Big data recommendation method and system applied to short video platform

Abstract

The embodiment of the present application provides a big data recommendation method and system applied to a short video platform, which can accurately capture the user's viewing habits and preferences by collecting and analyzing sample user viewing record data of short video interactive pages. The user preference prediction network is trained and learned using the reference learning data generated by these sample user viewing record data, so that the user preference prediction network can more accurately predict the user's preferences. Based on the prediction results, content that is more in line with their personal tastes and interests can be recommended to users, thereby improving the user's experience and satisfaction. It not only optimizes the recommendation algorithm, but also improves the user stickiness and activity of the short video platform, realizes personalized recommendation services, and significantly improves recommendation efficiency and user satisfaction.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a big data recommendation method and system applied to a short video platform. Background Art

[0002] With the rapid development of the Internet and the widespread popularity of smart phones, short video platforms, as a new way of entertainment and information acquisition, are becoming more and more popular among users. On short video platforms, users can watch various types of short video content and express their preferences and attitudes through interactive methods such as likes, comments, and reposts. However, in the face of massive short video content, how to provide users with accurate and personalized recommendation services has become a problem that short video platforms need to solve urgently.

[0003] Traditional recommendation methods often make recommendations based on users' historical viewing records, likes, comments and other behavioral data, but this method ignores the behavioral connections and feature relationships between users on different short video interactive pages, resulting in the accuracy and personalization of recommendation results to be improved. In order to more accurately capture user preferences and interests, it is necessary to mine deeper information from user behavior big data.

[0004] At present, although there are some recommendation algorithms that can use user behavior data for recommendation, these algorithms often only consider the data of a single video or a single user behavior, and do not fully utilize the feature connections between short video interactive pages and the correlation of user behavior. Therefore, the existing recommendation methods still have certain limitations in terms of accuracy and personalized services. Summary of the invention

[0005] In view of this, the purpose of this application is to provide a big data recommendation method and system applied to a short video platform.

[0006] In conjunction with the first aspect of the present application, a big data recommendation method applied to a short video platform is provided, which is applied to a big data recommendation system applied to a short video platform, and the method includes:

[0007] Collect sample user viewing record data of multiple short video interactive pages from the user behavior big data of the target short video platform;

[0008] Perform feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data;

[0009] Based on the multiple reference learning data, the user preference prediction network is subjected to knowledge learning, and after the user preference prediction is performed on any input user viewing record data based on the target user preference prediction network that has completed the knowledge learning, data recommendation is performed based on the user preference prediction results.

[0010] In a possible implementation manner of the first aspect, performing knowledge learning on the user preference prediction network based on the plurality of reference learning data includes:

[0011] Generate a pair of short video interaction pages corresponding to the reference learning data based on the short video interaction pages corresponding to the two sample user viewing record data in the reference learning data;

[0012] Assigning reference learning data of the same short video interactive page pair to the same reference learning data sequence;

[0013] The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data sequences.

[0014] In a possible implementation of the first aspect, the step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0015] Using a sample user behavior record set as supervision data, and performing feature association between the user viewing preference and the sample user behavior record set to generate first learning data;

[0016] The user preference prediction network includes a deep learning network, and the user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0017] Loading the user viewing preferences in the first learning data into the deep learning network, and generating a first predicted user behavior preference set through the deep learning network;

[0018] The deep learning network is subjected to knowledge learning based on the sample user behavior record set in the first learning data and the first predicted user behavior preference set.

[0019] In a possible implementation of the first aspect, the step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0020] Using the user's viewing preference as supervision data, and performing feature association between the sample user behavior record set and the user's viewing preference to generate second learning data;

[0021] The user preference prediction network includes a deep learning network, and the user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0022] Loading a set of sample user behavior records in the second learning data into the deep learning network, and generating a first predicted user viewing preference through the deep learning network;

[0023] The deep learning network is subjected to knowledge learning according to the user viewing preference in the second learning data and the first predicted user viewing preference.

[0024] In a possible implementation of the first aspect, the step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0025] Using the sample user behavior record set as supervision data, performing feature association between the user interaction pattern features and the sample user behavior record set to generate third learning data;

[0026] The user preference prediction network includes a feature extraction unit and a deep learning network. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0027] Loading the user interaction pattern features in the third learning data into the feature extraction unit, and generating a first interaction feature vector sequence through the feature extraction unit;

[0028] Loading the first interactive feature vector sequence into the deep learning network, and generating a second predicted user behavior preference set through the deep learning network;

[0029] Based on the sample user behavior record set in the third learning data and the second predicted user behavior preference set, knowledge learning is performed on the feature extraction unit and the deep learning network.

[0030] In a possible implementation of the first aspect, the step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0031] Using the user's viewing preference as supervision data, and performing feature association between the user's interaction pattern characteristics and the user's viewing preference to generate fourth learning data;

[0032] The user preference prediction network includes a feature extraction unit and a deep learning network. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0033] Loading the user interaction pattern features in the fourth learning data into the feature extraction unit, and generating a second interaction feature vector sequence through the feature extraction unit;

[0034] Loading the second interactive feature vector sequence into the deep learning network, and generating a second predicted user viewing preference through the deep learning network;

[0035] The feature extraction unit and the deep learning network are subjected to knowledge learning according to the user viewing preference in the fourth learning data and the second predicted user viewing preference.

[0036] In a possible implementation of the first aspect, the step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0037] Using the user interaction pattern feature as supervision data, performing feature association between the user viewing preference and the user interaction pattern feature, and generating fifth learning data;

[0038] The user preference prediction network includes a deep learning network and a multi-layer perceptron. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0039] Loading the user viewing preference in the fifth learning data into the deep learning network, and generating a first interactive behavior feature through the deep learning network;

[0040] Loading the first interactive behavior feature into the multi-layer perceptron, and generating a first predicted user interactive behavior pattern through the multi-layer perceptron;

[0041] Based on the user interaction pattern characteristics in the fifth learning data and the first predicted user interaction behavior pattern, knowledge learning is performed on the deep learning network and the multilayer perceptron.

[0042] In a possible implementation of the first aspect, the step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0043] Using the user interaction pattern feature as supervision data, performing feature association between the sample user behavior record set and the user interaction pattern feature to generate sixth learning data;

[0044] The user preference prediction network includes a deep learning network and a multi-layer perceptron. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0045] Loading a set of sample user behavior records in the sixth learning data into the deep learning network, and generating a second interactive behavior feature through the deep learning network;

[0046] Loading the second interactive behavior feature into the multi-layer perceptron, and generating a second predicted user interactive behavior pattern through the multi-layer perceptron;

[0047] Based on the user interaction pattern characteristics in the sixth learning data and the second predicted user interaction behavior pattern, knowledge learning is performed on the deep learning network and the multilayer perceptron.

[0048] In a possible implementation of the first aspect, the step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0049] Using the user interaction pattern feature as supervision data, performing feature association between the user viewing environment condition and the user interaction pattern feature, and generating seventh learning data;

[0050] The user preference prediction network includes a deep learning network and a conditional diffusion network. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0051] Loading the user viewing environment conditions in the seventh learning data into the deep learning network, performing feature extraction on the user viewing environment conditions in the seventh learning data through the deep learning network, and generating user viewing environment vector data;

[0052] Loading the user viewing environment vector data into the conditional diffusion network, and generating a third predicted user interaction behavior pattern through the conditional diffusion network;

[0053] Performing knowledge learning on the deep learning network and the conditional diffusion network according to the user interaction pattern characteristics in the seventh learning data and the third predicted user interaction behavior pattern;

[0054] Before loading the user viewing environment vector data into the conditional diffusion network, the method further includes:

[0055] Obtain multiple user interaction pattern features;

[0056] The conditional diffusion network is subjected to knowledge learning based on the plurality of user interaction pattern features.

[0057] In combination with the second aspect of the present application, a big data recommendation system applied to a short video platform is provided, wherein the big data recommendation system applied to the short video platform includes a machine-readable storage medium and a processor, wherein the machine-readable storage medium stores machine-executable instructions, and when the processor executes the machine-executable instructions, the big data recommendation system applied to the short video platform implements the aforementioned big data recommendation method applied to the short video platform.

[0058] In conjunction with the third aspect of the present application, a computer-readable storage medium is provided, in which computer-executable instructions are stored. When the computer-executable instructions are executed, the aforementioned big data recommendation method applied to the short video platform is implemented.

[0059] In combination with any of the above aspects, by collecting and analyzing sample user viewing record data of short video interactive pages, users' viewing habits and preferences can be accurately captured. The user preference prediction network is trained and learned using reference learning data generated by these sample user viewing record data, so that the user preference prediction network can more accurately predict user preferences. Based on the prediction results, content that is more in line with their personal tastes and interests can be recommended to users, thereby improving the user experience and satisfaction. This not only optimizes the recommendation algorithm, but also improves the user stickiness and activity of the short video platform, realizes personalized recommendation services, and significantly improves recommendation efficiency and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained by combining these drawings without paying creative work.

[0061] Figure 1 A flowchart of a big data recommendation method applied to a short video platform provided in an embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. According to the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or ends.

[0064] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0065] Figure 1 The flowchart of the big data recommendation method applied to the short video platform provided by the embodiment of the present application is shown. It should be understood that in other embodiments, the order of some steps in the big data recommendation method applied to the short video platform of this embodiment can be shared with each other according to actual needs, or some steps can be omitted or maintained. The details of the big data recommendation method applied to the short video platform include:

[0066] Step S110, collecting sample user viewing record data of multiple short video interactive pages from the user behavior big data of the target short video platform.

[0067] In this embodiment, the server is connected to the user behavior big data storage system of the target short video platform. This user behavior big data storage system records all the user's behaviors on the platform, including watching, liking, commenting, sharing, etc. The server uses a preset data collection program to filter out user viewing record data related to the short video interactive page from this big data storage system. These data include but are not limited to the short video ID watched by the user, viewing time, whether to like, comment, share, etc. The server organizes these data into structured sample user viewing record data and stores them in a local data warehouse for subsequent analysis.

[0068] Step S120: feature-linking the sample user viewing record data of any two short video interactive pages to generate reference learning data.

[0069] The server takes out two sample user viewing record data from the data warehouse. These two sample user viewing record data come from two different short video interactive pages. The server runs a feature connection algorithm to analyze the common features and difference features in these two sample user viewing record data. The common features may be the types of short videos that users like to watch, and the difference features may be the different behavior patterns that users show when watching different short videos. The server connects these features to generate a new data structure. This data structure contains the user behavior connection between the two short video interactive pages. This is the reference learning data. This process will be performed on all sample user viewing record data in the data warehouse to generate a large amount of reference learning data.

[0070] Step S130, based on the multiple reference learning data, the user preference prediction network is subjected to knowledge learning, and after the user preference prediction is performed on any input user viewing record data based on the target user preference prediction network that has completed the knowledge learning, data recommendation is performed based on the user preference prediction result.

[0071] The server first builds a user preference prediction network, which is a deep learning model that can predict user preferences by learning reference learning data. The server inputs the generated reference learning data into the user preference prediction network for knowledge learning. During the learning process, the user preference prediction network will continuously adjust its internal parameters to minimize the prediction error. When the network learning is completed, the server can use this user preference prediction network to predict user preferences for any input user viewing record data.

[0072] Specifically, when a new user's viewing record data is input into the user preference prediction network, the user preference prediction network will predict the type of short videos, viewing time, interactive behavior, etc. that the user may like based on the knowledge it has learned. Based on these prediction results, the server selects short videos that meet the user's preferences from the short video library and then pushes them to the user through the recommendation system. In this way, users can see more content that suits their preferences when watching short videos.

[0073] At the same time, the server will continue to collect user feedback data, such as the click rate, viewing time, interactive behavior, etc. of the user's recommended short videos, and then use these feedback data as new sample user viewing record data to re-contact features and learn knowledge, so as to continuously optimize the prediction accuracy of the user preference prediction network.

[0074] Based on the above steps, by collecting and analyzing the sample user viewing record data of the short video interactive page, the user's viewing habits and preferences can be accurately captured. The user preference prediction network is trained and learned using the reference learning data generated by these sample user viewing record data, so that the user preference prediction network can more accurately predict the user's preferences. Based on the prediction results, content that is more in line with their personal tastes and interests can be recommended to users, thereby improving the user's experience and satisfaction. This not only optimizes the recommendation algorithm, but also improves the user stickiness and activity of the short video platform, realizes personalized recommendation services, and significantly improves recommendation efficiency and user satisfaction.

[0075] In a possible implementation, the performing knowledge learning on the user preference prediction network based on the plurality of reference learning data includes:

[0076] Based on the short video interaction pages corresponding to the two sample user viewing record data in the reference learning data, a pair of short video interaction pages corresponding to the reference learning data is generated.

[0077] The reference learning data of the same short video interactive page pair are assigned to the same reference learning data sequence.

[0078] The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data sequences.

[0079] In this embodiment, the server first extracts the information of the short video interactive page corresponding to the two sample user viewing record data from the generated reference learning data. This information may include the content, style, interactive elements and other aspects of the short video. By comparing and analyzing this information, the server determines which sample user viewing record data appear in pairs, that is, they correspond to different user behaviors of the same short video interactive page. Based on this correspondence, the server organizes these pairs of sample user viewing record data to generate short video interactive page pairs.

[0080] Next, the server starts processing these short video interaction page pairs. For each short video interaction page pair, the server will classify all reference learning data (i.e. paired sample user viewing record data) in the same reference learning data sequence. The purpose of this is to ensure that when learning knowledge, the network can simultaneously consider the behavior patterns of different users under the same short video interaction page, so as to more accurately capture the commonalities and differences in user preferences.

[0081] When all reference learning data are assigned to the corresponding sequences, the server starts to train the user preference prediction network using these data sequences. In this process, the server will use each reference learning data sequence as a training sample and input it into the user preference prediction network. The network will learn how to predict the user's preferences based on their behavior on the short video interactive page by analyzing the user behavior patterns in these data sequences.

[0082] Specifically, the network tries to find the intrinsic connection between user behavior and user preferences, such as which behavior patterns are usually associated with specific preferences. By continuously adjusting its internal parameters and weights, the network gradually learns how to accurately predict user preferences based on the input user viewing record data. This process may continue for multiple training cycles until the network's prediction performance reaches a preset standard or no longer improves significantly.

[0083] Ultimately, after such a knowledge learning process, the user preference prediction network will be able to accurately predict the user's possible preferences based on their behavior on the short video interactive page, thereby providing strong support for subsequent personalized recommendations.

[0084] In a possible implementation, the feature association of sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0085] The sample user behavior record set is used as supervision data, and feature association is performed between the user viewing preference and the sample user behavior record set to generate first learning data.

[0086] The user preference prediction network includes a deep learning network, and the user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0087] The user viewing preferences in the first learning data are loaded into the deep learning network, and a first predicted user behavior preference set is generated through the deep learning network.

[0088] The deep learning network is subjected to knowledge learning based on the sample user behavior record set in the first learning data and the first predicted user behavior preference set.

[0089] In a possible implementation, the feature association of sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0090] The user viewing preference is used as supervision data, and feature association is performed between a sample user behavior record set and the user viewing preference to generate second learning data.

[0091] The user preference prediction network includes a deep learning network, and the user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0092] The sample user behavior record set in the second learning data is loaded into the deep learning network, and a first predicted user viewing preference is generated through the deep learning network.

[0093] The deep learning network is subjected to knowledge learning according to the user viewing preference in the second learning data and the first predicted user viewing preference.

[0094] In a possible implementation, the feature association of sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0095] The sample user behavior record set is used as supervision data, and feature association is performed between the user interaction pattern features and the sample user behavior record set to generate third learning data.

[0096] The user preference prediction network includes a feature extraction unit and a deep learning network. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0097] The user interaction pattern features in the third learning data are loaded into the feature extraction unit, and a first interaction feature vector sequence is generated by the feature extraction unit.

[0098] The first interactive feature vector sequence is loaded into the deep learning network, and a second predicted user behavior preference set is generated through the deep learning network.

[0099] Based on the sample user behavior record set in the third learning data and the second predicted user behavior preference set, knowledge learning is performed on the feature extraction unit and the deep learning network.

[0100] In a possible implementation, the feature association of sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0101] The user viewing preference is used as supervision data, and feature association is performed between the user interaction pattern features and the user viewing preference to generate fourth learning data.

[0102] The user preference prediction network includes a feature extraction unit and a deep learning network. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0103] The user interaction pattern features in the fourth learning data are loaded into the feature extraction unit, and a second interaction feature vector sequence is generated by the feature extraction unit.

[0104] The second interactive feature vector sequence is loaded into the deep learning network, and a second predicted user viewing preference is generated through the deep learning network.

[0105] The feature extraction unit and the deep learning network are subjected to knowledge learning according to the user viewing preference in the fourth learning data and the second predicted user viewing preference.

[0106] In a possible implementation, the feature association of sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0107] The user interaction pattern features are used as supervision data, and feature association is performed between the user viewing preference and the user interaction pattern features to generate fifth learning data.

[0108] The user preference prediction network includes a deep learning network and a multi-layer perceptron. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0109] The user viewing preference in the fifth learning data is loaded into the deep learning network, and a first interactive behavior feature is generated through the deep learning network.

[0110] The first interactive behavior feature is loaded into the multi-layer perceptron, and a first predicted user interactive behavior pattern is generated by the multi-layer perceptron.

[0111] Based on the user interaction pattern characteristics in the fifth learning data and the first predicted user interaction behavior pattern, knowledge learning is performed on the deep learning network and the multilayer perceptron.

[0112] In a possible implementation, the feature association of sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0113] The user interaction pattern features are used as supervision data, and feature association is performed between the sample user behavior record set and the user interaction pattern features to generate sixth learning data.

[0114] The user preference prediction network includes a deep learning network and a multi-layer perceptron. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0115] The sample user behavior record set in the sixth learning data is loaded into the deep learning network, and a second interactive behavior feature is generated through the deep learning network.

[0116] The second interactive behavior feature is loaded into the multi-layer perceptron, and a second predicted user interactive behavior pattern is generated by the multi-layer perceptron.

[0117] Based on the user interaction pattern characteristics in the sixth learning data and the second predicted user interaction behavior pattern, knowledge learning is performed on the deep learning network and the multilayer perceptron.

[0118] In a possible implementation, the feature association of sample user viewing record data of any two short video interactive pages to generate reference learning data includes:

[0119] The user interaction pattern features are used as supervision data, and feature association is performed between the user viewing environment conditions and the user interaction pattern features to generate seventh learning data.

[0120] The user preference prediction network includes a deep learning network and a conditional diffusion network. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including:

[0121] The user viewing environment conditions in the seventh learning data are loaded into the deep learning network, and the user viewing environment conditions in the seventh learning data are feature extracted by the deep learning network to generate user viewing environment vector data.

[0122] The user viewing environment vector data is loaded into the conditional diffusion network, and a third predicted user interaction behavior pattern is generated through the conditional diffusion network.

[0123] Based on the user interaction pattern characteristics in the seventh learning data and the third predicted user interaction behavior pattern, knowledge learning is performed on the deep learning network and the conditional diffusion network.

[0124] Before loading the user viewing environment vector data into the conditional diffusion network, the method further includes:

[0125] Get multiple user interaction mode features.

[0126] The conditional diffusion network is subjected to knowledge learning based on the plurality of user interaction pattern features.

[0127] In the above embodiments, the big data recommendation system applied to the short video platform for executing the above method embodiments has at least one processor, a control module (chip set) coupled to at least one of the (at least one) processors, a memory coupled to the control module, a non-volatile memory (NVM) / storage device coupled to the control module, at least one input / output device coupled to the control module, and a network interface coupled to the control module.

[0128] The processor may include at least one single-core or multi-core processor, and the processor may include any combination of general-purpose processors or special-purpose processors (such as graphics processors, application processors, baseband processors, etc.). For some alternative implementations, the big data recommendation system applied to the short video platform can be used as the big data recommendation system device applied to the short video platform such as the gateway described in the embodiments of this application.

[0129] For some alternative embodiments, the big data recommendation system applied to the short video platform may include at least one computer-readable medium (e.g., a memory or NVM / storage device) having instructions and at least one processor integrated with the at least one computer-readable medium and configured to execute instructions to implement a module to perform the actions described in the present disclosure.

[0130] For one embodiment, the control module may include any suitable interface controller to provide any suitable interface to at least one of the processor(s) and / or any suitable device or component in communication with the control module.

[0131] The control module may include a memory controller module to provide an interface to the memory. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0132] The memory may be used, for example, to load and store data and / or instructions for a big data recommendation system applied to a short video platform. For one embodiment, the memory may include any suitable volatile memory, for example, a suitable DRAM.

[0133] For one embodiment, the control module may include at least one input / output controller to provide an interface to the NVM / storage device and (at least one) input / output device.

[0134] For example, the NVM / storage device may be used to store data and / or instructions. The NVM / storage device may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable (at least one) non-volatile storage device (e.g., at least one hard disk drive (HDD), at least one compact disk (CD) drive, and / or at least one digital versatile disk (DVD) drive).

[0135] The NVM / storage device may include a storage resource that is physically part of a device on which the big data recommendation system for short video platforms is installed, or it may be accessible to the device without being part of the device. For example, the NVM / storage device may be accessed via (at least one) input / output device based on a network.

[0136] (At least one) input / output device may provide an interface for the big data recommendation system applied to the short video platform to communicate with any other appropriate device, and the input / output device may include a communication component, a pinyin component, a sensor component, etc. The network interface may provide an interface for the big data recommendation system applied to the short video platform to predict network communication based on at least one user preference. The big data recommendation system applied to the short video platform may wirelessly communicate with at least one component of the wireless network based on any standard and / or protocol of at least one wireless network standard and / or protocol, for example, access a wireless network based on a communication standard.

[0137] For one embodiment, at least one of the (at least one) processor may be loaded together with the logic of at least one controller of the control module (e.g., a memory controller module). For one embodiment, at least one of the (at least one) processor may be loaded together with the logic of at least one controller of the control module to form a system level load. For one embodiment, at least one of the (at least one) processor may be integrated on the same die with the logic of at least one controller of the control module. For one embodiment, at least one of the (at least one) processor may be integrated on the same die with the logic of at least one controller of the control module to form a system on chip (SoC).

[0138] The embodiments of the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

[0139] An embodiment of the present invention discloses a computer-readable storage medium, which stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps of the big data recommendation method applied to a short video platform described in the aforementioned embodiment.

[0140] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute the steps of the big data recommendation method applied to a short video platform described in the aforementioned embodiment.

[0141] The device embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple user preference prediction network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0142] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution can be essentially or partly contributed to the prior art in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other medium that can be used to be computer-readable with or store data.

[0143] Finally, it should be noted that what is disclosed above is only a preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A big data recommendation method applied to a short video platform, characterized in that: The method comprises: Collect sample user viewing record data of multiple short video interactive pages from the user behavior big data of the target short video platform; Perform feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data; Based on the plurality of reference learning data, knowledge learning is performed on the user preference prediction network, and after user preference prediction is performed on any input user viewing record data based on the target user preference prediction network that has completed knowledge learning, data recommendation is performed based on the user preference prediction result; Wherein, the user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including: Generate a pair of short video interaction pages corresponding to the reference learning data based on the short video interaction pages corresponding to the two sample user viewing record data in the reference learning data; Assigning reference learning data of the same short video interactive page pair to the same reference learning data sequence; Performing knowledge learning on the user preference prediction network according to the plurality of reference learning data sequences; The step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes: Using a sample user behavior record set as supervision data, and performing feature association between the user viewing preference and the sample user behavior record set to generate first learning data; The user preference prediction network includes a deep learning network, and the user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including: Loading the user viewing preferences in the first learning data into the deep learning network, and generating a first predicted user behavior preference set through the deep learning network; Performing knowledge learning on the deep learning network according to the sample user behavior record set in the first learning data and the first predicted user behavior preference set; The step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes: Using the user's viewing preference as supervision data, and performing feature association between the sample user behavior record set and the user's viewing preference to generate second learning data; The user preference prediction network includes a deep learning network, and the user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including: Loading a set of sample user behavior records in the second learning data into the deep learning network, and generating a first predicted user viewing preference through the deep learning network; performing knowledge learning on the deep learning network according to the user viewing preference in the second learning data and the first predicted user viewing preference; The step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes: Using the sample user behavior record set as supervision data, performing feature association between the user interaction pattern features and the sample user behavior record set to generate third learning data; The user preference prediction network includes a feature extraction unit and a deep learning network. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including: Loading the user interaction pattern features in the third learning data into the feature extraction unit, and generating a first interaction feature vector sequence through the feature extraction unit; Loading the first interactive feature vector sequence into the deep learning network, and generating a second predicted user behavior preference set through the deep learning network; Performing knowledge learning on the feature extraction unit and the deep learning network according to the sample user behavior record set in the third learning data and the second predicted user behavior preference set; The step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes: Using the user's viewing preference as supervision data, and performing feature association between the user's interaction pattern characteristics and the user's viewing preference to generate fourth learning data; The user preference prediction network includes a feature extraction unit and a deep learning network. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including: Loading the user interaction pattern features in the fourth learning data into the feature extraction unit, and generating a second interaction feature vector sequence through the feature extraction unit; Loading the second interactive feature vector sequence into the deep learning network, and generating a second predicted user viewing preference through the deep learning network; performing knowledge learning on the feature extraction unit and the deep learning network according to the user viewing preference in the fourth learning data and the second predicted user viewing preference; The step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes: Using the user interaction pattern feature as supervision data, performing feature association between the user viewing preference and the user interaction pattern feature, and generating fifth learning data; The user preference prediction network includes a deep learning network and a multi-layer perceptron. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including: Loading the user viewing preference in the fifth learning data into the deep learning network, and generating a first interactive behavior feature through the deep learning network; Loading the first interactive behavior feature into the multi-layer perceptron, and generating a first predicted user interactive behavior pattern through the multi-layer perceptron; Performing knowledge learning on the deep learning network and the multi-layer perceptron according to the user interaction pattern characteristics in the fifth learning data and the first predicted user interaction behavior pattern; The step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes: Using the user interaction pattern feature as supervision data, performing feature association between the sample user behavior record set and the user interaction pattern feature to generate sixth learning data; The user preference prediction network includes a deep learning network and a multi-layer perceptron. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including: Loading a set of sample user behavior records in the sixth learning data into the deep learning network, and generating a second interactive behavior feature through the deep learning network; Loading the second interactive behavior feature into the multi-layer perceptron, and generating a second predicted user interactive behavior pattern through the multi-layer perceptron; Performing knowledge learning on the deep learning network and the multilayer perceptron according to the user interaction pattern features in the sixth learning data and the second predicted user interaction behavior pattern; The step of performing feature association on sample user viewing record data of any two short video interactive pages to generate reference learning data includes: Using the user interaction pattern feature as supervision data, performing feature association between the user viewing environment condition and the user interaction pattern feature, and generating seventh learning data; The user preference prediction network includes a deep learning network and a conditional diffusion network. The user preference prediction network is subjected to knowledge learning based on the plurality of reference learning data, including: Loading the user viewing environment conditions in the seventh learning data into the deep learning network, performing feature extraction on the user viewing environment conditions in the seventh learning data through the deep learning network, and generating user viewing environment vector data; Loading the user viewing environment vector data into the conditional diffusion network, and generating a third predicted user interaction behavior pattern through the conditional diffusion network; Performing knowledge learning on the deep learning network and the conditional diffusion network according to the user interaction pattern characteristics in the seventh learning data and the third predicted user interaction behavior pattern; Before loading the user viewing environment vector data into the conditional diffusion network, the method further includes: Obtain multiple user interaction pattern features; The conditional diffusion network is subjected to knowledge learning based on the plurality of user interaction pattern features.

2. A big data recommendation system applied to a short video platform, characterized in that: It includes a processor and a computer-readable storage medium, wherein the computer-readable storage medium stores machine-executable instructions, and when the machine-executable instructions are executed by a computer, the big data recommendation method applied to a short video platform as described in claim 1 is implemented.

Images