN116028723A - Data recommendation method, device, equipment and computer storage medium - Google Patents
Data recommendation method, device, equipment and computer storage medium Download PDFInfo
- Publication number
- CN116028723A CN116028723A CN202111234461.1A CN202111234461A CN116028723A CN 116028723 A CN116028723 A CN 116028723A CN 202111234461 A CN202111234461 A CN 202111234461A CN 116028723 A CN116028723 A CN 116028723A
- Authority
- CN
- China
- Prior art keywords
- data
- user
- real
- time
- interest
- 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.)
- Pending
Links
Images
Abstract
The embodiment of the invention relates to the technical field of computer data processing and discloses a data recommendation method, which comprises the following steps: respectively determining real-time behavior data and historical behavior data corresponding to each user cluster; respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data; predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data; and respectively determining target recommended contents corresponding to the users according to the real-time interest data and the predicted interest data. By the mode, the data recommendation efficiency and accuracy are improved.
Description
Technical Field
The embodiment of the invention relates to the technical field of computer data processing, in particular to a data recommendation method, a device, equipment and a computer storage medium.
Background
At present, a method for analyzing massive historical user data or short-time streaming data to obtain a recommendation result is generally adopted when the recommendation data are determined.
The inventor of the application finds that in the process of implementing the invention, the current data recommendation method has the problems of low efficiency or low accuracy.
Disclosure of Invention
In view of the above problems, embodiments of the present invention provide a data recommendation method, apparatus, device, and computer storage medium, which are used to solve the problem in the prior art that the efficiency or accuracy of data recommendation is low.
According to an aspect of an embodiment of the present invention, there is provided a data recommendation method, including:
respectively determining real-time behavior data and historical behavior data corresponding to each user cluster;
respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data;
predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data;
and respectively determining target recommended contents corresponding to the users according to the real-time interest data and the predicted interest data.
In an alternative manner, the user cluster includes a plurality of users; the real-time behavior data comprise behavior data of each user in the user cluster aiming at a plurality of contents to be recommended in a first time interval; the first time interval represents a time interval with a first preset duration from the current time; one of the user clusters corresponds to one edge server; the method further comprises the steps of:
Analyzing the real-time behavior data through the edge server to obtain the real-time total interestingness of the content to be recommended in each user cluster; the real-time total interestingness is the sum of interestingness of all users in the user cluster to the content to be recommended;
and sequencing the content to be recommended according to the real-time total interest degree through the edge server to obtain the real-time interest data.
In an alternative manner, the historical behavior data includes behavior data of each of the users for the plurality of content to be recommended in a second time interval; the second time interval represents a time interval with a second preset duration from the current time; the second preset time period is longer than the first preset time period; each edge server is respectively connected with a central server; the method further comprises the steps of:
and collaborative filtering is carried out on all the historical behavior data through the central server to obtain the predicted interest data.
In an alternative, the method further comprises: splitting the predicted interest data through the central server to obtain predicted interest data corresponding to each user in the user cluster;
The predicted interest data are respectively sent to the edge servers corresponding to the user clusters through the central server;
and analyzing the predicted interest data and the real-time interest data corresponding to each user in the user cluster through the edge server to determine the target recommended content.
In an alternative, the method further comprises: weighting the predicted interest data according to the real-time interest data to obtain predicted user interest degrees corresponding to the contents to be recommended;
and determining the target recommended content from the content to be recommended according to the predicted user interest level.
In an alternative, the method further comprises: respectively carrying out noise adding compression processing on the predicted interest data corresponding to each user cluster to obtain compressed predicted interest data corresponding to each user cluster;
according to the real-time total interestingness of all the to-be-recommended contents in each user cluster, determining the interest weight of each to-be-recommended content in each user cluster; the real-time total interestingness is the sum of interestingness of all users in the user cluster to the content to be recommended;
Respectively weighting the compressed predicted interest data corresponding to each user cluster according to the interest weight to obtain encoded interest predicted data;
and decoding the encoded interest prediction data to obtain the predicted user interest degree of each user aiming at each content to be recommended in each user cluster.
In an alternative, the method further comprises: determining the maximum value in the real-time total interestingness corresponding to each user cluster;
normalizing the real-time total interestingness corresponding to all the contents to be recommended respectively according to the maximum value to obtain normalized interestingness corresponding to each content to be recommended respectively;
and determining the interest weight according to the normalized interest degree.
According to another aspect of the embodiment of the present invention, there is provided a data recommendation apparatus including:
the first determining module is used for respectively determining real-time behavior data and historical behavior data corresponding to each user cluster;
the second determining module is used for respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data;
the prediction module is used for predicting the predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data;
And the third determining module is used for respectively determining the target recommended content corresponding to each user according to the real-time interest data and the predicted interest data.
According to another aspect of the embodiment of the present invention, there is provided a data recommendation apparatus including: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
the memory is configured to store at least one executable instruction that causes the processor to perform operations of the data recommendation method.
According to still another aspect of the embodiments of the present invention, there is provided a computer-readable storage medium having stored therein at least one executable instruction for causing a data recommendation device to perform the operations of the data recommendation method as described below.
According to the embodiment of the invention, the real-time behavior data and the historical behavior data corresponding to each user cluster are respectively determined; respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data, wherein the real-time interest data are used for representing the real-time total interest degree of all users in each user cluster for each content to be recommended; predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data; and finally, respectively determining target recommended contents corresponding to each user according to the real-time interest data and the predicted interest data, so that the accuracy and the efficiency of data recommendation are improved by combining the real-time interest data with higher calculation efficiency and timeliness on the basis of the predicted interest data.
Different from the problems of low efficiency of determining recommended content according to historical behavior data or low accuracy of recommending according to real-time behavior data in the prior art, the embodiment of the invention divides a large number of users into a plurality of user clusters, and determines predicted interest data of each user according to the historical behavior data corresponding to all the user clusters; and then determining the real-time interest data of the users in the clusters aiming at each user cluster, and finally fusing the real-time interest data containing timeliness into predicted interest data predicted according to mass data, thereby improving the accuracy of data recommendation, and simultaneously improving the efficiency of data recommendation due to higher data processing efficiency of the user clusters by utilizing an edge computing environment.
The foregoing description is only an overview of the technical solutions of the embodiments of the present invention, and may be implemented according to the content of the specification, so that the technical means of the embodiments of the present invention can be more clearly understood, and the following specific embodiments of the present invention are given for clarity and understanding.
Drawings
The drawings are only for purposes of illustrating embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to designate like parts throughout the figures. In the drawings:
Fig. 1 shows a flow chart of a data recommendation method according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a data recommendation device according to an embodiment of the present invention;
fig. 3 shows a schematic structural diagram of a data recommendation device according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be limited to the embodiments set forth herein.
Before proceeding with the description of the embodiments of the present invention, related terms will be described:
edge calculation: the method is characterized in that an open platform integrating network, computing, storage and application core capabilities is adopted at one side close to an object or data source, so that nearest service is provided.
Edge device: the device which is fixed in life and work scenes and can acquire the user side data nearby for calculation has certain storage and caching capabilities. Each individual edge device also has a separate ID stored in the cloud.
DAE model: denoising Autoencoder, a noise reducing auto encoder, is used to add noise to a clean input signal to produce a corrupted signal, which is then fed into a conventional auto encoder to be reconstructed back into the original lossless signal. Wherein the automatic encoder is based on the fact that: the original input (set as x) is weighted (set as a weight parameter W, b), mapped to obtain y, and then reversely weighted and mapped to obtain z. By iteratively training the two sets (W, b) such that the error function is minimized, i.e. as close as possible z to x is guaranteed, i.e. x is perfectly reconstructed, the forward first set of weight parameters (W, b) can be said to be successful, and key features in the input data are well learned. The true concern of the self-encoder is the feature expression of the hidden layer, a good expression can capture the stable structure of the input signal, and the noise-reducing automatic encoder appears based on the purpose.
Fig. 1 shows a flowchart of a data recommendation method provided by an embodiment of the present invention, which is performed by a computer processing device. The computer processing device may include a cell phone, a notebook computer, and the like. As shown in fig. 1, the method comprises the steps of:
step 10: and respectively determining real-time behavior data and historical behavior data corresponding to each user cluster.
In one embodiment of the present invention, the user cluster includes a plurality of users, and the user cluster may be obtained by dividing the user cluster according to one or more specific attributes of the users. The specific attribute may be a user attribute that affects the user's interest in the content to be recommended, such as a geographic location, etc.
In one embodiment of the present invention, the real-time behavior data includes behavior data of each user in the user cluster for a plurality of content to be recommended in a first time interval; the first time interval represents a time interval with a first preset duration from the current time. Specifically, the first time interval may be shorter than 10 minutes, half hours, etc., so as to ensure that the timeliness of the user behavior represented by the real-time behavior data is stronger.
In one embodiment of the present invention, the content to be recommended may include multimedia content such as news, video, etc. Wherein the multimedia content may be distributed and disseminated in regions by geographic location. The user's behavior on the content to be recommended may include clicking, browsing, commenting, collecting, forwarding, etc.
In order to reduce information transmission cost and ensure data response time in edge calculation, the edge cluster is generally constructed according to region division and basically only processes the business in the region, so in a further embodiment of the invention, when the user cluster is divided according to the attribute related to the geographic position of the user, the real-time behavior data of the user cluster in the region where the server is located can be correspondingly obtained through the edge server, thereby meeting the requirements on the geographic position in the data recommendation related to the geographic position such as news and the like and improving the efficiency of data processing. One of the user clusters corresponds to one of the edge servers, wherein one of the edge servers is used for one of the edge clusters, the edge cluster comprises a plurality of edge terminals, and one of the edge terminals can be a terminal used by a user, such as a mobile phone, a router and the like.
In one embodiment of the invention, the historical behavior data includes behavior data of each user for the plurality of content to be recommended during a second time interval. The second time interval represents a time interval with a second preset duration from the current time, and the second preset duration is longer than the first preset duration. Optionally, the second preset duration may be 24 hours, three days, or half a year, so as to ensure that the reliability of the user interest feature characterized by the historical behavior data is high.
In still another embodiment of the present invention, the historical behavior data may be obtained by fusing real-time behavior data of each user cluster in a third preset duration collected at regular time by the edge server, specifically, the data reported by the edge server may be iteratively fused by the center server to obtain and store the historical behavior data. The third preset time period can be dynamically adjusted according to the type and the change of the content to be recommended, and can be 24 hours or three days. The central server is connected with each edge server respectively.
In yet another embodiment of the present invention, after the collected real-time behavior data is periodically reported to the central server, the data is emptied at each edge server, thereby reducing the pressure of the data storage of the edge servers.
Step 20: and respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data.
In one embodiment of the invention, for each content to be recommended, the real-time total interest degree of all users in the user cluster for the content to be recommended is determined according to the real-time behavior data. And then screening according to the real-time total interestingness of all the to-be-recommended contents to obtain a preset number of to-be-recommended contents with the highest real-time total interestingness corresponding to each user cluster as the real-time interest data of the user cluster. The real-time total interestingness can be determined according to the occurrence number of the behaviors of the content to be recommended under one or more behaviors, for example, the total click quantity or the total browsing quantity is determined to be the real-time total interestingness, or the occurrence numbers of different behaviors such as the total click quantity and the total browsing quantity are weighted and summed to obtain the real-time total interestingness.
In the further embodiment of the invention, the real-time behavior data of each user cluster is processed through the corresponding edge server of the user cluster to obtain real-time interest data, so that the information error caused by network factors is effectively reduced, the loss caused by false alarms is reduced, and the data processing efficiency is improved.
Thus, in yet another embodiment of the present invention, step 20 further comprises: step 201: and analyzing the real-time behavior data through the edge server to obtain the real-time total interestingness of the content to be recommended in each user cluster.
In one embodiment of the present invention, the real-time total interest level is the sum of the interest levels of all users in the user cluster in the content to be recommended. The interest level may be determined according to the click rate, the browse rate, and the like.
For example, the content to be recommended is news I 1 User cluster U 1 In the presence of n 1 Individual users, n 1 Individual user for news I 1 The total click rate is r 1 News I 1 Is r 1 。
Step 202: and sequencing the content to be recommended according to the real-time total interest degree through the edge server to obtain the real-time interest data.
In one embodiment of the present invention, the ranking may take the form of Top-N ranking, and the real-time interest data may be represented as a content-real-time total interest one-dimensional matrix: r= [ r ] 1 r 2 … r i ]。
Wherein r is i The real-time total interestingness of the content i to be recommended is obtained. It should be noted that, because the Top-N ordering mode is adopted, the real-time total interests of the other contents to be recommended are recorded as 0 in the one-dimensional matrix of the content-real-time total interests, except for the Top-N-bit contents to be recommended.
Step 30: predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data.
In one embodiment of the invention, the predicted interest data characterizes the predicted interest level of each user for each content to be recommended. For example, there are n+1 user clusters U 1- U n I+1 content I to be recommended 1- I n Wherein user u 1 Is U (U) 1- U n Any user in the list, all users u are included in the predicted interest data 1 For I 1- I n Is a predicted interestingness of (1).
In still another embodiment of the present invention, the historical behavior data may be predicted by a pre-trained CNN (Convolutional Neural Network ) model, to obtain predicted interest data corresponding to each user.
Considering that the historical behavior data is obtained on the basis of the historical behavior data in the previous statistical period and is obtained after continuous iterative updating, the data volume of the historical behavior data is large, and the analysis task volume is relatively large and centralized, in order to improve the efficiency of the historical behavior data processing, in still another embodiment of the invention, the historical behavior data after each updating can be analyzed and predicted through a central server which is respectively connected with each edge server, so as to obtain the predicted interest data.
In yet another embodiment of the present invention, collaborative filtering may be accomplished by pre-training a finished CNN (Convolutional Neural Network ) model. The predicted interest data may be represented in the form of a user-content prediction matrix in dimension u x i, where u is the total number of users in all user clusters, i is the total number of content to be recommended, and one element r in the user-content prediction matrix ab Representing the predicted interest level of the user a in the content b to be recommended.
Step 40: and respectively determining target recommended contents corresponding to the users according to the real-time interest data and the predicted interest data.
In one embodiment of the invention, the real-time interest data characterizes the real-time total interest degree of each content to be recommended in each user cluster, and the predicted interest data characterizes the predicted interest degree of each user for the content to be recommended, so that the adjustment can be performed according to the real-time interest data on the basis of the predicted interest degree, thereby adding the aging characteristic of the user interest reflected in the real-time interest data into the predicted interest data, and finally screening according to the predicted interest data added with the aging characteristic to obtain the target recommended content.
Thus, in yet another embodiment of the present invention, step 40 further comprises:
step 401: and splitting the predicted interest data through the central server to obtain predicted interest data corresponding to each user in the user cluster.
In one embodiment of the invention, the central server is used for matching the user information contained in each predicted interest data with the user information contained in each user cluster to obtain the predicted interest data corresponding to each user in each user cluster.
In yet another embodiment of the present invention, the predicted interest data may be represented as a user-content prediction matrix in u x i dimensions, where u is the number of users in the user cluster, i is the number of content to be recommended, and the user content matrixElement r in (a) ui The predicted interest degree of the user u for the content i to be recommended is obtained.
Step 402: and respectively sending the predicted interest data to the edge servers corresponding to the user clusters through the central server.
In one embodiment of the invention, the split predicted interest data is distributed to the edge servers of the user clusters through the central server, so that the edge servers only need to store and process the predicted interest data of the user clusters in the area where the edge servers are located, and the efficiency of user interest prediction is improved through edge calculation.
Step 403: and analyzing the predicted interest data and the real-time interest data corresponding to each user in the user cluster through the edge server to determine the target recommended content.
In one embodiment of the invention, the interest prediction information of each content to be recommended, which is included in the received predicted interest data, is weighted by each edge server according to the real-time interest data corresponding to the user in the area where the edge server is located, so as to obtain the prediction of each user on each content to be recommended.
Thus, in one embodiment of the present invention, step 40 further comprises: step 404: and weighting the predicted interest data according to the real-time interest data to obtain the predicted user interest degree corresponding to each content to be recommended.
In one embodiment of the invention, the real-time interest data and the predicted interest data can be fused through a DAE (Denoising Autoencoder) model, and the weight matrix of the coding layer is adjusted according to the real-time interest data in the fusion process, so that the weighting processing of the predicted interest data is realized through the coding layer in the DAE model.
Thus, in yet another embodiment of the present invention, step 404 further comprises: step 4041: and respectively carrying out noise adding compression processing on the predicted interest data corresponding to each user cluster to obtain compressed predicted interest data corresponding to each user cluster.
In one embodiment of the present invention, the predicted interest data corresponding to a user cluster is a matrix R ui At R ui Adding Gaussian noise in proportion q, i.e. let P (r ui =0) =q, and R will be ui The other uncorrupted elements in the matrix are amplified by 1/(1-q) times to obtain a matrix R 'after noise addition' ui Wherein r is ui Is R ui Any element in the above list.
R 'is then added' ui Compressing according to the following formula to obtain the predicted interest data h1 after compression from the dimension reduction of u×i to the dimension reduction of k×i, wherein the predicted interest data h1 after compression is obtained as follows:
wherein h1 is the predicted interest data after compression,
to activate the function +.>
Specifically, the sigmoid function, b is a bias matrix, and W is a weight matrix of the size of u×k.
Step 4042: and respectively determining the interest weights of the contents to be recommended in the user clusters according to the real-time total interests of the contents to be recommended in the user clusters.
In one embodiment of the present invention, the real-time total interest level is the sum of the interest levels of all users in the user cluster in the content to be recommended. After the real-time total interestingness of each content to be recommended is determined, carrying out normalization processing on the real-time total interestingness of all the content to be recommended in one user cluster to obtain corresponding weighting weights so as to carry out weighting processing, thereby improving the accuracy of the weighting processing.
Thus, in one embodiment of the invention, step 4042 further comprises:
step 421: and determining the maximum value in the real-time total interestingness corresponding to each user cluster.
Step 422: and carrying out normalization processing on the real-time total interestingness corresponding to all the contents to be recommended respectively according to the maximum value to obtain normalized interestingness corresponding to each content to be recommended respectively.
In one embodiment of the present invention, the normalization processing may be dividing the real-time total interestingness of each content to be recommended by a maximum value to obtain a corresponding normalized interestingness.
Step 423: and determining the interest weight according to the normalized interest degree.
In one embodiment of the invention, the normalized interestingness is used as the interestingness weight of each content to be recommended, and an interestingness weight matrix r' with the element value range of [0,1] is obtained.
Step 4043: and respectively weighting the compressed predicted interest data corresponding to each user cluster according to the interest weight to obtain encoded interest predicted data.
In one embodiment of the present invention, the compressed prediction interest data is weighted according to the following formula to obtain encoded prediction interest data h2.
Wherein r 'is' ii Is a matrix with the scale of i multiplied by i, r' ii Each diagonal element r' ii ＝1+r i ' the remainder being 0, wherein r i 'is the i-th element in the interest weight matrix r'.
Step 4044: and decoding the encoded interest prediction data to obtain the predicted user interest degree of each user aiming at each content to be recommended in each user cluster.
In one embodiment of the present invention, the decoding is performed according to a decoding layer in the DAE model to obtain predicted user interests of each content to be recommended under each user in each user cluster.
Step 405: and determining the target recommended content from the content to be recommended according to the predicted user interest level.
In one embodiment of the invention, the content to be recommended can be screened and ranked according to the predicted user interestingness of each user, and the top N pieces of content to be recommended can be recommended to each user.
In yet another embodiment of the present invention, steps 404-405 may be performed by the edge servers corresponding to the respective user clusters, as in step 403, so as to improve the efficiency of data processing.
The data recommendation method provided by the embodiment of the invention respectively determines the real-time behavior data and the historical behavior data corresponding to each user cluster; respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data, wherein the real-time interest data are used for representing the real-time total interest degree of all users in each user cluster for each content to be recommended; predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data; and finally, respectively determining target recommended contents corresponding to each user according to the real-time interest data and the predicted interest data, so that the accuracy and the efficiency of data recommendation are improved by combining the real-time interest data with higher calculation efficiency and timeliness on the basis of the predicted interest data.
Different from the problems of low efficiency of determining recommended content according to historical behavior data or low accuracy of recommending according to real-time behavior data in the prior art, the data recommendation method provided by the embodiment of the invention divides a large number of users into a plurality of user clusters, and determines predicted interest data of each user according to the historical behavior data corresponding to all the user clusters; and then determining the real-time interest data of the users in the clusters aiming at each user cluster, and finally fusing the real-time interest data containing timeliness into predicted interest data predicted according to mass data, thereby improving the accuracy of data recommendation, and simultaneously improving the efficiency of data recommendation due to higher data processing efficiency of the user clusters by utilizing an edge computing environment.
Fig. 2 shows a schematic structural diagram of a data recommendation device according to an embodiment of the present invention. As shown in fig. 2, the apparatus 50 includes: a first determination module 501, a second determination module 502, a prediction module 503, and a third determination module 504.
In an optional manner, the first determining module 501 is configured to determine real-time behavior data and historical behavior data corresponding to each user cluster respectively;
A second determining module 502, configured to determine real-time interest data corresponding to each user cluster according to the real-time behavior data;
a prediction module 503, configured to predict predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data;
and a third determining module 504, configured to determine target recommended content corresponding to each user according to the real-time interest data and the predicted interest data.
In yet another embodiment of the present invention, said user cluster comprises a plurality of said users; the real-time behavior data comprise behavior data of each user in the user cluster aiming at a plurality of contents to be recommended in a first time interval; the first time interval represents a time interval with a first preset duration from the current time; one of the user clusters corresponds to one edge server; the second determining module 502 is further configured to: analyzing the real-time behavior data through the edge server to obtain the real-time total interestingness of the content to be recommended in each user cluster; the real-time total interestingness is the sum of interestingness of all users in the user cluster to the content to be recommended;
And sequencing the content to be recommended according to the real-time total interest degree through the edge server to obtain the real-time interest data.
In yet another embodiment of the present invention, the historical behavior data includes behavior data of each of the users for the plurality of content to be recommended during a second time interval; the second time interval represents a time interval with a second preset duration from the current time; the second preset time period is longer than the first preset time period; each edge server is respectively connected with a central server; the prediction module 503 is further configured to:
and collaborative filtering is carried out on all the historical behavior data through the central server to obtain the predicted interest data.
In yet another embodiment of the present invention, the third determining module 504 is further configured to:
splitting the predicted interest data through the central server to obtain predicted interest data corresponding to each user in the user cluster;
the predicted interest data are respectively sent to the edge servers corresponding to the user clusters through the central server;
and analyzing the predicted interest data and the real-time interest data corresponding to each user in the user cluster through the edge server to determine the target recommended content.
In yet another embodiment of the present invention, the third determining module 504 is further configured to:
weighting the predicted interest data according to the real-time interest data to obtain predicted user interest degrees corresponding to the contents to be recommended;
and determining the target recommended content from the content to be recommended according to the predicted user interest level.
In yet another embodiment of the present invention, the third determining module 504 is further configured to:
respectively carrying out noise adding compression processing on the predicted interest data corresponding to each user cluster to obtain compressed predicted interest data corresponding to each user cluster;
according to the real-time total interestingness of all the to-be-recommended contents in each user cluster, determining the interest weight of each to-be-recommended content in each user cluster; the real-time total interestingness is the sum of interestingness of all users in the user cluster to the content to be recommended;
respectively weighting the compressed predicted interest data corresponding to each user cluster according to the interest weight to obtain encoded interest predicted data;
and decoding the encoded interest prediction data to obtain the predicted user interest degree of each user aiming at each content to be recommended in each user cluster.
In yet another embodiment of the present invention, the third determining module 504 is further configured to:
determining the maximum value in the real-time total interestingness corresponding to each user cluster;
normalizing the real-time total interestingness corresponding to all the contents to be recommended respectively according to the maximum value to obtain normalized interestingness corresponding to each content to be recommended respectively;
and determining the interest weight according to the normalized interest degree.
The data recommendation device provided by the embodiment of the invention respectively determines the real-time behavior data and the historical behavior data corresponding to each user cluster; respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data, wherein the real-time interest data are used for representing the real-time total interest degree of all users in each user cluster for each content to be recommended; predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data; and finally, respectively determining target recommended contents corresponding to each user according to the real-time interest data and the predicted interest data, so that the accuracy and the efficiency of data recommendation are improved by combining the real-time interest data with higher calculation efficiency and timeliness on the basis of the predicted interest data.
Different from the problem that in the prior art, the efficiency of determining recommended content according to historical behavior data is low or the accuracy of recommending according to real-time behavior data is low, the data recommending device provided by the embodiment of the invention divides a large number of users into a plurality of user clusters, and determines predicted interest data of each user according to the historical behavior data corresponding to all the user clusters; and then determining the real-time interest data of the users in the clusters aiming at each user cluster, and finally fusing the real-time interest data containing timeliness into predicted interest data predicted according to mass data, thereby improving the accuracy of data recommendation, and simultaneously improving the efficiency of data recommendation due to higher data processing efficiency of the user clusters by utilizing an edge computing environment.
Fig. 3 shows a schematic structural diagram of a data recommendation device according to an embodiment of the present invention, and the specific embodiment of the present invention is not limited to a specific implementation of the data recommendation device.
As shown in fig. 3, the data recommendation device may include: a processor 602, a communication interface (Communications Interface), a memory 606, and a communication bus 608.
Wherein: processor 602, communication interface 604, and memory 606 perform communication with each other via communication bus 608. Communication interface 604 is used to communicate with network elements of other devices, such as clients or other servers. The processor 602 is configured to execute the program 610, and may specifically perform the relevant steps in the embodiment of the data recommendation method described above.
In particular, program 610 may include program code comprising computer-executable instructions.
The processor 602 may be a central processing unit CPU or a specific integrated circuit ASIC (Application Specific Integrated Circuit) or one or more integrated circuits configured to implement embodiments of the present invention. The one or more processors included in the data recommendation device may be the same type of processor, such as one or more CPUs; but may also be different types of processors such as one or more CPUs and one or more ASICs.
A memory 606 for storing a program 610. The memory 606 may comprise high-speed RAM memory or may further comprise non-volatile memory (non-volatile memory), such as at least one disk memory.
The program 610 may be specifically invoked by the processor 602 to cause the data recommendation device to:
Respectively determining real-time behavior data and historical behavior data corresponding to each user cluster;
respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data;
predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data;
and respectively determining target recommended contents corresponding to the users according to the real-time interest data and the predicted interest data.
In an alternative manner, the user cluster includes a plurality of users; the real-time behavior data comprise behavior data of each user in the user cluster aiming at a plurality of contents to be recommended in a first time interval; the first time interval represents a time interval with a first preset duration from the current time; one of the user clusters corresponds to one edge server; the program 610 is invoked by the processor 602 to cause the data recommendation device to:
analyzing the real-time behavior data through the edge server to obtain the real-time total interestingness of the content to be recommended in each user cluster; the real-time total interestingness is the sum of interestingness of all users in the user cluster to the content to be recommended;
And sequencing the content to be recommended according to the real-time total interest degree through the edge server to obtain the real-time interest data.
In an alternative manner, the historical behavior data includes behavior data of each of the users for the plurality of content to be recommended in a second time interval; the second time interval represents a time interval with a second preset duration from the current time; the second preset time period is longer than the first preset time period; each edge server is respectively connected with a central server; the program 610 is invoked by the processor 602 to cause the data recommendation device to:
and collaborative filtering is carried out on all the historical behavior data through the central server to obtain the predicted interest data.
In an alternative, the program 610 is invoked by the processor 602 to cause the data recommendation device to:
splitting the predicted interest data through the central server to obtain predicted interest data corresponding to each user in the user cluster;
the predicted interest data are respectively sent to the edge servers corresponding to the user clusters through the central server;
And analyzing the predicted interest data and the real-time interest data corresponding to each user in the user cluster through the edge server to determine the target recommended content.
In an alternative, the program 610 is invoked by the processor 602 to cause the data recommendation device to:
weighting the predicted interest data according to the real-time interest data to obtain predicted user interest degrees corresponding to the contents to be recommended;
and determining the target recommended content from the content to be recommended according to the predicted user interest level.
In an alternative, the program 610 is invoked by the processor 602 to cause the data recommendation device to:
respectively carrying out noise adding compression processing on the predicted interest data corresponding to each user cluster to obtain compressed predicted interest data corresponding to each user cluster;
according to the real-time total interestingness of all the to-be-recommended contents in each user cluster, determining the interest weight of each to-be-recommended content in each user cluster; the real-time total interestingness is the sum of interestingness of all users in the user cluster to the content to be recommended;
Respectively weighting the compressed predicted interest data corresponding to each user cluster according to the interest weight to obtain encoded interest predicted data;
and decoding the encoded interest prediction data to obtain the predicted user interest degree of each user aiming at each content to be recommended in each user cluster.
In an alternative, the program 610 is invoked by the processor 602 to cause the data recommendation device to:
determining the maximum value in the real-time total interestingness corresponding to each user cluster;
normalizing the real-time total interestingness corresponding to all the contents to be recommended respectively according to the maximum value to obtain normalized interestingness corresponding to each content to be recommended respectively;
and determining the interest weight according to the normalized interest degree.
The data recommendation equipment provided by the embodiment of the invention respectively determines the real-time behavior data and the historical behavior data corresponding to each user cluster; respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data, wherein the real-time interest data are used for representing the real-time total interest degree of all users in each user cluster for each content to be recommended; predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data; and finally, respectively determining target recommended contents corresponding to each user according to the real-time interest data and the predicted interest data, so that the accuracy and the efficiency of data recommendation are improved by combining the real-time interest data with higher calculation efficiency and timeliness on the basis of the predicted interest data.
Different from the problem that in the prior art, the efficiency of determining recommended content according to historical behavior data is low or the accuracy of recommending according to real-time behavior data is low, the data recommending device provided by the embodiment of the invention divides a large number of users into a plurality of user clusters, and determines predicted interest data of each user according to the historical behavior data corresponding to all the user clusters; and then determining the real-time interest data of the users in the clusters aiming at each user cluster, and finally fusing the real-time interest data containing timeliness into predicted interest data predicted according to mass data, thereby improving the accuracy of data recommendation, and simultaneously improving the efficiency of data recommendation due to higher data processing efficiency of the user clusters by utilizing an edge computing environment.
The embodiment of the invention provides a computer readable storage medium, which stores at least one executable instruction, and the executable instruction enables a data recommendation device to execute the data recommendation method in any of the method embodiments.
The executable instructions may be specifically operable to cause a data recommendation device to:
Respectively determining real-time behavior data and historical behavior data corresponding to each user cluster;
respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data;
predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data;
and respectively determining target recommended contents corresponding to the users according to the real-time interest data and the predicted interest data.
In an alternative manner, the user cluster includes a plurality of users; the real-time behavior data comprise behavior data of each user in the user cluster aiming at a plurality of contents to be recommended in a first time interval; the first time interval represents a time interval with a first preset duration from the current time; one of the user clusters corresponds to one edge server; the executable instructions cause the data recommendation device to:
analyzing the real-time behavior data through the edge server to obtain the real-time total interestingness of the content to be recommended in each user cluster; the real-time total interestingness is the sum of interestingness of all users in the user cluster to the content to be recommended;
And sequencing the content to be recommended according to the real-time total interest degree through the edge server to obtain the real-time interest data.
In an alternative manner, the historical behavior data includes behavior data of each of the users for the plurality of content to be recommended in a second time interval; the second time interval represents a time interval with a second preset duration from the current time; the second preset time period is longer than the first preset time period; each edge server is respectively connected with a central server; the executable instructions cause the data recommendation device to:
and collaborative filtering is carried out on all the historical behavior data through the central server to obtain the predicted interest data.
In an alternative form, the executable instructions cause the data recommendation device to:
splitting the predicted interest data through the central server to obtain predicted interest data corresponding to each user in the user cluster;
the predicted interest data are respectively sent to the edge servers corresponding to the user clusters through the central server;
And analyzing the predicted interest data and the real-time interest data corresponding to each user in the user cluster through the edge server to determine the target recommended content.
In an alternative form, the executable instructions cause the data recommendation device to:
weighting the predicted interest data according to the real-time interest data to obtain predicted user interest degrees corresponding to the contents to be recommended;
and determining the target recommended content from the content to be recommended according to the predicted user interest level.
In an alternative form, the executable instructions cause the data recommendation device to:
respectively carrying out noise adding compression processing on the predicted interest data corresponding to each user cluster to obtain compressed predicted interest data corresponding to each user cluster;
according to the real-time total interestingness of all the to-be-recommended contents in each user cluster, determining the interest weight of each to-be-recommended content in each user cluster; the real-time total interestingness is the sum of interestingness of all users in the user cluster to the content to be recommended;
Respectively weighting the compressed predicted interest data corresponding to each user cluster according to the interest weight to obtain encoded interest predicted data;
and decoding the encoded interest prediction data to obtain the predicted user interest degree of each user aiming at each content to be recommended in each user cluster.
In an alternative form, the executable instructions cause the data recommendation device to:
determining the maximum value in the real-time total interestingness corresponding to each user cluster;
normalizing the real-time total interestingness corresponding to all the contents to be recommended respectively according to the maximum value to obtain normalized interestingness corresponding to each content to be recommended respectively;
and determining the interest weight according to the normalized interest degree.
The computer storage medium of the embodiment of the invention respectively determines the real-time behavior data and the historical behavior data corresponding to each user cluster; respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data, wherein the real-time interest data are used for representing the real-time total interest degree of all users in each user cluster for each content to be recommended; predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data; and finally, respectively determining target recommended contents corresponding to each user according to the real-time interest data and the predicted interest data, so that the accuracy and the efficiency of data recommendation are improved by combining the real-time interest data with higher calculation efficiency and timeliness on the basis of the predicted interest data.
In order to solve the problems that in the prior art, the efficiency of determining recommended content according to historical behavior data is low or the accuracy of recommending according to real-time behavior data is low, the computer storage medium of the embodiment of the invention divides a large number of users into a plurality of user clusters, and determines predicted interest data of each user according to the historical behavior data corresponding to all the user clusters; and then determining the real-time interest data of the users in the clusters aiming at each user cluster, and finally fusing the real-time interest data containing timeliness into predicted interest data predicted according to mass data, thereby improving the accuracy of data recommendation, and simultaneously improving the efficiency of data recommendation due to higher data processing efficiency of the user clusters by utilizing an edge computing environment.
The embodiment of the invention provides a data recommending device which is used for executing the data recommending method.
An embodiment of the present invention provides a computer program that can be invoked by a processor to cause a data recommendation device to perform the data recommendation method in any of the method embodiments described above.
An embodiment of the present invention provides a computer program product comprising a computer program stored on a computer readable storage medium, the computer program comprising program instructions which, when run on a computer, cause the computer to perform the data recommendation method of any of the method embodiments described above.
The algorithms or displays presented herein are not inherently related to any particular computer, virtual system, or other apparatus. Various general-purpose systems may also be used with the teachings herein. The required structure for a construction of such a system is apparent from the description above. In addition, embodiments of the present invention are not directed to any particular programming language. It will be appreciated that the teachings of the present invention described herein may be implemented in a variety of programming languages, and the above description of specific languages is provided for disclosure of enablement and best mode of the present invention.
In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.
Similarly, it should be appreciated that in the above description of exemplary embodiments of the invention, various features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be construed as reflecting the intention that: i.e., the claimed invention requires more features than are expressly recited in each claim.
Those skilled in the art will appreciate that the modules in the apparatus of the embodiments may be adaptively changed and disposed in one or more apparatuses different from the embodiments. The modules or units or components of the embodiments may be combined into one module or unit or component, and they may be divided into a plurality of sub-modules or sub-units or sub-components. Any combination of all features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or units of any method or apparatus so disclosed, may be used in combination, except insofar as at least some of such features and/or processes or units are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, third, etc. do not denote any order. These words may be interpreted as names. The steps in the above embodiments should not be construed as limiting the order of execution unless specifically stated.
Claims (10)
1. A data recommendation method, the method comprising:
respectively determining real-time behavior data and historical behavior data corresponding to each user cluster;
respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data;
predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data;
and respectively determining target recommended contents corresponding to the users according to the real-time interest data and the predicted interest data.
2. The method according to claim 1, wherein a plurality of said users are included in said user cluster; the real-time behavior data comprise behavior data of each user in the user cluster aiming at a plurality of contents to be recommended in a first time interval; the first time interval represents a time interval with a first preset duration from the current time; one of the user clusters corresponds to one edge server; the step of respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data comprises the following steps:
analyzing the real-time behavior data through the edge server to obtain the real-time total interestingness of the content to be recommended in each user cluster; the real-time total interestingness is the sum of interestingness of all users in the user cluster to the content to be recommended;
And sequencing the content to be recommended according to the real-time total interest degree through the edge server to obtain the real-time interest data.
3. The method of claim 2, wherein the historical behavior data comprises behavior data for the plurality of content to be recommended for each of the users during a second time interval; the second time interval represents a time interval with a second preset duration from the current time; the second preset time period is longer than the first preset time period; each edge server is respectively connected with a central server; predicting predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data, including:
and collaborative filtering is carried out on all the historical behavior data through the central server to obtain the predicted interest data.
4. The method of claim 3, wherein the determining the target recommended content corresponding to each user according to the real-time interest data and the predicted interest data includes:
splitting the predicted interest data through the central server to obtain predicted interest data corresponding to each user in the user cluster;
The predicted interest data are respectively sent to the edge servers corresponding to the user clusters through the central server;
and analyzing the predicted interest data and the real-time interest data corresponding to each user in the user cluster through the edge server to determine the target recommended content.
5. The method of claim 1, wherein the determining the target recommended content from the predicted interest data and the real-time interest data comprises:
weighting the predicted interest data according to the real-time interest data to obtain predicted user interest degrees corresponding to the contents to be recommended;
and determining the target recommended content from the content to be recommended according to the predicted user interest level.
6. The method of claim 5, wherein the weighting the predicted interest data according to the real-time interest data to obtain predicted user interests corresponding to the content to be recommended includes:
respectively carrying out noise adding compression processing on the predicted interest data corresponding to each user cluster to obtain compressed predicted interest data corresponding to each user cluster;
According to the real-time total interestingness of all the to-be-recommended contents in each user cluster, determining the interest weight of each to-be-recommended content in each user cluster; the real-time total interestingness is the sum of interestingness of all users in the user cluster to the content to be recommended;
respectively weighting the compressed predicted interest data corresponding to each user cluster according to the interest weight to obtain encoded interest predicted data;
and decoding the encoded interest prediction data to obtain the predicted user interest degree of each user aiming at each content to be recommended in each user cluster.
7. The method of claim 6, wherein the determining the interest weights of the to-be-recommended contents in the user clusters according to the real-time total interests of the to-be-recommended contents in the user clusters, respectively, comprises:
determining the maximum value in the real-time total interestingness corresponding to each user cluster;
normalizing the real-time total interestingness corresponding to all the contents to be recommended respectively according to the maximum value to obtain normalized interestingness corresponding to each content to be recommended respectively;
And determining the interest weight according to the normalized interest degree.
8. A data recommendation device, the device comprising:
the first determining module is used for respectively determining real-time behavior data and historical behavior data corresponding to each user cluster;
the second determining module is used for respectively determining real-time interest data corresponding to each user cluster according to the real-time behavior data;
the prediction module is used for predicting the predicted interest data corresponding to each user in all the user clusters according to all the historical behavior data;
and the third determining module is used for respectively determining the target recommended content corresponding to each user according to the real-time interest data and the predicted interest data.
9. A data recommendation device, characterized by comprising: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
the memory is configured to store at least one executable instruction that causes the processor to perform the operations of the data recommendation method according to any one of claims 1 to 7.
10. A computer readable storage medium having stored therein at least one executable instruction that, when executed on a data recommendation device, causes the data recommendation device to perform the operations of the data recommendation method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111234461.1A CN116028723A (en) | 2021-10-22 | 2021-10-22 | Data recommendation method, device, equipment and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111234461.1A CN116028723A (en) | 2021-10-22 | 2021-10-22 | Data recommendation method, device, equipment and computer storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116028723A true CN116028723A (en) | 2023-04-28 |
Family
ID=86089845
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111234461.1A Pending CN116028723A (en) | 2021-10-22 | 2021-10-22 | Data recommendation method, device, equipment and computer storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116028723A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116781949A (en) * | 2023-07-26 | 2023-09-19 | 新励成教育科技股份有限公司 | Recommendation method, system, equipment and storage medium for talent lecture live broadcast |
-
2021
- 2021-10-22 CN CN202111234461.1A patent/CN116028723A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116781949A (en) * | 2023-07-26 | 2023-09-19 | 新励成教育科技股份有限公司 | Recommendation method, system, equipment and storage medium for talent lecture live broadcast |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107612893B (en) | Short message auditing system and method and short message auditing model building method | |
| CN111026971A (en) | Content pushing method and device and computer storage medium | |
| CN112801132B (en) | Image processing method and device | |
| CN113379301A (en) | Method, device and equipment for classifying users through decision tree model | |
| CN111652661B (en) | Mobile phone client user loss early warning processing method | |
| CN111062431A (en) | Image clustering method, image clustering device, electronic device, and storage medium | |
| CN111797320A (en) | Data processing method, device, equipment and storage medium | |
| CN112464058A (en) | XGboost algorithm-based telecommunication internet fraud identification method | |
| CN111061948B (en) | User tag recommendation method and device, computer equipment and storage medium | |
| CN111179055B (en) | Credit line adjusting method and device and electronic equipment | |
| CN110245684B (en) | Data processing method, electronic device, and medium | |
| US20230004776A1 (en) | Moderator for identifying deficient nodes in federated learning | |
| CN116028723A (en) | Data recommendation method, device, equipment and computer storage medium | |
| CN114417174A (en) | Content recommendation method, device, equipment and computer storage medium | |
| US8713022B2 (en) | Community profiling for social media | |
| CN114064394A (en) | Safety monitoring method and device based on edge calculation and terminal equipment | |
| CN108846055A (en) | Search processing method, device, electronic equipment and readable storage medium storing program for executing | |
| CN114443943A (en) | Information scheduling method, device and equipment and computer readable storage medium | |
| US11087045B1 (en) | Apparatus and method for option data object performance prediction and modeling | |
| CN112861026B (en) | User preference mining method, device, medium and electronic equipment | |
| CN115935274A (en) | Training method, device, equipment and storage medium for reselling behavior recognition model | |
| CN114625612A (en) | User behavior analysis method and service system based on big data office | |
| CN114356712A (en) | Data processing method, device, equipment, readable storage medium and program product | |
| CN113849745A (en) | Object recommendation method, device, equipment and storage medium | |
| CN110705736A (en) | Macroscopic economy prediction method and device, computer equipment and storage medium |
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 |