Data recommendation method, device, storage medium and electronic device

Abstract

The present application relates to a data recommendation method, device, storage medium and electronic device. The method includes: obtaining device signal data, object behavior data and a universal recommendation model of a target recommendation object, the latter of which is obtained by training based on historical data. By vectorizing and fusing the device signal data and the object behavior data, an object feature fusion vector is generated, and the universal model is trained with the vector to perform personalized parameter training on the universal model to generate a personalized recommendation model. The system inputs new device signal and behavior data into the personalized model at a second time to generate initial device recommendation data. If the target recommendation object does not agree with the recommendation result, the system optimizes the recommendation strategy based on the feedback and generates an optimized recommendation model. At a third time, new data is input into the optimization model to obtain device target recommendation data, which is used to help the target recommendation object set the parameters of the media equipment. The use of this method can improve the accuracy of data recommendation, thereby achieving the goal of setting accurate device parameters for the target recommendation object in professional scenarios.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a data recommendation method, device, storage medium and electronic device. Background Art

[0002] With the development of computer technology, artificial intelligence has entered a stage of rapid development, especially in the fields of data recommendation systems, natural language processing, computer vision, etc., where models based on neural networks have gradually become mainstream. However, the neural network models in traditional technologies have certain limitations when dealing with complex and diverse data. For example, in the recommendation system, factors such as the behavior, device information, and context of the target recommendation object jointly affect the accuracy of data recommendation. For some scenarios with high professional and technical requirements, the lack of accuracy of data recommendation can easily lead to the target recommendation object setting seriously deviated device parameters in professional scenarios. Summary of the invention

[0003] Based on this, it is necessary to provide a data recommendation method, device, storage medium, electronic device and computer program product that can improve the accuracy of data recommendation in response to the above technical problems, so as to achieve the goal of setting accurate device parameters for the target recommendation object in professional scenarios.

[0004] In a first aspect, the present application provides a data recommendation method, comprising:

[0005] Acquire the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time; the universal recommendation model is trained by the historical device signal data and historical object behavior data of the target recommendation object;

[0006] Vectorizing and fusing the device signal data and the object behavior data to obtain an object feature fusion vector;

[0007] According to the object feature fusion vector, training the personalized model parameters in the universal recommendation model to obtain a personalized recommendation model for the target recommendation object;

[0008] Inputting the device signal data and the object behavior data corresponding to the target recommendation object at the second time into the personalized recommendation model to obtain initial device recommendation data for the target recommendation object;

[0009] In response to a non-recognition instruction from the target recommendation object regarding the initial device recommendation data, optimizing the recommendation strategy of the personalized recommendation model according to the device signal data corresponding to the second time, the object behavior data, and the initial device recommendation data to obtain an optimized recommendation model;

[0010] The device signal data and object behavior data corresponding to the target recommendation object at the third time are input into the optimization recommendation model to obtain device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting parameters of the media equipment.

[0011] In a second aspect, the present application also provides a data recommendation device, comprising:

[0012] An object data acquisition module, used to acquire the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time; the universal recommendation model is trained by the historical device signal data and historical object behavior data of the target recommendation object;

[0013] An object data fusion module, used for vectorizing and fusing the device signal data and the object behavior data to obtain an object feature fusion vector;

[0014] A recommendation model optimization module, used to train the personalized model parameters in the universal recommendation model according to the object feature fusion vector, so as to obtain a personalized recommendation model for the target recommendation object;

[0015] A recommendation data obtaining module, used for inputting the device signal data and the object behavior data corresponding to the target recommendation object at the second time into the personalized recommendation model to obtain initial device recommendation data for the target recommendation object;

[0016] The recommendation model optimization module is further configured to optimize the recommendation strategy of the personalized recommendation model according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time in response to the target recommendation object's instruction of not agreeing with the initial device recommendation data, so as to obtain an optimized recommendation model;

[0017] The recommendation data acquisition module is also used to input the device signal data and object behavior data corresponding to the target recommendation object at the third time into the optimization recommendation model to obtain device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting parameters of the media equipment.

[0018] In a third aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0019] Acquire the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time; the universal recommendation model is trained by the historical device signal data and historical object behavior data of the target recommendation object;

[0020] Vectorizing and fusing the device signal data and the object behavior data to obtain an object feature fusion vector;

[0021] According to the object feature fusion vector, training the personalized model parameters in the universal recommendation model to obtain a personalized recommendation model for the target recommendation object;

[0022] Inputting the device signal data and the object behavior data corresponding to the target recommendation object at the second time into the personalized recommendation model to obtain initial device recommendation data for the target recommendation object;

[0023] In response to a non-recognition instruction from the target recommendation object regarding the initial device recommendation data, optimizing the recommendation strategy of the personalized recommendation model according to the device signal data corresponding to the second time, the object behavior data, and the initial device recommendation data to obtain an optimized recommendation model;

[0024] The device signal data and object behavior data corresponding to the target recommendation object at the third time are input into the optimization recommendation model to obtain device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting parameters of the media equipment.

[0025] In a fourth aspect, the present application further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0026] Acquire the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time; the universal recommendation model is trained by the historical device signal data and historical object behavior data of the target recommendation object;

[0027] Vectorizing and fusing the device signal data and the object behavior data to obtain an object feature fusion vector;

[0028] According to the object feature fusion vector, training the personalized model parameters in the universal recommendation model to obtain a personalized recommendation model for the target recommendation object;

[0029] Inputting the device signal data and the object behavior data corresponding to the target recommendation object at the second time into the personalized recommendation model to obtain initial device recommendation data for the target recommendation object;

[0030] In response to a non-recognition instruction from the target recommendation object regarding the initial device recommendation data, optimizing the recommendation strategy of the personalized recommendation model according to the device signal data corresponding to the second time, the object behavior data, and the initial device recommendation data to obtain an optimized recommendation model;

[0031] The device signal data and object behavior data corresponding to the target recommendation object at the third time are input into the optimization recommendation model to obtain device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting parameters of the media equipment.

[0032] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0033] Acquire the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time; the universal recommendation model is trained by the historical device signal data and historical object behavior data of the target recommendation object;

[0034] Vectorizing and fusing the device signal data and the object behavior data to obtain an object feature fusion vector;

[0035] According to the object feature fusion vector, training the personalized model parameters in the universal recommendation model to obtain a personalized recommendation model for the target recommendation object;

[0036] Inputting the device signal data and the object behavior data corresponding to the target recommendation object at the second time into the personalized recommendation model to obtain initial device recommendation data for the target recommendation object;

[0037] In response to a non-recognition instruction from the target recommendation object regarding the initial device recommendation data, optimizing the recommendation strategy of the personalized recommendation model according to the device signal data corresponding to the second time, the object behavior data, and the initial device recommendation data to obtain an optimized recommendation model;

[0038] The device signal data and object behavior data corresponding to the target recommendation object at the third time are input into the optimization recommendation model to obtain device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting parameters of the media equipment.

[0039] The above-mentioned data recommendation method, device, storage medium, electronic device and computer program product obtain the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time; the universal recommendation model is obtained by training the historical device signal data and historical object behavior data of the target recommendation object; the device signal data and the object behavior data are vectorized and fused to obtain the object feature fusion vector; according to the object feature fusion vector, the personalized model parameters in the universal recommendation model are trained to obtain the personalized recommendation model for the target recommendation object; the device signal data and the object behavior data corresponding to the target recommendation object at the second time are input into the personalized recommendation model to obtain the initial device recommendation data for the target recommendation object; in response to the target recommendation object's instruction of not agreeing with the initial device recommendation data, the recommendation strategy of the personalized recommendation model is optimized according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time to obtain the optimized recommendation model; the device signal data and the object behavior data corresponding to the target recommendation object at the third time are input into the optimized recommendation model to obtain the device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting the parameters of the media equipment.

[0040] By building a progressive recommendation system from a universal recommendation model to a personalized recommendation model and then to an optimized recommendation model, in the initial stage, the device signal data and the behavior data of the target recommendation object are integrated for feature vectorization processing to achieve preliminary training of the personalized model, making the recommendation results more targeted. When the target recommendation object disagrees with the initial recommendation result, the system further adjusts the recommendation strategy by receiving feedback, optimizes the personalized model, and enables it to adapt to the changes in the needs of the target recommendation object, improves the accuracy of data recommendation, and thus achieves the target recommendation object to set accurate equipment parameters in professional scenarios, avoids solidifying the data recommendation results, and more accurately predicts the target recommendation object's media equipment usage preferences in subsequent interactions, providing intelligent support for the setting of media equipment parameters, further improving the flexibility of the recommendation system and the satisfaction of the target recommendation object, while also enhancing the equipment usage efficiency and interactive effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 An application environment diagram of a data recommendation method in one embodiment;

[0043] Figure 2 A schematic diagram of a flow chart of a data recommendation method in one embodiment;

[0044] Figure 3 A schematic diagram of a flow chart of a method for obtaining an optimized recommendation model in one embodiment;

[0045] Figure 4 A schematic diagram of a process of analyzing trend information in one embodiment;

[0046] Figure 5 A schematic diagram of a process flow of a trend information analysis method in another embodiment;

[0047] Figure 6 A schematic diagram of a process for obtaining a personalized recommendation model in one embodiment;

[0048] Figure 7 A schematic diagram of a process for obtaining a personalized recommendation model in another embodiment;

[0049] Figure 8 A schematic diagram of a flow chart of a method for obtaining an object feature fusion vector in one embodiment;

[0050] Fig. 9 is a structural block diagram of a data recommendation device in one embodiment;

[0051] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0053] The data recommendation method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown, the media device 102 communicates with the server 104 through a network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. Server 104 obtains the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time; the universal recommendation model is obtained by training the historical device signal data and historical object behavior data of the target recommendation object; the device signal data and the object behavior data are vectorized and fused to obtain the object feature fusion vector; according to the object feature fusion vector, the personalized model parameters in the universal recommendation model are trained to obtain the personalized recommendation model for the target recommendation object; the device signal data and the object behavior data corresponding to the target recommendation object at the second time are input into the personalized recommendation model to obtain the initial device recommendation data for the target recommendation object; in response to the target recommendation object's instruction of not agreeing with the initial device recommendation data, the recommendation strategy of the personalized recommendation model is optimized according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time to obtain the optimized recommendation model; the device signal data and the object behavior data corresponding to the target recommendation object at the third time are input into the optimized recommendation model to obtain the device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting the parameters of the media equipment. The media device 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0054] In an exemplary embodiment, Figure 2 As shown, a data recommendation method is provided, which is applied to Figure 1 The server in the example is used to illustrate, including the following steps 202 to 212. Among them:

[0055] Step 202: Obtain device signal data, object behavior data, and a universal recommendation model corresponding to the target recommendation object at the first time.

[0056] The target recommendation object may be the subject for which the system makes personalized recommendations, that is, the target recommendation object or device operator currently identified as needing recommended media device operation or configuration suggestions.

[0057] Among them, the device signal data can be various parameters from wireless communication during the operation of the media device, which introduces physical layer signal characteristics and enhances the ability to perceive the status of the target recommendation object.

[0058] Among them, the object behavior data can be information such as the operating habits, usage patterns, interaction methods, shooting behaviors, etc. exhibited by the target recommended object when using the device, usually including the target recommended object's operating instructions for the device, usage frequency, preference settings, etc.

[0059] Among them, the universal recommendation model can be a general recommendation model trained by analyzing the historical device signal data and behavior data of a large number of target recommendation objects, which is suitable for most similar objects.

[0060] The first time, the second time and the third time are different times for the sensor to collect device signal data and object behavior data, respectively. In the order of the time axis, the first time precedes the second time, and the second time precedes the third time.

[0061] Among them, media devices can be electronic devices used to obtain, generate, disseminate, transmit and display information, and are widely used in various media forms such as radio, television, the Internet, billboards, public display screens, etc. They can include televisions, projectors, digital billboards, audio equipment, virtual reality equipment, mobile terminals, etc., and their main function is to present information, content or advertisements to the audience in a visual, auditory or other sensory way. Media devices not only support the display of content, but also can interact with target recommendation objects.

[0062] Specifically, at the first time node, the system obtains the real-time signal data of the device from the device of the target recommendation object, including but not limited to the current operating status of the device, sensor data, operation log, etc., and collects the behavior data of the target recommendation object, such as interaction habits, operation methods, usage frequency and other variables in the device environment (temperature, pressure, etc.). These data are captured in real time through the sensor network or data acquisition module. At the same time, the system loads a previously trained universal recommendation model, which is trained based on the historical device signal data and historical behavior data of other objects with similar device usage characteristics as the target recommendation object, and can provide basic behavior patterns and device operation recommendations for the recommendation process. Among them, the real-time signal data is the physical layer signal of the media device, which can come from various parameters of wireless communication, such as channel state information (CSI), received signal strength indicator (RSSI), angle of arrival (AoA), time delay (ToF) between the receiving end and the sending end, etc. Usually, devices that support communication protocols such as Wi-Fi, 5G, LoRa, etc. are used to capture signals through network cards, wireless communication modules, etc. For example, CSI data can be collected in real time through CSI tools or APIs, and RSSI can be read directly from the wireless communication module. Signal data should be collected at a high frequency (thousands of times per second) to ensure real-time reflection of the communication status. The actual collection frequency can be adjusted according to the application scenario. Among them, the object behavior data can be the operation record of the target recommended object on the media equipment during the interview process using media equipment or in the scene of live broadcast using drones (such as adjusting the temperature, starting certain devices, etc.), or it can be combined with GPS or Wi-Fi positioning and other technologies to obtain the geographic location information of the target recommended object.

[0063] Step 204 , vectorize and fuse the device signal data and the object behavior data to obtain an object feature fusion vector.

[0064] Among them, vectorized fusion can be the process of processing device signal data and object behavior data, converting them into a unified numerical representation, and merging them.

[0065] Among them, the object feature fusion vector can be a high-dimensional vector generated by vectorized fusion of device signal data and object behavior data, which comprehensively reflects the individual characteristics of the target object.

[0066] Specifically, based on the preprocessed data, the system uses specific vectorization techniques (such as embedding vectors and feature extraction algorithms) to convert device signal data and object behavior data into numerical vectors respectively. Then, the system uses fusion algorithms (such as weighted average or attention mechanism) to merge these numerical vectors into an object feature fusion vector. The object feature fusion vector not only retains the main information of the device status and object behavior, but also reflects the comprehensive characteristics of the target recommended object when using the device by combining the two types of data.

[0067] Step 206 , training the personalized model parameters in the universal recommendation model according to the object feature fusion vector, to obtain a personalized recommendation model for the target recommendation object.

[0068] The personalized model parameters may be model parameters adjusted according to differences in individual shooting and operation of the target recommendation object.

[0069] Among them, the personalized recommendation model can be a model trained on the basis of the universal recommendation model, combining the device signal data, object behavior data and personalized parameters of the target recommendation object. It can generate device recommendation data specifically for a certain object to meet personalized needs.

[0070] Specifically, after obtaining the object feature fusion vector of the target recommendation object, the system uses it as input and combines it with the personalized parameters that already exist in the universal recommendation model to conduct further model training. This training process can use machine learning methods such as reinforcement learning and supervised learning to gradually adjust the weights, parameter values, and parameter representation of the personalized model parameters in the model so that it can better capture the personalized preferences of the target recommendation object. The goal of the training process is to maximize the model's matching degree with the device usage pattern of the current target recommendation object, so that the model can output personalized device recommendations that are more in line with the needs of the target recommendation object and obtain a personalized recommendation model. The system verifies the accuracy of the model by reviewing historical data, and performs cross-validation as needed to prevent overfitting.

[0071] Step 208 , input the device signal data and the object behavior data corresponding to the target recommended object at the second time into the personalized recommendation model to obtain initial device recommendation data for the target recommended object.

[0072] The initial device recommendation data may be the first set of recommendation results generated based on the personalized recommendation model. These data include the system's appropriate parameter settings or operation suggestions for the target recommendation object when currently using the device, as the preliminary output of the recommendation.

[0073] Specifically, at the second time node, the system collects the device signal data and object behavior data of the target recommendation object at the current time again, and inputs these data into the recommendation model that has been personalized trained. After integrating the personalized characteristics of the target recommendation object in the early stage, the model can generate a set of initial device recommendation data based on the real-time data at the current time point. These recommendation data contain the optimal operating parameters or setting options for the target recommendation object when the system currently uses the media device, such as the device's operating mode, power consumption setting, parameter adjustment, etc. This process is completed within a millisecond response time to ensure that the recommendation can be synchronized with the actual needs of the target recommendation object.

[0074] Step 210, in response to the target recommendation object's instruction of not agreeing with the initial device recommendation data, the recommendation strategy of the personalized recommendation model is optimized according to the device signal data corresponding to the second time, the object behavior data and the initial device recommendation data to obtain an optimized recommendation model.

[0075] Among them, the recommendation strategy can be the model decision rules and optimization methods that the recommendation system relies on when generating device recommendation data.

[0076] Among them, the optimized recommendation model can be a recommendation model that is updated after the system continuously receives feedback from the target recommendation object and adjusts the recommendation strategy based on real-time data. This model can more accurately reflect the needs and preferences of the target object and provide more personalized and accurate device recommendations.

[0077] Specifically, if the target recommendation object is not satisfied with the initial device recommendation data generated by the system, the system will detect the signal of disapproval by capturing the feedback instructions of the target recommendation object. The system will further optimize the model twice according to the operation instructions that the target recommendation object disagrees with, wherein the device signal data, object behavior data and initial device recommendation data corresponding to the second time are input into the personalized recommendation model together, and the weights and parameters of the recommendation strategy are adjusted by using feedback reinforcement learning or gradient descent-based optimization algorithms. The core of this process is to re-evaluate the loss function of the model through the feedback of the target recommendation object, gradually optimize the parameters of the recommendation strategy of the model, and obtain the optimized recommendation model, so that it can better meet the personalized needs of the target recommendation object in future recommendations, and improve the accuracy of the recommendation and the satisfaction of the target recommendation object.

[0078] Step 212: input the device signal data and the object behavior data corresponding to the target recommendation object at the third time into the optimization recommendation model to obtain device target recommendation data for the target recommendation object.

[0079] The device target recommendation data may be the final device recommendation result generated by the optimized recommendation model. It includes optimized personalized device parameter settings, aiming to provide more accurate device usage suggestions for the target recommendation object and improve the experience of the target recommendation object.

[0080] Specifically, at the third time point, the system again obtains the device signal data and object behavior data of the target recommendation object, and inputs these data into the personalized recommendation model that has been optimized through feedback. Since the model has combined the feedback of the target recommendation object and readjusted the recommendation strategy, the device target recommendation data generated by the system will be more accurate. These data are not only based on the current device status and object behavior, but also through the model optimization process, they are more in line with the actual usage preferences and device operation habits of the target recommendation object. These target recommendation data can provide personalized device parameter setting suggestions for the target recommendation object, help them configure media device parameters more efficiently in complex device operation environments, and greatly improve the intelligent operation experience of the device.

[0081] In the above-mentioned data recommendation method, the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time are obtained; the universal recommendation model is obtained by training the historical device signal data and historical object behavior data of the target recommendation object; the device signal data and the object behavior data are vectorized and fused to obtain the object feature fusion vector; according to the object feature fusion vector, the personalized model parameters in the universal recommendation model are trained to obtain the personalized recommendation model for the target recommendation object; the device signal data and the object behavior data corresponding to the target recommendation object at the second time are input into the personalized recommendation model to obtain the initial device recommendation data for the target recommendation object; in response to the target recommendation object's non-acceptance instruction for the initial device recommendation data, the recommendation strategy of the personalized recommendation model is optimized according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time to obtain the optimized recommendation model; the device signal data and the object behavior data corresponding to the target recommendation object at the third time are input into the optimized recommendation model to obtain the device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting the parameters of the media equipment.

[0082] By building a progressive recommendation system from a universal recommendation model to a personalized recommendation model and then to an optimized recommendation model, in the initial stage, the device signal data and the behavior data of the target recommendation object are integrated for feature vectorization processing to achieve preliminary training of the personalized model, making the recommendation results more targeted. When the target recommendation object disagrees with the initial recommendation result, the system further adjusts the recommendation strategy by receiving feedback, optimizes the personalized model, and enables it to adapt to the changes in the needs of the target recommendation object, improves the accuracy of data recommendation, and thus achieves the target recommendation object to set accurate equipment parameters in professional scenarios, avoids solidifying the data recommendation results, and more accurately predicts the target recommendation object's media equipment usage preferences in subsequent interactions, providing intelligent support for the setting of media equipment parameters, further improving the flexibility of the recommendation system and the satisfaction of the target recommendation object, while also enhancing the equipment usage efficiency and interactive effect.

[0083] In an exemplary embodiment, Figure 3 As shown, according to the device signal data, object behavior data and initial device recommendation data corresponding to the second time, the recommendation strategy of the personalized recommendation model is optimized to obtain an optimized recommendation model, including steps 302 to 306. Among them:

[0084] Step 302: identifying dynamic demand information of a target recommended object according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time.

[0085] The dynamic demand information may be the immediate demand of the target recommendation object which is generated as time or environment changes during the process of using the media device.

[0086] Specifically, at the second time point, the system collects the real-time signal data of the device and the current behavior data of the target object, and combines it with the initial device recommendation data to dynamically analyze the current demand changes of the target recommendation object. This process compares the difference between current data and historical data to identify the specific needs of the target object at the moment and generate dynamic demand information, such as changes in operating habits and changes in equipment usage conditions, to ensure that the recommendation strategy can respond to the latest needs of the object in a timely manner.

[0087] Step 304: Analyze the behavior trend information and signal trend information of the target recommendation object according to the dynamic demand information and the initial device recommendation data.

[0088] The behavior trend information may be the possibility of behavior changes of the target recommendation object caused by operating the device in the current demand of the target recommendation object.

[0089] The signal trend information may be obtained by analyzing the signal data during the operation of the device, and may be the possibility of change of the signal information in a future period of time.

[0090] Specifically, the system extracts features from the dynamic demand information of the target recommendation object and identifies the information with the highest relevance to the target recommendation object’s needs. It combines the initial device recommendation data and uses the data trend analysis algorithm to mine the behavior change trend of the target recommendation object and the device signal change trend. This may include the change pattern of the device signal over time and the change of the target recommendation object’s operating habits.

[0091] Step 306, optimizing the recommendation strategy of the personalized recommendation model according to the dynamic demand information, the behavior trend information and the signal trend information to obtain an optimized recommendation model.

[0092] Specifically, based on the dynamic demand information, behavior trend information, and signal trend information obtained through analysis, the system combines the existing personalized recommendation model and updates or adjusts the parameters of the recommendation strategy part of the model through algorithms such as gradient descent or Bayesian optimization. It can also modify the priority in the recommendation strategy based on the feedback of the target recommendation object, such as adjusting the recommended parameter range or setting a recommendation rule that is more suitable for the current target recommendation object behavior; if necessary, it can reset the implementation logic of the parameters or recommendation strategy to improve the model's responsiveness to the personalized needs of the target object, so that the recommendation strategy can better adapt to the object's behavior and the changing trend of the device signal, and finally generate an optimized recommendation model. This optimized recommendation model can more accurately reflect the actual needs of the target object and provide device parameter recommendations that are more in line with the preferences of the target recommendation object in future recommendations.

[0093] In this embodiment, by dynamically collecting device signal data, object behavior data, and initial recommendation results, the real-time needs of the target recommendation object are accurately identified, and its behavior trend and signal trend information are further analyzed. Through the fusion of these multi-dimensional data, the system can dynamically adjust and optimize the recommendation strategy of the personalized recommendation model to ensure that the recommendation results can adapt to the changes in the needs of the target recommendation object in real time, and provide more accurate and personalized device recommendations. This not only improves the response efficiency and accuracy of the recommendation system, but also significantly enhances the experience of the target recommendation object, making the recommendation results more targeted and real-time.

[0094] In an exemplary embodiment, Figure 4 As shown, according to the dynamic demand information and the initial device recommendation data, the behavior trend information and the signal trend information of the target recommendation object are analyzed, including steps 402 to 406. Among them:

[0095] Step 402: Identify actual scene data of the target recommendation object based on the dynamic demand information.

[0096] The actual scene data may be a specific scene of the current location of the target recommendation object, such as spatial information, cultural information, and environmental information of the target recommendation object.

[0097] Specifically, after obtaining the dynamic demand information of the target recommendation object, the system combines the current operating status of the device, sensor data (such as temperature, humidity, light, etc.), operating environment information (such as indoor / outdoor, time period) and the current operating behavior of the target recommendation object (such as frequently adjusting a parameter or turning on a certain mode) to obtain analysis results. This process involves the integration and processing of multi-source data, and uses scene recognition algorithms (such as context-based scene classification or correlation analysis between device status and external environment) to determine the specific usage scenario of the target recommendation object. The actual scene data not only includes physical environment data, but also reflects the current task or goal of the target recommendation object (such as home use, work scene, entertainment mode, etc.), and finally establishes the actual scene data under the current scene.

[0098] Step 404: Analyze the scene prediction trajectory of the target recommendation object based on the actual scene data and the initial device recommendation data.

[0099] The scenario prediction trajectory can be a future behavior prediction path generated based on the historical operation mode of the target recommendation object, the current actual scenario data, and the device recommendation data. It reflects the possible operation behavior evolution and device status change trend of the target recommendation object in a specific scenario.

[0100] Specifically, by combining the actual scenario data and the initial device recommendation data, the system will use a prediction model (such as a machine learning model or time series model based on historical data) to predict the possible future operation trajectory of the target recommendation object. Specifically, based on the actual scenario data and the initial device recommendation data, the operation mode of the target recommendation object in similar scenarios is analyzed, the operation or movement rules in the historical behavior are extracted, and the device usage characteristics of the current scenario are combined to infer how the target recommendation object may operate the device or mobile device. This process will also take into account the difference between the initial recommendation data and the feedback from the target recommendation object, and infer whether the target recommendation object will continue to accept the parameters recommended by the system to move during the movement of the device, or will tend to move according to the wishes of the target recommendation object, and finally obtain the scene prediction trajectory. The scene prediction trajectory depicts the possible operation path of the target object in a certain period of time in the future, helping the system to adjust the device recommendations in advance before recommendation.

[0101] Step 406, calculating behavior trend information and signal trend information based on dynamic demand information, actual scenario data, and scenario prediction trajectory.

[0102] Specifically, the system analyzes the behavior pattern of the target recommendation object in a specific environment based on the dynamic demand information, actual scene data, and scene prediction trajectory of the target recommendation object. For example, the system will comprehensively analyze the target recommendation object's operation frequency, habitual operations (such as regular adjustment of device parameters), and operation preferences under scene changes, extract the target recommendation object's possible future behavior trends, and obtain behavior trend information. At the same time, the system will track the device's historical signal data, infer the device's signal fluctuations in future operations based on the scene prediction trajectory, and calculate the signal change trend of the media device during operation to obtain signal trend information. Through these calculations, the system can effectively predict the target recommendation object's future operation preferences and device status changes, providing a basis for further optimization of the personalized recommendation model.

[0103] In this embodiment, the actual scenario data of the target recommendation object is accurately identified through dynamic demand information, and the scene prediction trajectory of the object is deeply analyzed in combination with the initial device recommendation data. Through these steps, the system can effectively predict the behavior trend and device signal change trend of the target recommendation object, and predict the operation preference and device status of the target recommendation object in advance. This prediction capability enables the recommendation system to more intelligently adapt to the actual demand changes of the target recommendation object, provide more accurate and timely recommendations, and significantly improve the accuracy, personalization and satisfaction of the device recommendation object.

[0104] In an exemplary embodiment, Figure 5 As shown, according to the dynamic demand information, the actual scene data and the scene prediction trajectory, the behavior trend information and the signal trend information are calculated, including steps 502 to 506. Among them:

[0105] Step 502: Generate a physical state prediction space of several target recommendation objects on the scene prediction trajectory according to the dynamic demand information and the actual scene data.

[0106] Among them, the physical state prediction space can be a three-dimensional space that predicts the target recommendation object may be in at different time points or operation scenarios in the future based on the dynamic demand information and actual scene data of the target recommendation object.

[0107] Specifically, the system generates multiple physical state prediction spaces based on the dynamic demand information and actual scenario data of the target recommendation object, with the possible operation trajectory (scenario prediction trajectory) of the target recommendation object as the axis, and selects a certain three-dimensional space area near the axis at a set time interval. Each physical state prediction space represents the media device usage status and operating environment of the target recommendation object at different future moments or operating conditions, covering the operating status of the device (such as power, temperature, speed) and external environmental data (such as temperature, humidity), and provides multiple possible future device states and operating scenarios.

[0108] Step 504 , for any physical state prediction space, using the historical physical state data of the target recommended object in the historical physical state space as reference data, and generating the physical state prediction data and physical state prediction error of the target recommended object according to the initial device recommendation data.

[0109] The historical physical state space may be a three-dimensional space used by a collection of all physical states experienced by a device in different scenarios and times during past use of the target recommendation object.

[0110] Among them, the historical physical state data can be specific data points collected from the past device usage process of the target recommendation object, reflecting the movement state of the target recommendation object in a specific time and scenario.

[0111] Among them, the physical state prediction data can be the motion state of the predicted target recommendation object at a specific time and scene in the future.

[0112] The physical state prediction error may be a prediction error of the motion state of the target recommendation object at a specific time and scene in the future.

[0113] Specifically, for each physical state prediction space, the system uses the historical physical state data of the target recommended object as a reference, combined with the current initial device recommendation data, to predict the physical state changes of the target recommended object at future moments. This includes establishing a physical state model (used to calculate the physical state of the target recommended object, such as the relationship between action, orientation and time) through historical physical state data, and inputting the current environment, device usage information and initial device recommendation data into the model to generate corresponding physical state prediction data, such as future device power consumption, temperature, etc. At the same time, the system calculates the credibility of the system prediction through the difference between the physical state prediction data and the historical physical state data, and further obtains the physical state prediction error based on the credibility.

[0114] Step 506: Fit each physical state prediction data and each physical state prediction error to generate behavior trend information and signal trend information.

[0115] Specifically, the system takes the physical state prediction data and its corresponding prediction errors in multiple physical state prediction spaces as input, and performs fitting using regression analysis, neural network fitting or other machine learning algorithms. Through fitting analysis, the system can generate smooth and continuous activity information of the physical state of the target recommendation object in each physical state prediction space. Furthermore, due to the direct interaction between the media device and the target recommendation object, the system can extract the behavioral trend information of the media device and the signal trend information of the device from the smooth and continuous activity information of the target recommendation object. The behavioral trend information predicts the future evolution of the operation mode of the target recommendation object, such as the frequency of use and operation preferences of the target recommendation object in different scenarios; while the signal trend information reflects how the operating state of the media device changes over time or usage.

[0116] In this embodiment, multiple physical state prediction spaces are generated through dynamic demand information and actual scenario data, and the physical state prediction data and prediction errors are calculated using historical physical state data and initial device recommendation data. By fitting these data, the system can accurately generate behavior trend information and signal trend information. This process enables the system to predict the changing trends of the target recommended object's operation and device status in advance, thereby providing more accurate, dynamic and personalized device recommendations in complex scenarios, improving the prediction accuracy of recommendations, reducing errors, and enhancing the real-time and intelligent experience of the target recommended object.

[0117] In an exemplary embodiment, Figure 6 As shown, according to the object feature fusion vector, the personalized model parameters in the universal recommendation model are trained to obtain a personalized recommendation model for the target recommendation object, including steps 602 to 606. Among them:

[0118] Step 602: Calculate the association degree between the target recommended object and the media device in the universal recommendation model according to the object feature fusion vector to obtain object device association information.

[0119] Among them, the object device association information can be the association data obtained by analyzing the matching degree between the object feature fusion vector of the target recommended object and the functional characteristics of the device. This information reflects the usage habits, needs and adaptability of the target recommended object to a specific device, helping the system evaluate the interaction effect between the target recommended object and the device, thereby generating more accurate personalized recommendation results.

[0120] Specifically, the system inputs the object feature fusion vector of the target recommended object into the universal recommendation model to obtain the recommendation data of the universal recommendation model, and based on the recommendation data, evaluates the degree of match between the characteristics of the target recommended object and the functions of the media device through association analysis methods (such as cosine similarity, matrix decomposition or deep neural network). This association analysis not only examines the current media device usage of the target recommended object, but also considers historical interaction data and device characteristics, and derives the adaptability of the target recommended object and the device in different usage scenarios, generating "object device association information", which is an important reference for evaluating the target recommended object's needs and the accuracy of device recommendation.

[0121] Step 604: Use the behavior signal coordination model to calculate the coordination effect of the object feature fusion vector at different time scales to obtain behavior signal coordination data.

[0122] Among them, the behavior signal synergy model can be a model for analyzing the synergistic effect of the target recommendation object behavior and device signals at different time scales. By comparing the relationship between the target recommendation object's operation mode and device response in the short and long term, the model reveals the time dimension characteristics of the target recommendation object's behavior and the way it interacts with the device status, helping the system to better understand the synergy between the target recommendation object's operation habits and the device.

[0123] Among them, the behavior signal synergy data can be data calculated by a behavior signal synergy model, reflecting the synergy effect between the behavior of the target recommendation object and the device signal at different time scales.

[0124] Specifically, the system further uses the "behavior signal synergy model" to analyze the synergy effect of the object feature fusion vector of the target recommendation object at different time scales. The model examines the relationship between short-term behavior (such as the target recommendation object's recent device usage pattern) and long-term behavior (such as the target recommendation object's long-term device usage trend) by processing the target recommendation object's behavior changes and device signal reactions in different time periods. Through algorithms such as time series analysis or collaborative filtering, the system captures the synchronization and consistency of the target recommendation object's behavior signals in these different time dimensions, quantifies the synergy effect between the target recommendation object's operation behavior and device signals, and generates behavior signal synergy data. This data reflects how the target recommendation object interacts with the device in different time periods, helping the system to further identify the target recommendation object's preferences and habits.

[0125] Step 606 , optimizing the personalized model parameters in the universal recommendation model according to the object device association information and the behavior signal collaborative data, to obtain a personalized recommendation model.

[0126] Specifically, the system inputs the object device association information and behavior signal collaborative data into the universal recommendation model to optimize the parameters of the personalized model. By repeatedly adjusting the model's weights, parameters, and influencing factors of behavioral signals (such as weight settings for behaviors at different time scales), the system uses machine learning optimization algorithms (such as gradient descent or reinforcement learning) to fine-tune the model to make it more adaptable to the specific needs and preferences of the target recommendation object. After this optimization process, the system is able to generate a personalized recommendation model. This model not only inherits the basic recommendation capabilities of the universal model, but also makes the recommendation results more accurate through personalized adjustments, and can provide the target recommendation object with device usage suggestions that are highly matched with its actual needs.

[0127] In this embodiment, by calculating the object feature fusion vector, the association degree between the target recommended object and the media device is accurately evaluated, the object device association information is generated, and the behavior signal coordination model is combined to analyze the behavior coordination effect of the target recommended object at different time scales to obtain the behavior signal coordination data. Through the optimization of this information, the system can adjust the personalized parameters of the universal recommendation model, and finally generate a personalized recommendation model for the individual needs of the target recommended object. This process enhances the depth of the recommendation model's understanding of the needs of the target recommended object, making the recommendation results more personalized and dynamically adaptable, and significantly enhancing the accuracy of the recommendation and the satisfaction of the target recommended object.

[0128] In an exemplary embodiment, Figure 7 As shown, according to the object device association information and the behavior signal collaborative data, the personalized model parameters in the universal recommendation model are optimized to obtain the initial personalized recommendation model, including steps 702 to 708. Among them:

[0129] Step 702: Determine a personalized adjustment strategy for the universal recommendation model according to the object device association information.

[0130] Among them, the personalized adjustment strategy can be the adjustment direction and method formulated by the system to optimize the recommendation model based on the object device association information of the target recommendation object. This strategy mainly focuses on the adaptability between the target recommendation object and the device, and determines how to adjust the parameters of the model according to the needs and preferences of the target recommendation object to ensure that the device function recommendation matches the actual usage of the target recommendation object. For example, the strategy may increase the weight of the device functions frequently used by the target recommendation object to improve the relevance of the recommendation.

[0131] Specifically, since the association information provides the target recommendation object's preferences for device functions, usage frequency, and operating modes in different scenarios, the object device association information can be used to analyze the interaction characteristics and degree of adaptation between the target recommendation object and the media device. Based on the analysis results, the system formulates a personalized adjustment strategy, the goal is to optimize the relevant parameters in the model so that the recommendation model is more suitable for the target recommendation object's usage habits and device requirements. Among them, the personalized adjustment strategy includes selecting which device characteristics or functions should receive more attention (such as increasing the weight of a certain function or giving priority to specific parameters), so that the model's recommendation results are more in line with the actual operating scenarios of the target recommendation object.

[0132] Step 704: Determine personalized adjustment content for the universal recommendation model based on the behavioral signal collaborative data.

[0133] Among them, the personalized adjustment content can be determined based on the behavioral signal and collaborative data of the target recommendation object, reflecting how the system adjusts the recommendation model at different time scales based on the operation behavior of the target recommendation object. The adjustment content mainly focuses on the dynamic changes in the behavior of the target recommendation object, and determines how the model sensitively responds to the short-term and long-term operation habits of the target recommendation object, so as to ensure that the recommendation can be updated in time with the changes in the behavior pattern of the target recommendation object.

[0134] Specifically, since the behavioral signal synergy data reflects the target recommendation object's operating mode and the response of the device signal at different time scales, the short-term and long-term synergy effects of the target recommendation object's operating behavior and the device signal are analyzed based on the behavioral signal synergy data to help the system determine the personalized adjustment content. These adjustments include how the model should respond when dealing with short-term behavioral changes (such as quickly adjusting device settings) and long-term usage patterns (such as gradually preferring a certain type of function) of the target recommendation object. This process ensures that the model can flexibly adjust the recommended content according to the dynamic operation of the target recommendation object through behavioral pattern modeling, making the recommendation results more timely and flexible.

[0135] Step 706: When the personalized adjustment strategy and the personalized adjustment content do not match, the personalized model parameters in the universal recommendation model are adjusted using the personalized adjustment strategy and the personalized adjustment content to obtain a first adjustment model and a second adjustment model.

[0136] Among them, the first adjustment model is the model generated after adjusting the universal recommendation model according to the personalized adjustment strategy. This model focuses on adjusting the model parameters to improve the adaptability of device function recommendations by optimizing the matching degree between the target recommendation object and the device, so that the recommendation results are more in line with the preferences and needs of the target recommendation object, and the focus is on the matching of device functions with the needs of the target recommendation object. The second adjustment model is the model generated after the system adjusts the universal recommendation model according to the personalized adjustment content. This model focuses on adjusting the dynamic response capability of the model so that it can sensitively adapt to changes in the behavior of the target recommendation object, whether it is short-term operational adjustments or long-term usage trends, thereby improving the real-time and flexibility of the recommendation system.

[0137] Specifically, if the system finds that there is an inconsistency between the personalized adjustment strategy and the personalized adjustment content (that is, there is a deviation between the personalized adjustment strategy and the personalized adjustment content in the target direction of adjusting the universal recommendation model), it means that the personalized adjustment strategy of the universal recommendation model (based on the object device association information) and the actual behavior change response mechanism (based on the behavior signal collaborative data) have different focuses. At this time, the system will apply these two adjustment paths respectively to generate two independent adjustment models. First, the personalized adjustment strategy focuses on optimizing the matching degree between the target recommendation object and the device, generating the first adjustment model to ensure that the device function recommendation is highly adapted to the needs of the target recommendation object. Secondly, the personalized adjustment content focuses on adjusting the response sensitivity of the universal recommendation model to the behavior of the target recommendation object, generating a second adjustment model, so that the model can quickly adapt to the short-term changes and long-term trends of the behavior of the target recommendation object.

[0138] Step 708: Perform loss-level fusion on the first adjustment model and the second adjustment model to obtain a personalized recommendation model.

[0139] Specifically, the system merges the results of the first adjustment model and the second adjustment model through the loss-level fusion method. Loss-level fusion is a fusion technology based on the model loss function. It calculates the error (loss value) of each model in the prediction and performs weighted average or other fusion operations on these loss values ​​to generate a final personalized recommendation model. This fusion process ensures that the system can maintain a high degree of adaptability between the target recommendation object and the device function, and can sensitively respond to changes in the operational behavior of the target recommendation object, and finally generate a balanced recommendation model to provide more accurate and personalized device recommendation results.

[0140] In this embodiment, the personalized adjustment strategy is determined by the object device association information, and the personalized adjustment content is formulated through the behavioral signal collaborative data. When the strategy and the adjustment content are inconsistent, the system adjusts the model parameters separately, generates two independent optimization models, and merges the two into the final personalized recommendation model through loss-level fusion. This method ensures that the recommendation model can not only accurately adapt to the needs of the target recommendation object, but also flexibly respond to the dynamic changes in the behavior of the target recommendation object, thereby improving the overall accuracy and adaptability of the model, and ultimately providing more accurate and personalized recommendation results for the target recommendation object.

[0141] In an exemplary embodiment, Figure 8 As shown, the device signal data and the object behavior data are vectorized and fused to obtain the object feature fusion vector, including steps 802 to 808. Among them:

[0142] Step 802: Perform spatial channel mapping processing on the device signal data to obtain signal behavior spatial data.

[0143] Among them, spatial channel mapping can be a signal processing technology that visualizes and quantifies the propagation path, intensity variation, interference and other physical characteristics of device signals in physical space.

[0144] Among them, the signal behavior spatial data can be the manifestation of the device signal generated by spatial channel mapping in the physical space, reflecting the strength, change and propagation of the signal at different spatial positions.

[0145] Specifically, the system uses spatial channel mapping technology to convert these signal data into spatial distribution. Spatial channel mapping is a signal processing method that visualizes and quantifies the propagation path, interference, and signal changes of device signals in physical space. Through this process, the system generates signal behavior spatial data that reflects the changes of device signals with spatial positions. These data provide important references for the performance of devices in different locations or environments.

[0146] Step 804: Perform behavior-environment linkage analysis on the object behavior data to obtain object behavior-environment relationship data.

[0147] Among them, behavior-environment linkage analysis can be a method for studying the relationship between the behavior data of the target object and the operating environment. By analyzing the correlation between the operating habits of the target recommendation object and environmental factors (such as temperature, humidity, light, etc.), the system can identify the pattern of the target recommendation object's behavior being affected by changes in the external environment, thereby optimizing the understanding of the target recommendation object's needs.

[0148] Among them, the object behavior environment relationship data can be generated through behavior environment linkage analysis, and is used to describe the relationship between the behavior of the target object and its operating environment. This data shows how environmental conditions (such as temperature, humidity, etc.) affect the way the target recommended object operates the device, providing a more accurate basis for the system to recommend device functions.

[0149] Specifically, the behavior-environment linkage analysis is performed in combination with the object behavior data corresponding to the target recommendation object and the data of the environment in which the device is located (such as external temperature, humidity, brightness, etc.). The environmental linkage analysis reveals how environmental conditions affect the operating habits of the target recommendation object by identifying the correlation between the behavior of the target recommendation object and environmental changes. For example, the target recommendation object may operate the device differently at different temperatures, or change the settings of the device under different lighting conditions. Through this analysis, the system generates object behavior-environment relationship data, which describes the complex interaction between the behavior of the target recommendation object and environmental conditions, providing a basis for further recommendation optimization.

[0150] Step 806, embed the signal behavior space data and the object behavior environment relationship data into the time-space domain respectively to obtain the signal time-space embedding vector and the behavior environment time-space embedding vector.

[0151] Among them, the signal spatiotemporal embedding vector can be a high-dimensional feature vector formed by embedding the signal behavior spatial data into the time and space dimensions, which captures the dynamic changes of the device signal at different time points and spatial positions.

[0152] Among them, the environmental spatiotemporal embedding vector can be a feature vector generated by embedding the object behavior environment relationship data into the time and space dimensions, which describes the dynamic changes of the target recommendation object behavior and the environment interaction in different time and space. This vector reflects how the target recommendation object operation behavior changes with the changes in time and environment.

[0153] Specifically, the system uses spatiotemporal embedding technology to embed signal behavior space data and object behavior environment relationship data into the spatiotemporal domain respectively. The spatiotemporal embedding technology captures the dynamic changes of device signals and target recommended object behaviors by tracking the evolution of data in time and space dimensions. For example, the signal spatiotemporal embedding vector generates a feature representation of device signals changing with time and space (signal spatiotemporal embedding vector) by analyzing the changing trends of signal data at different times and spatial locations. The behavior environment spatiotemporal embedding vector forms a spatiotemporal feature representation of the interaction between behavior and environment (environmental spatiotemporal embedding vector) by analyzing the changes in the behavior of the target recommended object under different environmental conditions and time points. These two embedding vectors reflect the comprehensive evolution of device signals and target recommended object behaviors over time and space, enhancing the understanding of the spatiotemporal dimension of data.

[0154] Step 808, using the cross-modal attention mechanism, feature fusion is performed on the signal spatiotemporal embedding vector and the behavior environment spatiotemporal embedding vector to obtain an object feature fusion vector.

[0155] Among them, the cross-modal attention mechanism can be a technology used to fuse data of different modalities (such as signal data and behavioral data). By identifying and strengthening the key connections between the modalities, the mechanism dynamically weights the importance of different modalities to ensure that the system can focus on the features that are most critical to the task.

[0156] Specifically, since the cross-modal attention mechanism is a technology used to process multimodal data, it can identify and emphasize important connections between different modalities. The system first assigns different weights to the signal spatiotemporal embedding vector and the behavior environment spatiotemporal embedding vector, focusing on the most critical features for the recommendation task. Then, through the weighted calculation of the attention mechanism, these two types of spatiotemporal features are effectively fused to finally generate an object feature fusion vector. This fusion vector combines the spatiotemporal features of the target recommended object behavior and the device signal, providing accurate input for the personalized recommendation model and improving the recommendation system's perception of the target recommended object's preferences and device status.

[0157] In this embodiment, by performing spatial channel mapping on device signal data and behavioral environment linkage analysis on object behavior data, signal behavior space data and object behavior environment relationship data are generated respectively, and embedded into the spatiotemporal domain to capture the dynamic changes of data in time and space. Subsequently, the system performs feature fusion on the signal spatiotemporal embedding vector and the behavioral environment spatiotemporal embedding vector through a cross-modal attention mechanism to generate an object feature fusion vector. This multi-dimensional fusion method enables the system to more accurately capture the complex relationship between the target recommended object behavior and the device status, thereby providing more comprehensive and accurate input for personalized recommendations and improving the accuracy and intelligence of recommendations.

[0158] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0159] Based on the same inventive concept, the embodiment of the present application also provides a data recommendation device for implementing the data recommendation method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more data recommendation device embodiments provided below can refer to the limitations of a data recommendation method above, and will not be repeated here.

[0160] In an exemplary embodiment, Fig. 9 As shown, a data recommendation device is provided, including: an object data acquisition module 902, an object data fusion module 904, a recommendation model optimization module 906 and a recommendation data acquisition module 908, wherein:

[0161] The object data acquisition module 902 is used to acquire the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time; the universal recommendation model is trained by the historical device signal data and historical object behavior data of the target recommendation object;

[0162] The object data fusion module 904 is used to vectorize and fuse the device signal data and the object behavior data to obtain an object feature fusion vector;

[0163] The recommendation model optimization module 906 is used to train the personalized model parameters in the universal recommendation model according to the object feature fusion vector to obtain a personalized recommendation model for the target recommendation object;

[0164] The recommendation data obtaining module 908 is used to input the device signal data and the object behavior data corresponding to the target recommendation object at the second time into the personalized recommendation model to obtain the initial device recommendation data for the target recommendation object;

[0165] The recommendation model optimization module 906 is further configured to optimize the recommendation strategy of the personalized recommendation model according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time in response to the target recommendation object's instruction of not agreeing with the initial device recommendation data, so as to obtain an optimized recommendation model;

[0166] The recommendation data acquisition module 908 is also used to input the device signal data and object behavior data corresponding to the target recommendation object at the third time into the optimization recommendation model to obtain device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting parameters of the media equipment.

[0167] In one embodiment, the recommendation model optimization module 906 is also used to identify the dynamic demand information of the target recommendation object based on the device signal data, object behavior data and initial device recommendation data corresponding to the second time; analyze the behavior trend information and signal trend information of the target recommendation object based on the dynamic demand information and the initial device recommendation data; optimize the recommendation strategy of the personalized recommendation model based on the dynamic demand information, behavior trend information and signal trend information to obtain an optimized recommendation model.

[0168] In one embodiment, the recommendation model optimization module 906 is also used to identify the actual scenario data of the target recommendation object based on dynamic demand information; analyze the scenario prediction trajectory of the target recommendation object based on the actual scenario data and the initial device recommendation data; and calculate the behavior trend information and signal trend information based on the dynamic demand information, the actual scenario data and the scenario prediction trajectory.

[0169] In one embodiment, the recommendation model optimization module 906 is also used to generate a physical state prediction space of several target recommendation objects on the scene prediction trajectory according to dynamic demand information and actual scene data; for any physical state prediction space, the historical physical state data of the target recommendation object in the historical physical state space is used as reference data, and the physical state prediction data and physical state prediction error of the target recommendation object are generated according to the initial device recommendation data; each physical state prediction data and each physical state prediction error are fitted to generate behavior trend information and signal trend information.

[0170] In one embodiment, the recommendation model optimization module 906 is also used to calculate the degree of association between the target recommendation object and the media device in the universal recommendation model based on the object feature fusion vector, and obtain object device association information; use the behavior signal collaboration model to calculate the collaborative effect of the object feature fusion vector at different time scales to obtain behavior signal collaboration data; according to the object device association information and the behavior signal collaboration data, optimize the personalized model parameters in the universal recommendation model to obtain a personalized recommendation model.

[0171] In one embodiment, the recommendation model optimization module 906 is also used to determine a personalized adjustment strategy for the universal recommendation model based on object device association information; determine personalized adjustment content for the universal recommendation model based on behavioral signal collaborative data; when the personalized adjustment strategy and the personalized adjustment content do not match, use the personalized adjustment strategy and the personalized adjustment content to adjust the personalized model parameters in the universal recommendation model respectively to obtain a first adjustment model and a second adjustment model; and perform loss-level fusion on the first adjustment model and the second adjustment model to obtain a personalized recommendation model.

[0172] In one embodiment, the object data fusion module 904 is also used to perform spatial channel mapping processing on the device signal data to obtain signal behavior space data; perform behavior environment linkage analysis on the object behavior data to obtain object behavior environment relationship data; embed the signal behavior space data and the object behavior environment relationship data into the spatiotemporal domain respectively to obtain a signal spatiotemporal embedding vector and a behavior environment spatiotemporal embedding vector; use a cross-modal attention mechanism to perform feature fusion on the signal spatiotemporal embedding vector and the behavior environment spatiotemporal embedding vector to obtain an object feature fusion vector.

[0173] Each module in the above-mentioned data recommendation device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module.

[0174] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Fig.10 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store server data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external media device through a network connection. When the computer program is executed by the processor, a data recommendation method is implemented.

[0175] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0176] In one embodiment, a computer storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0177] In one embodiment, an electronic device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0178] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in the above-mentioned method embodiments.

[0179] It should be noted that the target recommendation object information (including but not limited to the target recommendation object device information, target recommendation object personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the target recommendation object or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with relevant regulations.

[0180] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0181] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0182] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A data recommendation method, characterized in that: The method comprises: Acquire the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time; the universal recommendation model is trained by the historical device signal data and historical object behavior data of the target recommendation object; Vectorizing and fusing the device signal data and the object behavior data to obtain an object feature fusion vector; According to the object feature fusion vector, training the personalized model parameters in the universal recommendation model to obtain a personalized recommendation model for the target recommendation object; Inputting the device signal data and the object behavior data corresponding to the target recommendation object at the second time into the personalized recommendation model to obtain initial device recommendation data for the target recommendation object; In response to a non-recognition instruction from the target recommendation object regarding the initial device recommendation data, optimizing the recommendation strategy of the personalized recommendation model according to the device signal data corresponding to the second time, the object behavior data, and the initial device recommendation data to obtain an optimized recommendation model; The step of optimizing the recommendation strategy of the personalized recommendation model according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time to obtain the optimized recommendation model includes: identifying dynamic demand information of the target recommendation object according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time; Analyzing the behavior trend information and the signal trend information of the target recommendation object according to the dynamic demand information and the initial device recommendation data; Optimizing the recommendation strategy of the personalized recommendation model according to the dynamic demand information, the behavior trend information, and the signal trend information to obtain the optimized recommendation model; The step of analyzing the behavior trend information and the signal trend information of the target recommendation object according to the dynamic demand information and the initial device recommendation data includes: According to the dynamic demand information, identifying actual scene data of the target recommendation object; Analyzing the scene prediction trajectory of the target recommendation object according to the actual scene data and the initial device recommendation data; Calculating the behavior trend information and the signal trend information according to the dynamic demand information, the actual scenario data, and the scenario prediction trajectory; The device signal data and object behavior data corresponding to the target recommendation object at the third time are input into the optimization recommendation model to obtain device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting parameters of the media equipment.

2. The method according to claim 1, characterized in that The calculating the behavior trend information and the signal trend information according to the dynamic demand information, the actual scenario data and the scenario prediction trajectory includes: Generating a plurality of physical state prediction spaces of the target recommended objects on the scene prediction trajectory according to the dynamic demand information and the actual scene data; For any of the physical state prediction spaces, using the historical physical state data of the target recommended object in the historical physical state space as reference data, generating the physical state prediction data and the physical state prediction error of the target recommended object according to the initial device recommendation data; The physical state prediction data and the physical state prediction errors are fitted to generate the behavior trend information and the signal trend information.

3. The method according to claim 1, characterized in that The step of training the personalized model parameters in the universal recommendation model according to the object feature fusion vector to obtain a personalized recommendation model for the target recommendation object includes: Calculating the association degree between the target recommended object and the media device in the universal recommendation model according to the object feature fusion vector to obtain object device association information; Using a behavior signal coordination model to calculate the coordination effect of the object feature fusion vector at different time scales, to obtain behavior signal coordination data; According to the object device association information and the behavior signal collaborative data, the personalized model parameters in the universal recommendation model are optimized to obtain the personalized recommendation model.

4. The method according to claim 3, characterized in that The optimizing the personalized model parameters in the universal recommendation model according to the object device association information and the behavior signal collaborative data to obtain the personalized recommendation model includes: Determining a personalized adjustment strategy for the universal recommendation model according to the object device association information; Determining personalized adjustment content for the universal recommendation model according to the behavioral signal collaborative data; In the case where the personalized adjustment strategy and the personalized adjustment content do not match, respectively using the personalized adjustment strategy and the personalized adjustment content to adjust the personalized model parameters in the universal recommendation model to obtain a first adjustment model and a second adjustment model; The first adjustment model and the second adjustment model are subjected to loss-level fusion to obtain the personalized recommendation model.

5. The method according to claim 1, characterized in that The vectorizing and fusing the device signal data and the object behavior data to obtain an object feature fusion vector includes: Performing spatial channel mapping processing on the device signal data to obtain signal behavior spatial data; Performing behavior-environment linkage analysis on the object behavior data to obtain object behavior-environment relationship data; Embed the signal behavior space data and the object behavior environment relationship data into the spatiotemporal domain respectively to obtain a signal spatiotemporal embedding vector and a behavior environment spatiotemporal embedding vector; The signal spatiotemporal embedding vector and the behavior environment spatiotemporal embedding vector are feature fused using a cross-modal attention mechanism to obtain the object feature fusion vector.

6. A data recommendation device, characterized in that: The device comprises: An object data acquisition module, used to acquire the device signal data, object behavior data and universal recommendation model corresponding to the target recommendation object at the first time; the universal recommendation model is trained by the historical device signal data and historical object behavior data of the target recommendation object; An object data fusion module, used for vectorizing and fusing the device signal data and the object behavior data to obtain an object feature fusion vector; A recommendation model optimization module, used to train the personalized model parameters in the universal recommendation model according to the object feature fusion vector, so as to obtain a personalized recommendation model for the target recommendation object; A recommendation data obtaining module, used for inputting the device signal data and the object behavior data corresponding to the target recommendation object at the second time into the personalized recommendation model to obtain initial device recommendation data for the target recommendation object; The recommendation model optimization module is further configured to optimize the recommendation strategy of the personalized recommendation model according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time in response to the target recommendation object's instruction of not agreeing with the initial device recommendation data, so as to obtain an optimized recommendation model; The step of optimizing the recommendation strategy of the personalized recommendation model according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time to obtain the optimized recommendation model includes: identifying dynamic demand information of the target recommendation object according to the device signal data, the object behavior data and the initial device recommendation data corresponding to the second time; Analyzing the behavior trend information and the signal trend information of the target recommendation object according to the dynamic demand information and the initial device recommendation data; Optimizing the recommendation strategy of the personalized recommendation model according to the dynamic demand information, the behavior trend information, and the signal trend information to obtain the optimized recommendation model; The step of analyzing the behavior trend information and the signal trend information of the target recommendation object according to the dynamic demand information and the initial device recommendation data includes: According to the dynamic demand information, identifying actual scene data of the target recommendation object; Analyzing the scene prediction trajectory of the target recommendation object according to the actual scene data and the initial device recommendation data; Calculating the behavior trend information and the signal trend information according to the dynamic demand information, the actual scenario data, and the scenario prediction trajectory; The recommendation data acquisition module is also used to input the device signal data and object behavior data corresponding to the target recommendation object at the third time into the optimization recommendation model to obtain device target recommendation data for the target recommendation object; the target recommendation data is used to assist the target recommendation object in setting parameters of the media equipment.

7. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

Images