Data processing method and device and electronic equipment
By performing interval processing and model prediction on data that do not have data boundaries in the electronic device operation data, the problem of indefinite data processing in the prior art is solved, and accurate control of the power consumption of the communication chip is achieved.
Patent Information
- Application Number
- CN202510190028.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-20
- Publication Date
- 2025-05-27
AI Technical Summary
In the power consumption control of mobile phone communication chips, data without data boundaries cannot be effectively processed, resulting in data processing not being reliable enough and it is difficult to achieve accurate control of chip power consumption.
By obtaining the operation data of the electronic device, the first data without the data boundary is processed according to the processing method corresponding to different data intervals, the first characteristic value is obtained, and input it into the target model to obtain the prediction result, and the operation power consumption of the communication chip is controlled based on the prediction result.
It improves the reliability of data processing, can accurately control the power consumption of components in electronic devices, and is suitable for data types that do not have data boundaries.
Smart Images
Figure CN120050757A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of data processing, and particularly to a data processing method, apparatus, and electronic device. Background Art
[0002] In the control scheme for communication chips in mobile phones, the operation data of the running chips is first preprocessed, and then the preprocessed data is input into an intelligent model, and the power consumption of the communication chips is controlled using the prediction results output by the intelligent model.
[0003] However, in the current preprocessing, only data with upper limit values can be normalized. Summary of the Invention
[0004] In view of this, this application provides a data processing method, apparatus, and electronic device, as follows:
[0005] A data processing method includes:
[0006] Obtaining the operation data of an electronic device;
[0007] Processing the first data in the operation data in a first processing manner to obtain a first feature value; the first processing manner corresponding to the first data in different data intervals is different; the first data is a data type without data boundaries;
[0008] At least using the first feature value as the input data of a target model to obtain the prediction result output by the target model;
[0009] Controlling the operation power consumption of a first component according to the prediction result; the first component is disposed in the electronic device.
[0010] In the above method, preferably, the operation data further includes second data; the second data is processed by a second processing manner to obtain a second feature value; the second processing manner is different from the first processing manner;
[0011] Wherein, the method further includes:
[0012] Using the second feature value as the input data of the target model to enable the target model to output the prediction result.
[0013] In the above method, preferably, the first data includes data in multiple data dimensions;
[0014] Wherein, the processing parameters corresponding to the first data in different data dimensions in the first processing manner are different.
[0015] In the above method, preferably, the first component is a communication chip used to implement data interaction; at least the data related to the communication chip is included in the operation data;
[0016] Controlling the operation power consumption of the first component includes:
[0017] Controlling the first component to execute a processing operation related to a target communication behavior; the processing operation causes the operation power consumption of the first component to change.
[0018] In the above method, preferably, the prediction result includes: the probability of behavior anomaly of the communication chip in the target communication behavior;
[0019] If the behavior anomaly probability is not higher than the target anomaly probability, set the communication parameters for the first component to perform the target communication behavior;
[0020] If the behavior anomaly probability is higher than the target anomaly probability, prohibit the first component from performing the target communication behavior.
[0021] In the above method, preferably, the first data represents the duration of the second component in the first state; the power consumption of the second component in the first state is less than the power consumption of the second component in the second state.
[0022] In the above method, preferably, processing the first data in the operation data in a first processing manner to obtain a first eigenvalue includes:
[0023] Dividing the first data in the operation data into a target interval according to a target threshold; the target threshold is determined based on the data dimension corresponding to the first data;
[0024] Processing the first data in the target interval into a first eigenvalue according to the first processing manner corresponding to the target interval.
[0025] In the above method, preferably, when the target interval is an interval greater than or equal to a first threshold, the first processing manner is: processing the first data into an upper limit value;
[0026] When the target interval is an interval less than or equal to a second threshold, the first processing manner is: processing the first data into a lower limit value; the second threshold is less than the first threshold, and the intermediate value between the first threshold and the second threshold is the target threshold;
[0027] When the target interval is greater than the second threshold and less than the first threshold, the first processing manner is: processing the first data into a value between the upper limit value and the lower limit value.
[0028] A data processing device, comprising:
[0029] A data acquisition unit for acquiring the operation data of an electronic device;
[0030] A data processing unit for processing the first data in the operation data in a first processing manner to obtain a first eigenvalue; the first processing manner for the first data corresponding to different data regions is different; the first data is a data type without data boundaries;
[0031] A model processing unit for using at least the first eigenvalue as input data of a target model to obtain a prediction result output by the target model;
[0032] An operation control unit for controlling the operation function of a first component according to the prediction result; the first component is provided in the electronic device.
[0033] An electronic device, comprising:
[0034] A first component;
[0035] A controller for acquiring the operation data of the electronic device; processing the first data in the operation data in a first processing manner to obtain a first eigenvalue; the first processing manner for the first data in different data intervals is different; the first data is a data type without data boundaries; using at least the first eigenvalue as input data of a target model to obtain a prediction result output by the target model; controlling the operation power consumption of the first component according to the prediction result.
[0036] A computer device / system, comprising: a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the data processing method described in any one of the foregoing.
[0037] A computer-readable storage medium, on which a computer program / instruction is stored, and when the computer program / instruction is executed by a processor, the data processing method described in any one of the foregoing is implemented.
[0038] A computer program product, comprising a computer program / instruction, and when the computer program / instruction is executed by a processor, the data processing method described in any one of the foregoing is implemented. Description of the Drawings
[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to these drawings.
[0040] Figure 1 It is a flowchart for implementing a data processing method provided by an embodiment of the present application;
[0041] Figure 2 It is a partial flowchart of a data processing method provided by an embodiment of the present application;
[0042] Figure 3 It is a schematic structural diagram of a data processing device provided by an embodiment of the present application;
[0043] Figure 4 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application;
[0044] Figure 5 It is a schematic diagram of a function curve for normalizing the mobile phone sleep time Ts in an embodiment of the present application. Detailed implementation manners
[0045] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.
[0046] Refer to Figure 1 As shown, it is a flowchart for implementing a data processing method provided by an embodiment of the present application. This method can be applied to an electronic device with a first component. The electronic device can be a mobile phone, a tablet device, a notebook, etc. The first component can be a component capable of implementing corresponding functions. For example, the first component can be a communication chip used for data interaction, and the first component can perform one or more communication behaviors, such as switching communication networks, configuring adjacent communication networks, etc. The technical solutions in this embodiment are mainly used to improve the data processing reliability when controlling the power consumption of the component, and further achieve accurate control of the component power consumption.
[0047] Specifically, the method in this embodiment can include the following steps:
[0048] Step 101: Obtain the operation data of the electronic device.
[0049] Among them, the operation data can include data in multiple data dimensions during the operation of the electronic device. Taking a mobile phone as an example, the operation data of the mobile phone can include data in multiple data dimensions such as screen-off duration, data traffic volume, communication frequency, total power consumption, etc.
[0050] Specifically, in this embodiment, the operation data of the electronic device can be obtained by reading data from the operation log; alternatively, in this embodiment, data monitoring and data statistics can be respectively performed on each operating component in the electronic device to obtain the operation data of the electronic device. Of course, in this embodiment, the operation data of the electronic device can also be obtained by other means.
[0051] Step 102: Process the first data in the operation data in a first processing manner to obtain a first eigenvalue.
[0052] Among them, the first processing manner corresponding to the first data in different data intervals is different. The first data is a data type without data boundaries.
[0053] It should be noted that the data type without data boundaries can be understood as: as time goes by, the growth of the data has no upper limit, or the decrease of the data has no lower limit. For example, taking the screen-off duration as an example, the mobile phone may remain in the screen-off state continuously, and as time goes by, the screen-off duration keeps increasing without an upper limit; another example is the data traffic. The mobile phone interacts with the Internet through the communication chip, and the data traffic used by the mobile phone keeps increasing without an upper limit as time goes by.
[0054] Among them, the first data may be in different data intervals. For example, taking the target threshold as the boundary of the data interval, the first data may be in any arbitrarily divided data interval. In step 102, for the first data in different data intervals, the first data is processed in different first processing manners to obtain the corresponding first eigenvalues. The first eigenvalues obtained from the first data in different data intervals may be different.
[0055] In one implementation manner, in step 102, the data of the data type without data boundaries in the operation data, that is, the first data, can be screened out first, then the data interval where the first data is located is judged, and then the first data is processed in the first processing manner corresponding to the data interval where the first data is located to obtain the corresponding first eigenvalue.
[0056] It should be noted that the operation data may also include second data, and the second data can be understood as other data in the operation data except the first data without data boundaries. The second data is a data type with data boundaries. Based on this, in this embodiment, the second data can also be processed in a second processing manner, and the second data is processed by the second processing manner to obtain a second eigenvalue.
[0057] Among them, the second processing method is different from the first processing method. That is to say, in this embodiment, the second data with data boundaries and the first data without data boundaries in the running data are processed in different ways respectively to obtain corresponding eigenvalue. Further, for the first data without data boundaries, the first processing methods corresponding to the first data in different data intervals are different.
[0058] Step 103: Use at least the first eigenvalue as the input data of the target model to obtain the prediction result output by the target model.
[0059] Among them, the target model can be an intelligent model capable of realizing data prediction, such as a communication behavior anomaly detection model, a large language model, etc. The target model can process the input data to obtain corresponding prediction results. Taking the communication behavior anomaly detection model as an example, this model can output the prediction result of whether there is an abnormal communication behavior in the communication chip.
[0060] It should be noted that the running data may also include second data. The second data is processed by the second processing method to obtain the second eigenvalue. Correspondingly, in step 103, while using the first eigenvalue as the input data of the target model, the second eigenvalue is also used as the input data of the target model, so that the target model outputs the prediction result according to the first eigenvalue and the second eigenvalue.
[0061] Step 104: Control the operating power consumption of the first component according to the prediction result.
[0062] Among them, the first component is arranged in the electronic device. For example, the first component is a communication chip.
[0063] In one implementation, in this embodiment, the operating parameters of the first component can be set according to the prediction result to change the operating power consumption of the first component, such as reducing it.
[0064] In another implementation, in this embodiment, the first component can be controlled to perform corresponding communication behaviors according to the prediction result to change the operating power consumption of the first component.
[0065] In another implementation, in this embodiment, the first component can be controlled to prohibit performing corresponding communication behaviors according to the prediction result to change the operating power consumption of the first component.
[0066] It should be noted that in this embodiment, the operating power consumption of the first component can also be controlled by other means.
[0067] As can be seen from the above technical solution, in a data processing method provided by an embodiment of the present application, by obtaining the operation data of an electronic device and processing the first data without data boundaries in the operation data in a first processing manner, after obtaining the first eigenvalue, the first eigenvalue is used as the input data of the target model to obtain the prediction result output by the target model. Finally, according to the prediction result, the operating power consumption of the first component disposed in the electronic device is controlled. And the first processing manners corresponding to the first data in different data intervals are different. In this way, even for the first data without data boundaries, it can be processed into the corresponding first eigenvalue, and the situation where it cannot be processed into an eigenvalue due to the lack of data boundaries in the first data will not occur. Thereby, the reliability of data processing is improved, and further, the accurate control of the power consumption of components in the electronic device is realized.
[0068] In one implementation, the first data may include data in multiple data dimensions. And the processing parameters corresponding to the first data in different data dimensions are different in the first processing manner.
[0069] In a specific implementation, the processing parameter may be a threshold parameter for dividing intervals of the first data in the first processing manner. For example, taking the processing parameter as the target threshold, the target thresholds corresponding to the first data in different data dimensions are different. The target threshold is used to divide the corresponding first data into at least two data intervals, and the first data is processed based on the first processing manner corresponding to the data interval where the first data is divided to obtain the first eigenvalue.
[0070] For example, taking the first data as the screen-off duration, the target threshold may be 3 minutes, that is, 180 seconds. Based on 180 seconds and the corresponding floating value, three data intervals are divided: a data interval less than or equal to 180 degrees minus the floating value, a data interval greater than 180 seconds minus the floating value and less than 180 seconds plus the floating value, and a data interval greater than or equal to 180 seconds plus the floating value.
[0071] For another example, taking the first data as the data traffic, the target threshold may be 500K or 1M. Based on 500K, two data intervals are divided: a data interval greater than 500K, and a data interval less than or equal to 500K.
[0072] In one implementation, the first component can be a communication chip used to implement data interaction, and the running data at least includes data related to the communication chip, such as data traffic. The first data represents the duration of the second component in the first state, and the power consumption of the second component in the first state is less than that in the second state. The second component is a component in an electronic device, such as the Central Processing Unit (CPU) in a mobile phone. The first state can be a state where the second component does not perform a specific process, and the second state can be a state where the second component performs a specific process. For example, the first state is a state where the CPU does not trigger the display screen to display an image, that is, the screen-off state; the second state is a state where the CPU triggers the display screen to display an image, that is, the screen-on state.
[0073] Based on this, when controlling the running power consumption of the first component in step 104, it can be achieved in the following way:
[0074] Control the first component to execute a processing operation related to the target communication behavior. The processing operation causes the running power consumption of the first component to change.
[0075] Among them, the prediction result can include: the behavior anomaly probability of the communication chip in the target communication behavior. The target communication behavior can be understood as the communication behavior implemented by the communication chip for data interaction. For example, the communication behavior of adding or deleting a 5G secondary cell in the ENDC (E-UTRAN New Radio Dual Connectivity) mode; another example is the communication behavior of switching between a 4G cell and a 5G secondary cell in the ENDC mode. The target model can predict whether there is an anomaly in these target communication behaviors to obtain a prediction result, and the prediction result is composed of the behavior anomaly probabilities corresponding to these target communication behaviors. The magnitude of the behavior anomaly probability represents whether there is an anomaly in the communication chip in the target communication behavior. Correspondingly, in step 104, based on whether the behavior anomaly probability is higher than the target anomaly probability, corresponding processing operations are performed to control the change in the running power consumption of the first component.
[0076] In one case, if the behavior anomaly probability is not higher than the target anomaly probability, then the communication parameters for the first component to perform the target communication behavior can be set so that the first component can perform the target communication behavior.
[0077] For example, taking the communication behavior of handover between a 4G cell and a 5G secondary cell in ENDC mode as an example, if the behavior anomaly probability corresponding to this communication behavior does not exceed the target anomaly probability, it indicates that the communication behavior of handover of the mobile phone between the 4G cell and the 5G secondary cell in ENDC mode is normal. At this time, the signal strength threshold for the communication chip to perform cell network handover can be reduced so that the communication chip can continue to perform handover between the 4G cell and the 5G secondary cell in ENDC mode; if the behavior anomaly probability corresponding to this communication behavior exceeds the target anomaly probability, it indicates that the communication behavior of handover of the mobile phone between the 4G cell and the 5G secondary cell in ENDC mode is abnormal. At this time, the signal strength threshold for the communication chip to perform cell network handover can be increased so that the difficulty of the communication chip to perform handover between the 4G cell and the 5G secondary cell in ENDC mode is increased, avoiding frequent handover between the 4G cell and the 5G secondary cell in ENDC mode, thereby reducing the operating power consumption of the communication chip.
[0078] In another case, if the behavior anomaly probability is higher than the target anomaly probability, in step 104, the first component can be prohibited from performing the target communication behavior. Specifically, in step 104, the first component can be directly prohibited from performing the target communication behavior, or in step 104, the first component can be prohibited from performing the pre-order behavior corresponding to the target communication behavior.
[0079] For example, taking the communication behavior of adding or deleting a 5G secondary cell in ENDC mode as an example, if the behavior anomaly probability corresponding to this communication behavior exceeds the target anomaly probability, it indicates that the communication behavior of adding or deleting a 5G secondary cell in ENDC mode by the mobile phone is abnormal, such as frequently adding or deleting a 5G secondary cell in ENDC mode, etc. At this time, the communication chip can be prohibited from adding or deleting a 5G secondary cell in ENDC mode, or the communication chip can be prohibited from reporting the measurement report of the 5G secondary cell in ENDC mode. Thus, the communication chip can be prevented from adding or deleting a 5G secondary cell in ENDC mode too frequently, which can reduce the operating power consumption of the communication chip.
[0080] In one implementation, when processing the first data in the processing operation data in the first processing manner to obtain the first eigenvalue in step 102, it can be implemented in the following manner, such as Figure 2 as shown in
[0081] Step 201: Divide the first data in the operation data into a target interval according to a target threshold.
[0082] Among them, the target threshold is determined based on the data dimension corresponding to the first data. For example, taking the screen-off duration as an example, the target threshold is 180 seconds; taking the data traffic as an example, the target threshold is 500K.
[0083] In one implementation, in this embodiment, two data intervals can be divided according to the target threshold: a first interval greater than the target threshold and a second interval less than or equal to the target threshold. Based on this, the first data is divided into the first interval or the second interval, that is, the target interval, according to the data value of the first data.
[0084] For example, taking data traffic as an example, the target threshold is 500K. The first interval divided is the traffic interval greater than 500K, and the second interval is the traffic interval less than or equal to 500K. Based on this, if the first data is 350K, then the first data is divided into the second interval; if the first data is 650K, then the first data is divided into the first interval.
[0085] In another implementation, in this embodiment, the target threshold corresponds to a floating value. Three data intervals can be divided according to the target threshold: a data interval less than or equal to the target threshold minus the floating value, a data interval greater than the target threshold minus the floating value and less than the target threshold plus the floating value, and a data interval greater than or equal to the target threshold plus the floating value.
[0086] For example, taking the screen-off duration as an example, the target threshold can be 3 minutes, that is, 180 seconds, and the floating value can be 3π. Accordingly, the first interval divided is: a data interval less than or equal to 180 degrees minus the floating value, the second interval is: a data interval greater than 180 seconds minus the floating value and less than 180 seconds plus the floating value, and the third interval is: a data interval greater than or equal to 180 seconds plus the floating value. Based on this, taking the floating value of 3π as an example, if the first data is 200 seconds, then the first data is divided into the third interval; if the second data is 100 seconds, then the first data is divided into the first interval; if the first data is 180 seconds, then the first data is divided into the second interval.
[0087] Step 202: Process the first data in the target interval into a first eigenvalue according to the first processing method corresponding to the target interval.
[0088] Among them, when the target intervals into which the first data is divided are different, the first processing methods used are different.
[0089] Taking the division of three data intervals by the target threshold as an example, when the target interval is an interval greater than or equal to the first threshold, the first processing method is: process the first data into the upper limit value. The first threshold can be the target threshold plus the floating value.
[0090] When the target interval is an interval less than or equal to the second threshold, the first processing method is: processing the first data into the lower limit value. The second threshold is less than the first threshold, and the second threshold can be the target threshold minus the floating value. The intermediate value between the first threshold and the second threshold is the target threshold.
[0091] When the target interval is greater than the second threshold and less than the first threshold, the first processing method is: processing the first data into a value between the upper limit value and the lower limit value.
[0092] In specific implementation, when processing the first data into a value between the upper limit value and the lower limit value, the function value of the first data can be calculated through the target function to obtain the value between the upper limit value and the lower limit value.
[0093] The target function is a curve function determined based on the first threshold and the second threshold, and the maximum value output by the target function is the upper limit value, and the minimum value output by the target function is the lower limit value.
[0094] For example, the target function can be a sine function, a cosine function or other linear functions. For example, taking the screen-off duration as an example, the target threshold can be 3 minutes, that is, 180 seconds, the floating value can be 3π, the first threshold is 180 seconds plus 3π, the second threshold is 180 seconds minus 3π, and the target function can be a sine function, as shown in formula (1):
[0095]
[0096] Based on this, when the target interval where x is located is greater than 180 + 3×π and less than 180 - 3×π, the first eigenvalue is obtained through g(x), and the first eigenvalue is a value between -1 and 1. When the target interval where x is located is less than or equal to 180 - 3×π, x is processed as -1; when the target interval where x is located is greater than or equal to 180 + 3×π, x is processed as 1.
[0097] Reference Figure 3 , which is a schematic structural diagram of a data processing device provided by an embodiment of the present application. The device can be applied to an electronic device having a first component. The electronic device can be a mobile phone, a tablet device, a notebook, etc. The first component can be a component capable of implementing corresponding functions. For example, the first component can be a communication chip used for data interaction, and the first component can perform one or more communication behaviors, such as switching communication networks, configuring adjacent communication networks, etc. The technical solution in this embodiment is mainly used to improve the data processing reliability when controlling the power consumption of the component, and further achieve accurate control of the component power consumption.
[0098] Specifically, the device in this embodiment may include the following units:
[0099] A data acquisition unit 301, configured to acquire operation data of an electronic device;
[0100] A data processing unit 302, configured to process first data in the operation data in a first processing manner to obtain a first eigenvalue; different first data corresponding to different data regions are processed in different first processing manners; the first data is a data type without data boundaries;
[0101] A model processing unit 303, configured to use at least the first eigenvalue as input data of a target model to obtain a prediction result output by the target model;
[0102] An operation control unit 304, configured to control an operation function of a first component according to the prediction result; the first component is disposed in the electronic device.
[0103] As can be seen from the above technical solution, in a data processing device provided in an embodiment of the present application, by acquiring operation data of an electronic device and processing first data without data boundaries in the operation data in a first processing manner, after obtaining a first eigenvalue, the first eigenvalue is used as input data of a target model to obtain a prediction result output by the target model. Finally, according to the prediction result, the operation power consumption of the first component disposed in the electronic device is controlled. And different first data corresponding to different data intervals are processed in different first processing manners, so that even for first data without data boundaries, it can be processed into corresponding first eigenvalues, and the situation that it cannot be processed into eigenvalues due to the lack of data boundaries of the first data will not occur, thereby improving the reliability of data processing and further realizing accurate control of the power consumption of components in the electronic device.
[0104] In an implementation manner, the operation data further includes second data; the second data is processed by a second processing manner to obtain a second eigenvalue; the second processing manner is different from the first processing manner;
[0105] Wherein, in this embodiment, the model processing unit 303 is further configured to use the second eigenvalue as input data of the target model to enable the target model to output the prediction result.
[0106] In an implementation manner, the first data includes data in multiple data dimensions; wherein, the processing parameters corresponding to the first data in different data dimensions in the first processing manner are different.
[0107] In an implementation manner, the first component is a communication chip used to implement data interaction; the operation data at least includes data related to the communication chip;
[0108] The operation control unit 304 is specifically configured to: control the first component to perform processing operations related to the target communication behavior; the processing operations cause the operating power consumption of the first component to change.
[0109] Wherein, the prediction result includes: the behavior anomaly probability of the communication chip in the target communication behavior; if the behavior anomaly probability is not higher than the target anomaly probability, the operation control unit 304 sets the communication parameters for the first component to perform the target communication behavior; if the behavior anomaly probability is higher than the target anomaly probability, the operation control unit 304 prohibits the first component from performing the target communication behavior.
[0110] Based on the above implementation, the first data represents the duration of the second component in the first state; the power consumption of the second component in the first state is less than the power consumption of the second component in the second state.
[0111] In one implementation, the data processing unit 302 is specifically configured to: divide the first data in the operation data into a target interval according to a target threshold; the target threshold is determined based on the data dimension corresponding to the first data; and process the first data in the target interval into a first eigenvalue according to the first processing method corresponding to the target interval.
[0112] Wherein, when the target interval is an interval greater than or equal to the first threshold, the first processing method is: processing the first data into an upper limit value; when the target interval is an interval less than or equal to the second threshold, the first processing method is: processing the first data into a lower limit value; the second threshold is less than the first threshold, and the intermediate value between the first threshold and the second threshold is the target threshold; when the target interval is greater than the second threshold and less than the first threshold, the first processing method is: processing the first data into a value between the upper limit value and the lower limit value.
[0113] It should be noted that the specific implementation manners of the units in this embodiment can refer to the corresponding content in the foregoing, and will not be elaborated herein.
[0114] Refer to Figure 4 , which is a schematic structural diagram of an electronic device provided by an embodiment of the present application. The electronic device may include the following structures:
[0115] A first component 401;
[0116] A controller 402 is configured to obtain the operation data of the electronic device; process the first data in the operation data in a first processing manner to obtain a first eigenvalue; the first processing manner corresponding to the first data in different data intervals is different; the first data is a data type without data boundaries; at least use the first eigenvalue as input data of a target model to obtain a prediction result output by the target model; and control the operating power consumption of a first component 401 according to the prediction result.
[0117] As can be seen from the above technical solution, in an electronic device provided by an embodiment of the present application, by obtaining the operation data of the electronic device and processing the first data without data boundaries in the operation data in a first processing manner, after obtaining the first eigenvalue, the first eigenvalue is used as input data of the target model to obtain a prediction result output by the target model. Finally, according to the prediction result, the operating power consumption of the first component provided in the electronic device is controlled. And the first processing manner corresponding to the first data in different data intervals is different, so that even for the first data without data boundaries, it can be processed into corresponding first eigenvalues, and the situation that the first data cannot be processed into eigenvalues due to the lack of data boundaries will not occur, thereby improving the reliability of data processing and further realizing accurate control of the power consumption of components in the electronic device.
[0118] Taking the scenario of using a neural network model to control the power consumption of a communication chip, i.e., a modem, in a mobile phone as an example, the technical solution of the present application will be illustrated as follows:
[0119] Currently, the mainstream data preprocessing / normalization methods for neural network models include the extreme value method (interval scaling method), the proportion method, standardization, and some non-linear normalization methods. In the power-saving solution for mobile phone modem development, the sleep duration of the CPU needs to be considered. Usually, this duration is distinguished by a threshold, and different logical processes, i.e., different power consumption control methods, will be generated for the duration above or below the threshold. However, the current data normalization methods cannot reflect the existence of the threshold, resulting in low efficiency during the forward and backward propagation of the neural network, consuming a large amount of computing, and at the same time, the prediction result is not good. Moreover, some data will have no boundaries. The current neural network models mainly process graphic or sound data, which all have fixed boundaries, such as the color grayscale of image pixels or the amplitude of audio sampling points. However, some data will have an unpredictable upper limit, such as the sleep time of a mobile phone, which may be 10 seconds, 10 hours, or even 10 days. When processing this kind of data, if the above extreme value method or proportion method is used, many normal mobile phone sleep times will be extremely compressed and overly offset, which is not conducive to data preprocessing for the neural network model.
[0120] In view of this, the present application proposes a new data normalization method for neural network models:
[0121] This normalization method is for the collected mobile phone sleep duration Ts. Ts has no data boundary. Based on this, generally, when Ts is lower than the target threshold T thrd such as 180 seconds, power-saving actions may not be taken. In order to enable the neural network model to accurately process the sleep duration and other operation data, Ts needs to be normalized. Specifically, in the present application, there is no maximum value for Ts, and the corresponding normalization method is the following piecewise function, as shown in formula (2):
[0122]
[0123] These 3 functions make interval divisions for the mobile phone screen off time x (i.e., the sleep duration Ts). For those higher than the target threshold 180 by a certain range (such as 3π), they are regarded as 1; for those lower than the target threshold 180 by a certain range (such as 3π), they are regarded as -1; and within the short time (6π) between the two, g(x) is used to replace them to achieve a smooth transition between the two from 1 to -1, as Figure 5 shown. It is the function curve of the piecewise function. Among them, the abscissa is the sleep time Ts, and the abscissa is the first eigenvalue, which is a value between -1 and 1.
[0124] It can be seen that adopting the technical solution of the present application has the following advantages:
[0125] 1. The target threshold can be reflected, and the eigenvalues are evenly distributed about 0. Those lower than the target threshold are all negative numbers, and those higher than the threshold are positive numbers, which is convenient for the subsequent learning and classification of the neural network model.
[0126] 2. There is no upper limit requirement for the input data, and the sigmoid function is used to transform extremely large inputs to within 1.
[0127] 3. The amount of training data required can be greatly reduced, and only representative examples around the threshold are needed.
[0128] In this specification, each embodiment is described in a progressive manner. The key points of each embodiment are the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method part.
[0129] Those skilled in the art may further realize that the units and algorithm steps of each example described in connection with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application.
[0130] The steps of the methods or algorithms described in connection with the embodiments disclosed herein can be directly implemented by hardware, software modules executed by a processor, or a combination of the two. The software modules can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well-known in the technical field.
[0131] The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application will not be limited to the embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.
Claims
1. A data processing method, comprising: Obtaining operation data of electronic equipment; Processing first data in the operating data in a first processing manner to obtain a first characteristic value; The first processing methods corresponding to the first data in different data intervals are different; The first data is a data type without data boundaries; At least using the first characteristic value as input data of a target model to obtain a prediction result output by the target model; According to the prediction result, the operating power consumption of a first component is controlled; the first component is arranged in the electronic device.
2. The method according to claim 1, wherein the operation data further includes second data; the second data is processed by a second processing method to obtain a second characteristic value; the second processing method is different from the first processing method; in, The method further comprises: The second feature value is used as input data of the target model so that the target model outputs the prediction result.
3. The method according to claim 1 or 2, wherein the first data includes data in multiple data dimensions; in, The first data in different data dimensions have corresponding processing parameters different in the first processing method.
4. The method according to claim 1, wherein the first component is a communication chip used to implement data interaction; The operation data at least includes data related to the communication chip; The controlling the operating power consumption of the first component includes: Controlling the first component to perform a processing operation related to the target communication behavior; The processing operation causes a change in the operating power consumption of the first component.
5. The method according to claim 4, wherein the prediction result comprises: Probability of abnormal behavior of the communication chip on the target communication behavior; If the abnormal behavior probability is not higher than the target abnormal probability, setting the communication parameters for the first component to perform the target communication behavior; If the behavior abnormality probability is higher than the target abnormality probability, the first component is prohibited from performing the target communication behavior. 6 . The method according to claim 4 , wherein the first data represents a duration of time that the second component is in a first state; and the power consumption of the second component in the first state is less than the power consumption of the second component in the second state.
7. The method according to claim 1, processing the first data in the operation data in a first processing manner to obtain a first characteristic value, comprising: According to the target threshold, first data in the operating data is divided into a target interval; The target threshold is determined based on the data dimension corresponding to the first data; According to a first processing method corresponding to the target interval, the first data in the target interval is processed into a first eigenvalue.
8. The method according to claim 7, when the target interval is an interval greater than or equal to a first threshold value, the first processing manner is: processing the first data as an upper limit value; When the target interval is an interval less than or equal to the second threshold, the first processing method is: processing the first data as a lower limit value; the second threshold value is less than the first threshold value, and the middle value between the first threshold value and the second threshold value is the target threshold value; When the target interval is greater than the second threshold value and less than the first threshold value, the first processing manner is: processing the first data into a value between the upper limit value and the lower limit value.
9. A data processing device, comprising: A data acquisition unit, used to acquire operation data of the electronic device; A data processing unit, configured to process first data in the operation data in a first processing manner to obtain a first characteristic value; the first processing manners corresponding to processing the first data in different data regions are different; The first data is a data type without data boundaries; A model processing unit, configured to use at least the first feature value as input data of a target model to obtain a prediction result output by the target model; An operation control unit is used to control the operation function of a first component according to the prediction result; the first component is arranged in the electronic device.
10. An electronic device comprising: first component; A controller, used to obtain operation data of the electronic device; Processing first data in the operating data in a first processing manner to obtain a first characteristic value; The first processing methods corresponding to the first data in different data intervals are different; The first data is a data type without data boundaries; at least the first characteristic value is used as input data of a target model to obtain a prediction result output by the target model; and according to the prediction result, the operating power consumption of the first component is controlled.