Application program recommendation method and related device thereof

Abstract

The invention discloses an application program recommendation method and a related device thereof. The method comprises the following steps: in response to a side edge sliding operation of a user, obtaining equipment state data of multiple dimensions; performing multi-dimensional feature fusion on the multi-dimensional equipment state data to obtain a user state feature; carrying out weight distribution on each application program based on the user state characteristics, and determining a sorted application program recommendation list; and popping up a screen sidebar, and displaying the application recommendation list in the screen sidebar. The method comprises the following steps: in response to a side edge sliding operation of a user, collecting equipment state data of multiple dimensions, and carrying out multi-dimensional feature fusion on the equipment state data to generate a state feature of the current user; and according to the user state characteristics, the weights are allocated to the applications, the sorted application recommendation list is generated and displayed in the screen sidebar, and the technical problem that the accuracy and flexibility of the display content of the screen sidebar are low can be solved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to an application recommendation method and related devices thereof. Background Art

[0002] With the popularity of smart devices, users' frequency of use and dependence on smart devices are increasing. As a convenient operation interface, the sidebar function of smart devices is usually used to quickly access users' frequently used applications or functions.

[0003] In the prior art, the main form of the sidebar is mostly static display, and users need to manually configure commonly used applications or functions for quick access in the sidebar. However, this static configuration method has obvious limitations and is difficult to meet the needs of users in different usage scenarios. This sidebar design that lacks dynamic adjustment not only increases the time cost of user operation, but also reduces operation efficiency.

[0004] Therefore, it is necessary to propose a solution to improve the pertinence and personalization of the content displayed in the screen sidebar. Summary of the invention

[0005] The main technical problem solved by the present application is to provide an application recommendation method and related devices, which can solve the technical problems of low accuracy and flexibility of content displayed in the screen sidebar.

[0006] In order to solve the above technical problems, a technical solution adopted by the present application is: the application recommendation method includes:

[0007] In response to a side sliding operation by a user, acquiring device status data of multiple dimensions; wherein the device status data includes at least one of scenario data, sensor data, preference data, social communication data, and hardware data;

[0008] Performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features;

[0009] Based on the user status characteristics, weights are assigned to the applications, and a ranked application recommendation list is determined;

[0010] A screen sidebar pops up, and the application recommendation list is displayed in the screen sidebar.

[0011] Optionally, performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features includes:

[0012] Extracting a shared feature layer of the device status data;

[0013] Determine at least one perspective attention of each dimension using multi-head self-attention;

[0014] A target shared feature layer is screened from each of the shared feature layers through the perspective, and multi-scale feature fusion is performed on the target shared feature layer to obtain the user state feature.

[0015] Optionally, performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features includes:

[0016] Using factor analysis method, extracting low-level features and high-level features from the device status data in multiple dimensions;

[0017] Performing layer-by-layer splicing on the low-level features to form a first splicing feature, and performing layer-by-layer splicing on the high-level features to form a second splicing feature;

[0018] The first splicing feature and the second splicing feature are fused to obtain the user state feature.

[0019] Optionally, the assigning weights to the applications based on the user status features to determine a ranked application recommendation list includes:

[0020] Obtain historical behavior characteristics of similar users;

[0021] Perform sliding window and attenuation function detection on the historical behavior characteristics and the user status characteristics to generate recommendation weights corresponding to the current applications;

[0022] The application recommendation list is generated according to the recommendation weights.

[0023] Optionally, obtaining historical behavior characteristics of similar users includes:

[0024] Obtain historical behavior datasets;

[0025] Perform collaborative filtering on the data in the historical behavior data set, and calculate similar user groups to obtain the historical behavior data of the similar users;

[0026] Corresponding implicit features are extracted from the historical behavior data to obtain the historical behavior features.

[0027] Optionally, the step of assigning weights to the applications based on the user status features and determining a ranked application recommendation list may then include:

[0028] In response to the number of operations of at least one target application exceeding a preset number within a preset time period, determining a target function of the target application from a pre-built function association database;

[0029] matching relevant applications associated with the target function;

[0030] Based on the target application and the related applications, the application recommendation list is updated, and other applications that have a low correlation with the target application are removed.

[0031] Optionally, the process of constructing the function association database includes:

[0032] Classifying and describing the functions of each of the application programs to obtain function information;

[0033] Based on the functional information, each of the application programs is cross-clustered to construct the functional association database.

[0034] In order to solve the above technical problems, another technical solution adopted by the present application is: an application recommendation device, the application recommendation device comprising:

[0035] An acquisition module, configured to acquire device status data of multiple dimensions in response to a side sliding operation of a user; wherein the device status data includes at least one of scenario data, sensor data, preference data, social communication data, and hardware data;

[0036] A fusion module, used for performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features;

[0037] An allocation module, configured to allocate weights to the applications based on the user status characteristics and determine a ranked application recommendation list;

[0038] The display module is used to pop up a screen sidebar and display the application recommendation list in the screen sidebar.

[0039] In order to solve the above technical problems, another technical solution adopted by the present application is: an electronic device, the electronic device comprising:

[0040] A memory for storing executable program codes;

[0041] The processor is used to call and run the executable program code from the memory, so that the electronic device executes the application recommendation method as described in any one of the above.

[0042] To solve the above technical problems, another technical solution adopted in the present application is: a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the application recommendation method as described in any one of the above.

[0043] Different from the prior art, an embodiment of the present application provides an application recommendation method, which includes: in response to a user's side sliding operation, obtaining device status data of multiple dimensions; wherein the device status data includes at least one of scenario data, sensor data, preference data, social communication data, and hardware data; performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features; assigning weights to each application based on the user status features to determine a sorted application recommendation list; and popping up a screen sidebar to display the application recommendation list in the screen sidebar.

[0044] By responding to the user's side sliding operation, device status data in multiple dimensions is collected, and multi-dimensional feature fusion is performed on the device status data to generate the status characteristics of the current user; then, based on the user status characteristics, weights are assigned to each application, and a sorted application recommendation list is generated and displayed in the screen sidebar, which can solve the technical problems of low accuracy and flexibility of the content displayed in the screen sidebar. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a flowchart of the first embodiment of the application recommendation method of the present application;

[0046] Figure 2 This is a flowchart of the second embodiment of the application recommendation method of the present application;

[0047] Figure 3 It is a flowchart of the third embodiment of the application recommendation method of the present application;

[0048] Figure 4 This is a flowchart of the fourth embodiment of the application recommendation method of the present application;

[0049] Figure 5 This is a flowchart of the fifth embodiment of the application recommendation method of the present application;

[0050] Figure 6 is an exemplary structural block diagram of an electronic device of the application recommendation method of the present application;

[0051] Figure 7 It is an exemplary structural block diagram of a computer-readable storage medium of the application recommendation method of the present application. DETAILED DESCRIPTION

[0052] The present application is described in detail below with reference to the accompanying drawings and implementation methods.

[0053] See also Figure 1 , Figure 11 is a flowchart of the first embodiment of the application recommendation method of the present application. It should be noted that if there is substantially the same result, the method of the present application is not limited to the first embodiment of the application recommendation method of the present application. Figure 1 The process sequence shown is limited. Figure 1 As shown, the application recommendation method includes:

[0054] Step S110, in response to the side sliding operation of the user, obtaining device status data of multiple dimensions; wherein the device status data includes at least one of scenario data, sensor data, preference data, social communication data, and hardware data;

[0055] This embodiment takes into account that most of the functions of the sidebar of smart phones are currently static displays, and users need to manually configure common applications or functions. However, in different scenarios, user needs will change, but the existing sidebar cannot actively perceive and adapt to these changes, and there are limitations. Therefore, this embodiment uses the status data of different dimensions of smart devices to construct a complete user portrait from the side, accurately predict user preferences, and thus improve the flexibility of displaying content on the sidebar of the smart device screen.

[0056] Among them, device status data can represent the operating status of smart devices. Specifically, contextual data may include time and location, device connection status, and network status; for example, in a weak network environment, the system can reduce the weight of applications that are highly dependent on the network, and increase the weight of applications such as online video and cloud storage in a strong network environment. Sensor data may include light sensor data, sound sensor data, temperature sensor data, and motion sensor data. Preference data may include user interest tags, such as music, sports, shopping, etc., as well as user operational feedback on recommended content. Social communication data may include the user's recent contact frequency, schedule, and calendar information. Hardware data may include battery power.

[0057] Specifically, the user can slide the side of the screen of the smart device, and the system can automatically trigger the data acquisition module to obtain the device status data of the smart device in real time. The gesture detection interface can be used to detect the user's sliding gesture. For example, by monitoring the edge sliding gesture, after detecting the side sliding operation, the sidebar can be triggered to pop up.

[0058] Step S120, performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features;

[0059] Specifically, multi-dimensional feature fusion can comprehensively analyze the state data of smart devices in each dimension, so that the user state feature can express multi-dimensional information through the state data of smart devices. The fused user state feature helps the system understand the potential correlation between multiple states, so as to make more accurate predictions. Among them, the user state feature can be used to characterize the user's current state information.

[0060] In an exemplary embodiment, a feature layer shared between different dimensions can be extracted from the device status data; then, a multi-head self-attention mechanism can be used to identify key information related to the shared feature layer and the dimension, i.e., perspective attention; then, based on perspective attention, the shared feature layer can be subjected to multi-dimensional fusion processing to integrate information from different dimensions and obtain user status features containing information from each dimension.

[0061] In another exemplary embodiment, a factor analysis method can be used to extract low-level features and high-level features from device status data of multiple dimensions, and separate feature information at different levels; then, all low-level features are spliced ​​by layer to generate a first spliced ​​feature; similarly, high-level features are spliced ​​by layer to generate a second spliced ​​feature; the first spliced ​​feature and the second spliced ​​feature are fused to obtain a user status feature containing comprehensive features.

[0062] Step S130, assigning weights to the applications based on the user status characteristics, and determining a ranked application recommendation list;

[0063] In this embodiment, weight allocation can be performed based on machine learning or deep learning models, such as using models such as multi-layer perceptron (MLP) and long short-term memory network (LSTM), taking into account historical behavior data and user status characteristics, and calculating the application weight in the current scenario. Then, the applications can be arranged in order from high to low weights to form a sorted recommendation list. By assigning weights, the system can automatically filter out applications that do not match the current situation and only display high-weight applications, which simplifies the interface and improves operational efficiency.

[0064] Step S140: a screen sidebar pops up, and the application recommendation list is displayed in the screen sidebar.

[0065] Among them, the sidebar can be located in the area at the edge of the smart device screen, used to display related applications and functions.

[0066] Specifically, the sidebar can be displayed through a graphical user interface. Compared with the traditional homepage or multitasking view, the design of the sidebar reduces the steps for users to switch applications. For example, a list view component can be added to the sidebar layout, and the recommended applications can be sorted by weight and loaded into the list one by one; at the same time, a click event listener can be added to the list item so that the user can directly open the application when clicking.

[0067] This embodiment collects device status data of multiple dimensions by responding to the user's side sliding operation, and performs multi-dimensional feature fusion on the device status data to generate the status characteristics of the current user; then, based on the user status characteristics, a weight is assigned to each application, and a sorted application recommendation list is generated and displayed in the screen sidebar, which can solve the technical problem of low accuracy and flexibility of the content displayed in the screen sidebar.

[0068] See also Figure 2 , Figure 2 This is a flow chart of the second embodiment of the application recommendation method of the present application. The method comprises the following steps:

[0069] Step S210, in response to the side sliding operation of the user, obtaining device status data of multiple dimensions; wherein the device status data includes at least one of scenario data, sensor data, preference data, social communication data, and hardware data;

[0070] The above step S110 may be performed and will not be described in detail here.

[0071] Step S220, extracting the shared feature layer of the device status data;

[0072] Specifically, the shared feature layer can contain information shared between dimensions to represent the current operating status of the device. The original device status data can be initially extracted and converted into a feature space representation to reduce the data dimension and improve computational efficiency, while preparing a unified feature representation for subsequent processing.

[0073] Step S230, using multi-head self-attention, determining at least one perspective focus of each dimension;

[0074] Specifically, multi-head self-attention can be a deep learning mechanism that determines the focus on each dimension by focusing on different features or patterns on different attention heads. Perspective attention is used to selectively control the flow of information. That is, through multiple independent attention heads, different weights are assigned to each dimension of the shared feature layer, so that the most relevant information about the device status can be found in multiple dimensions, allowing the deep learning model to capture multiple feature patterns in the data and improve the accuracy of complex data processing.

[0075] Step S240 , screening a target shared feature layer from each of the shared feature layers through the perspective focus, and performing multi-scale feature fusion on the target shared feature layer to obtain the user state feature.

[0076] Specifically, unimportant shared feature layers can be filtered out through visual attention, and the most relevant target shared feature layers can be retained. For example, a threshold can be set to filter out shared feature layers with higher attention values ​​and discard shared feature layers with lower attention values.

[0077] At the same time, each viewpoint attention can be regarded as a gating unit. By activating different gating units, the important target shared feature layer in the current state of the smart device can be retained. Then, according to the viewpoint attention, the target shared feature layer can generate feature representations of different scales, that is, resolution or abstraction level, and these feature representations can be fused to obtain user state features, which can capture the user's feature information from multiple granularities.

[0078] Step S250, assigning weights to the applications based on the user status characteristics, and determining a ranked application recommendation list;

[0079] Specifically, the above step S130 may be performed, which will not be described in detail here.

[0080] Step S260: a screen sidebar pops up, and the application recommendation list is displayed in the screen sidebar.

[0081] The above step S140 may be performed and will not be described in detail here.

[0082] In this embodiment, the original device status data is converted into a user status feature rich in semantics and multi-dimensional information through shared feature extraction, focusing of attention mechanism and multi-dimensional fusion.

[0083] See also Figure 3 , Figure 3 This is a flow chart of the third embodiment of the application recommendation method of the present application. The method comprises the following steps:

[0084] Step S310, in response to the side sliding operation of the user, obtaining device status data of multiple dimensions; wherein the device status data includes at least one of scenario data, sensor data, preference data, social communication data, and hardware data;

[0085] The above step S110 may be performed and will not be described in detail here.

[0086] Step S320, extracting low-level features and high-level features from the device status data of multiple dimensions using a factor analysis method;

[0087] Specifically, factor analysis can extract representative factors or latent variables from the original equipment status data to analyze the multi-dimensional characteristics of the equipment status data, and decompose these data into features at different levels, making the equipment status data structure clearer.

[0088] Among them, low-level features are used to describe surface state changes, and can contain basic information of device state data, such as directly observable device states, such as instantaneous readings of sensor data. High-level features can be patterns or trends extracted from device state data, which are used to characterize state changes of smart devices over a longer period of time or at a higher level, such as trend information or overall health status.

[0089] Step S330, performing layer stitching on the low-level features to form a first stitching feature, and performing layer stitching on the high-level features to form a second stitching feature;

[0090] Among them, the first splicing feature can be used to represent all low-level features as a whole, so that the model can more comprehensively capture the real-time status of the device; the second splicing feature can effectively capture higher-level or abstract status information, thereby forming a more general and stable status description. Through hierarchical splicing, a unified representation of low-level and high-level features is constructed respectively, which improves the accuracy of multi-dimensional feature fusion and can avoid information loss to a certain extent.

[0091] Step S340: fusing the first splicing feature and the second splicing feature to obtain the user state feature.

[0092] Specifically, the first concatenated feature and the second concatenated feature may be fused using weighted average, convolution operation, or multi-layer perceptron. Through fusion, the model combines low-level and high-level features, and can take into account both the details of the data and the underlying trends or patterns during analysis, thereby providing a more comprehensive device status description and making the user status feature more representative and highly accurate in prediction.

[0093] Step S350, assigning weights to the applications based on the user status characteristics, and determining a ranked application recommendation list;

[0094] The above step S130 may be performed and will not be described in detail here.

[0095] Step S360: a screen sidebar pops up, and the application recommendation list is displayed in the screen sidebar.

[0096] The above step S140 may be performed and will not be described in detail here.

[0097] In this embodiment, low-level and high-level features are separated by factor analysis, and unified state features are formed through hierarchical splicing and fusion, so that the final user state feature integrates different levels of device state information, including both the instantaneous state at the detail level and the trend and abstract information at a higher level, which can provide accurate and multi-dimensional information support for applications such as device state monitoring, user behavior analysis or prediction, etc.

[0098] See also Figure 4 , Figure 4 This is a flowchart of the fourth embodiment of the application recommendation method of the present application. The method includes the following steps:

[0099] Step S410, in response to the side sliding operation of the user, obtaining device status data of multiple dimensions; wherein the device status data includes at least one of scenario data, sensor data, preference data, social communication data, and hardware data;

[0100] The above step S110 may be performed and will not be described in detail here.

[0101] Step S420, performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features;

[0102] The above step S120 may be performed and will not be described in detail here.

[0103] Step S430, obtaining historical behavior characteristics of similar users;

[0104] Specifically, historical behavior features may include usage frequency, usage time, usage type preference, etc. of each application. By aggregating the behavior features of multiple similar users, the deep learning model can more accurately predict the application type or function that the current user may like.

[0105] Further, step S430 includes: obtaining a historical behavior data set; performing collaborative filtering on the data in the historical behavior data set, and calculating similar user groups to obtain historical behavior data of the similar users; extracting corresponding implicit features from the historical behavior data to obtain the historical behavior features.

[0106] Specifically, the historical behavior data set may include the interaction records of multiple users on the social platform, such as clicks, browsing, searches, purchases, comments and other information. Then, collaborative filtering can be used to filter out the historical behavior data corresponding to similar user groups from the historical behavior data set. This set of behavior data of similar user groups contains similar behavior habits or preferences of multiple users. Then, implicit features can be extracted from the historical behavior data of similar users through at least one of dimensionality reduction, matrix decomposition or autoencoder to map the behavior data of similar users to a low-dimensional feature space and generate a feature vector containing user behavior habits or preferences.

[0107] Step S440, performing sliding window and attenuation function detection on the historical behavior characteristics and the user status characteristics, and generating a recommendation weight corresponding to each of the current applications;

[0108] Specifically, the deep learning model can be used to perform sliding window detection and attenuation function processing on the user's historical behavior characteristics and current user status characteristics. Among them, the sliding window can be used to extract time series information from the feature data, focusing on the changes in the characteristics over a period of time. The sliding window captures the user's short-term behavior characteristics by calculating the frequency, mean and other characteristics of the user's behavior within a specific time interval. Then, the detection results can be weighted averaged or nonlinearly combined to combine the historical behavior feature weights and the status feature weights into the recommendation weights.

[0109] Step S450: generating the application recommendation list according to the recommendation weight.

[0110] Specifically, the application recommendation list can be a ranking generated by combining the comprehensive weights of different application categories and individual applications, wherein applications with higher weights are placed at the front of the list, effectively increasing the possibility of user selection.

[0111] This embodiment combines the historical behaviors of similar users and the real-time status characteristics of users, uses sliding windows and decay functions to process data, generates dynamic recommendation weights to output an application recommendation list, and can effectively balance the user's long-term interests and recent changes, thereby providing application recommendations that better meet the user's immediate needs.

[0112] See also Figure 5 , Figure 5 This is a flowchart of the fifth embodiment of the application recommendation method of the present application. After step S140, the following steps are also included:

[0113] Step S510, in response to the number of operations of at least one target application exceeding a preset number within a preset time period, determining a target function of the target application from a pre-built function association database;

[0114] The threshold may be a fixed value or may be dynamically adjusted according to the user's operation behavior. For example, if the user opens the navigation application multiple times within an hour, the system will identify the navigation application as the target application.

[0115] Specifically, by setting an operation count threshold, when the number of operations of a user on a certain application exceeds this threshold within a preset time period, the target application that the user is frequently using can be identified. At the same time, a function association database can be pre-built in the smart device, which stores the association between each application and its specific function. By determining the user's specific target function, it can help to recommend related applications more accurately, so that the recommendation is not only based on the application category, but also based on the specific user function requirements.

[0116] Step S520, matching relevant applications associated with the target function;

[0117] Specifically, after identifying the target functions of the target application, the system will match other applications with similar or highly related functions through the function association database. The matched related applications can help users access the tools they may need more quickly and reduce the time users spend searching for other applications. This method effectively improves the accuracy and diversity of recommendations, allowing users to easily obtain other applications that are closely related to their current needs while using the main application.

[0118] Step S530: based on the target application and the related applications, update the application recommendation list, and remove other applications that have a low correlation with the target application.

[0119] After obtaining the target application and related applications, the recommendation list can be rearranged to put these applications in the front position and increase their recommendation priority. At the same time, in order to optimize the recommendation effect, the system will remove applications that are less relevant to the target application to reduce user distraction. For example, when users frequently use navigation applications, music or video applications may not be relevant temporarily, so they can be temporarily removed from the recommendation list. Among them, real-time data frameworks such as data monitoring mechanisms can be used to ensure that recommended content is automatically updated as user status changes.

[0120] In some embodiments, the process of constructing the function association database includes:

[0121] The functions of each of the application programs are classified and described to obtain function information; based on the function information, each of the application programs is cross-clustered to construct the function association database.

[0122] Specifically, function classification is used to help the system understand the purpose and application scenarios of the application; and based on the function classification, a detailed description is written for each application function. The description information should not only include the function name, but also cover the specific purpose of the function, target users, frequency of use, and potential usage scenarios.

[0123] In this embodiment, cross clustering can be used to classify applications according to multiple functional dimensions at the same time, so as to find application groups that have common points in different functions. Through cross clustering, association relationships containing multiple different categories of applications can be generated. For example, map navigation applications and taxi applications are clustered in navigation functions. During the clustering process, the system will select several key functional dimensions for clustering based on the previously extracted functional information. These dimensions will serve as the main axis of clustering to help the system identify the functional association between different applications. Among them, K-means clustering, hierarchical clustering and other algorithms can be used for clustering. After clustering is completed, the system will generate association relationships based on the applications in each cluster. These association relationships reflect the functional similarities between applications. For example, through clustering, it is found that taxi applications and map applications are closely related in navigation functions, and the system will record this association relationship. The association relationship can be a mark of different weights such as strong association and weak association to help the system determine the priority when recommending.

[0124] The structure of the functional association database may include fields such as application name, functional category, list of associated applications, and association weight. These associations are stored in a tree or graph structure for quick retrieval during recommendation. In addition, as new applications are added or user needs change, the functional association database may need to be updated regularly. Automated machine learning processes or manual reviews can be used to add new applications to appropriate clusters and dynamically adjust the associations of existing applications. For example, when a new application that combines navigation and social networking is launched, the system can associate it with a group of related applications for navigation and social networking functions.

[0125] This embodiment, by updating the recommendation list in real time, can not only reflect the user's current needs in real time, but also reduce the interference of irrelevant applications, so that users can find applications that meet their needs more quickly and avoid users having difficulty in choosing due to too much recommended content; at the same time, through the function association database, corresponding applications can be recommended based on functional requirements, thereby improving the accuracy of recommendations.

[0126] In addition, the embodiment of the present application also provides an application recommendation device, the application recommendation device comprising:

[0127] An acquisition module, configured to acquire device status data of multiple dimensions in response to a side sliding operation of a user; wherein the device status data includes at least one of scenario data, sensor data, preference data, social communication data, and hardware data;

[0128] A fusion module, used for performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features;

[0129] An allocation module, configured to allocate weights to the applications based on the user status characteristics and determine a ranked application recommendation list;

[0130] The display module is used to pop up a screen sidebar and display the application recommendation list in the screen sidebar.

[0131] See also Figure 6 , Figure 6 is an exemplary structural block diagram of an electronic device of the application recommendation method of the present application. Figure 6 As shown, the electronic device 600 of the present application may include a processor 601 and a memory 602, wherein the processor 601 and the memory 602 communicate with each other via a bus. The memory 602 stores program instructions for application recommendation, and when the program instructions are executed by the processor 601, the processor executes the above-mentioned related method steps to implement an application recommendation method in the above-mentioned embodiment.

[0132] See also Figure 7 , Figure 7 is an exemplary structural block diagram of a computer-readable storage medium of the application recommendation method of the present application. Figure 7 As shown, the computer-readable storage medium 700 stores a computer program 701. When the computer program 701 is executed by a processor on a computer, the computer executes the above-mentioned related method steps to implement an application recommendation method in the above-mentioned embodiment.

[0133] The above scheme obtains device status data of multiple dimensions in response to the user's side sliding operation; performs multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features; assigns weights to each application based on the user status features to determine a ranked application recommendation list; pops up a screen sidebar to display the application recommendation list in the screen sidebar. By responding to the user's side sliding operation, collecting device status data of multiple dimensions, and performing multi-dimensional feature fusion on the device status data to generate the status features of the current user; and then assigning weights to each application based on the user status features, generating a ranked application recommendation list, and displaying it in the screen sidebar, the technical problems of low accuracy and flexibility of the content displayed in the screen sidebar can be solved.

[0134] In the several embodiments provided in the present application, it should be understood that the disclosed methods, electronic devices and storage media can be implemented in other ways. For example, the device implementation described above is only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0135] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0136] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0137] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to execute all or part of the steps of the application recommendation method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

[0138] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for recommending an application, characterized in that: The application recommendation method comprises: In response to a side sliding operation by a user, acquiring device status data of multiple dimensions; wherein the device status data includes at least one of scenario data, sensor data, preference data, social communication data, and hardware data; Performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features; Based on the user status characteristics, weights are assigned to the applications, and a ranked application recommendation list is determined; A screen sidebar pops up, and the application recommendation list is displayed in the screen sidebar.

2. The application recommendation method according to claim 1, characterized in that: The performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features includes: Extracting a shared feature layer of the device status data; Determine at least one perspective attention of each dimension using multi-head self-attention; A target shared feature layer is screened from each of the shared feature layers through the perspective, and multi-scale feature fusion is performed on the target shared feature layer to obtain the user state feature.

3. The application recommendation method according to claim 1, characterized in that: The performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features includes: Using factor analysis method, extracting low-level features and high-level features from the device status data in multiple dimensions; Performing layer-by-layer splicing on the low-level features to form a first splicing feature, and performing layer-by-layer splicing on the high-level features to form a second splicing feature; The first splicing feature and the second splicing feature are fused to obtain the user state feature.

4. The application recommendation method according to claim 1, characterized in that: The step of allocating weights to the applications based on the user status characteristics and determining a ranked application recommendation list includes: Obtain historical behavior characteristics of similar users; Perform sliding window and attenuation function detection on the historical behavior characteristics and the user status characteristics to generate recommendation weights corresponding to the current applications; The application recommendation list is generated according to the recommendation weights.

5. The application recommendation method according to claim 4, characterized in that: The obtaining of historical behavior characteristics of similar users includes: Obtain historical behavior datasets; Perform collaborative filtering on the data in the historical behavior data set, and calculate similar user groups to obtain the historical behavior data of the similar users; Corresponding implicit features are extracted from the historical behavior data to obtain the historical behavior features.

6. The application recommendation method according to claim 1, characterized in that: The step of allocating weights to the applications based on the user status characteristics and determining a ranked application recommendation list comprises: In response to the number of operations of at least one target application exceeding a preset number within a preset time period, determining a target function of the target application from a pre-built function association database; matching relevant applications associated with the target function; Based on the target application and the related applications, the application recommendation list is updated, and other applications that have a low correlation with the target application are removed.

7. The application recommendation method according to claim 6, characterized in that: The process of constructing the functional association database includes: Classifying and describing the functions of each of the application programs to obtain function information; Based on the functional information, each of the application programs is cross-clustered to construct the functional association database.

8. An application recommendation device, characterized in that: The application recommendation device comprises: An acquisition module, configured to acquire device status data of multiple dimensions in response to a side sliding operation of a user; wherein the device status data includes at least one of scenario data, sensor data, preference data, social communication data, and hardware data; A fusion module, used for performing multi-dimensional feature fusion on the device status data of multiple dimensions to obtain user status features; An allocation module, configured to allocate weights to the applications based on the user status characteristics and determine a ranked application recommendation list; The display module is used to pop up a screen sidebar and display the application recommendation list in the screen sidebar.

9. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable program codes; A processor is used to call and run the executable program code from the memory, so that the electronic device executes the application recommendation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the application recommendation method according to any one of claims 1 to 7 is implemented.

Images