Portable wireless CPE low-power-consumption communication control method and device

Abstract

The invention provides a portable wireless CPE low-power-consumption communication control method and device, and relates to the technical field of communication control, and the method comprises the steps: calling historical communication behavior data of a target user, and constructing a reference communication behavior feature; retrieving to obtain a plurality of similar users; performing communication behavior joint differential privacy calling to obtain a plurality of differential privacy communication behavior data; capturing a time sequence mode communication rule, and constructing a periodic communication time window sequence; and in the process that the periodic communication time window sequence controls communication on-off of the portable wireless CPE of the target user, confidence action optimization of communication awakening opportunity is carried out according to a plurality of short-term communication event flows of a plurality of similar users. The technical problem that wireless communication control in the prior art generally depends on a fixed communication period or on the basis of a simple rule to control on-off of equipment and cannot dynamically adapt to actual demand fluctuation, so that energy consumption is wasted is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication control, in particular to a low-power communication control method and device for a portable wireless CPE. BACKGROUND

[0002] With the development of mobile Internet and the continuous progress of wireless communication technology, portable wireless CPEs have been widely used in various scenarios, especially in home, office, public places and other environments. Portable wireless CPEs have become an important bridge connecting users and networks. However, the popularity and frequent use of portable wireless CPEs have also brought the demand for low-power and efficient communication. Traditional wireless communication control usually relies on fixed communication cycles or simple rules to control the on-off of devices. The biggest problem of this method is that it lacks dynamic adaptability to actual communication needs of users. In actual application, many users' communication needs have obvious time regularity, and they may communicate frequently in some time periods and hardly use the device in other time periods. Fixed time windows cannot dynamically adapt to such demand fluctuations, resulting in high power consumption of devices in low demand periods, causing unnecessary energy waste. SUMMARY

[0003] The present application provides a low-power communication control method and device for a portable wireless CPE, aiming to solve the technical problem that the wireless communication control of the prior art usually relies on fixed communication cycles or simple rules to control the on-off of devices, and cannot dynamically adapt to actual demand fluctuations, resulting in energy waste.

[0004] The first aspect of the present application provides a low-power communication control method for a portable wireless CPE, the method comprising: retrieving historical communication behavior data of a target user, and constructing a reference communication behavior feature based on the historical communication behavior data; taking the service identity information of the target user and the reference communication behavior feature as hierarchical clustering retrieval conditions to retrieve a plurality of similar users; performing communication behavior joint differential privacy calling on the plurality of similar users to obtain a plurality of differential privacy communication behavior data; performing time sequence mode communication rule capturing on the historical communication behavior data and the plurality of differential privacy communication behavior data to construct a periodic communication time window sequence; and in the process of controlling the on-off of the portable wireless CPE of the target user using the periodic communication time window sequence, performing confidence action optimization of communication wake-up time according to a plurality of short-term communication event streams of the plurality of similar users.

[0005] In a second aspect, the application discloses a portable wireless CPE low-power communication control device, which is used for the above-mentioned portable wireless CPE low-power communication control method, and comprises: a behavior feature construction module, which is used for calling historical communication behavior data of a target user and constructing a reference communication behavior feature based on the historical communication behavior data; a similar user retrieval module, which is used for taking the service identity information of the target user and the reference communication behavior feature as hierarchical clustering retrieval conditions to retrieve a plurality of similar users; a privacy calling module, which is used for performing joint differential privacy calling on the communication behaviors of the plurality of similar users to obtain a plurality of differential privacy communication behavior data; a communication rule capturing module, which is used for capturing the time sequence mode communication rule of the historical communication behavior data and the plurality of differential privacy communication behavior data to construct a periodic communication time window sequence; and a confidence action optimization module, which is used for performing confidence action optimization on a communication wake-up time of the target user according to a plurality of short-term communication event streams of the plurality of similar users in a process of controlling the communication on-off of the portable wireless CPE of the target user by using the periodic communication time window sequence.

[0006] The one or more technical solutions provided in the application have at least the following beneficial effects: By analyzing the historical communication behavior data of the target user and constructing the reference communication behavior feature, the communication mode of the user can be accurately captured, and a reliable data foundation is provided for subsequent user behavior prediction, similar user identification and communication scheduling; the service identity information and the communication behavior feature of the target user are subjected to hierarchical clustering analysis with other users, so that a user group with similar communication demands is found, which provides an accurate reference group for subsequent communication event stream optimization, and errors caused by prediction based on a single user behavior model can be effectively avoided; the communication behaviors of the plurality of similar users are analyzed and processed by using the differential privacy technology, effective user communication behavior modes can be extracted while ensuring data privacy security, the problem of privacy leakage is avoided, data is more compliant during sharing and analysis, and the subsequent user behavior prediction and scheduling still maintains high accuracy; the historical communication behavior data and the differential privacy data are subjected to time sequence mode capturing, and a periodic communication time window can be established based on the communication rule of the user, the periodic time window reflects the communication demand intensity of the user in certain time periods, the communication peak period and the valley period of the user can be more accurately predicted, the scheduling strategy of the communication resource is optimized, resource waste is avoided, and the communication efficiency of the system is improved; the communication wake-up time of the target user is optimized according to the short-term communication event streams of the plurality of similar users, which means that, in the communication on-off control of the portable wireless CPE, the communication wake-up time can be predicted and adjusted according to the historical communication behavior and the real-time demands of the similar users, unnecessary communication wake-up is reduced, energy consumption is reduced, and communication activities are ensured to be woken up at the most appropriate time, so that the energy efficiency and the service quality of the communication system are improved.

[0007] The above description is only a summary of the technical solutions of the present application. In order to enable one skilled in the art to better understand the technical means of the present application and to implement the same according to the contents of the description, and in order to make the above and other purposes, features and advantages of the present application more apparent and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 A flowchart of a low-power-consumption communication control method of a portable wireless CPE is provided for the embodiments of the present application.

[0009] Figure 2 A structural diagram of a low-power-consumption communication control device of a portable wireless CPE is provided for the embodiments of the present application.

[0010] The reference signs are explained as follows: a behavior feature construction module 10, a similar user retrieval module 20, a privacy calling module 30, a communication rule capturing module 40, and a confidence action optimization module 50. DETAILED DESCRIPTION

[0011] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined application purposes, the specific embodiments, structures, features and effects according to the present application are described in detail below in combination with the drawings and preferred embodiments.

[0012] Embodiment one, as shown in the present application, provides a low-power-consumption communication control method of a portable wireless CPE, which comprises: Figure 1 retrieving historical communication behavior data of a target user, and constructing a reference communication behavior feature based on the historical communication behavior data.

[0013] The historical communication behavior data of the target user is extracted and analyzed, specifically, key data such as communication timestamp, communication duration, uplink and downlink data volume, and used communication protocol type is obtained from the communication record of the target user, and feature extraction is performed on these data to form the communication behavior mode of the user, for example, the communication time distribution feature of the target user is extracted, such as peak period and off-peak period; the traffic mode is extracted, such as daily traffic fluctuation; and the service association feature is extracted, such as the use condition of different types of services. These features are combined to constitute the reference communication behavior feature, which provides a basis for the clustering retrieval and similar user matching in the subsequent steps.

[0014] The service identity information of the target user and the reference communication behavior feature are taken as hierarchical clustering retrieval conditions, and a plurality of similar users are retrieved and obtained.

[0015] ​Basic information such as CPE device type, industry attribute and service level is extracted from the target user's business identity information, which helps to determine the user's business background and further filter out similar user groups. The business identity information is converted into a digital vector using one-hot encoding to obtain a business identity encoding vector. Through the similarity based on the target user's baseline communication behavior characteristics and business identity encoding vector, a hierarchical clustering algorithm is used to coarsely filter users to obtain multiple similar users.

[0016] The multiple similar users are subjected to a communication behavior joint differential privacy call to obtain multiple differential privacy communication behavior data.

[0017] For the filtered multiple similar users, differential privacy protection technology is used to protect their communication behavior data. The basic idea of differential privacy is to perturb the user's data so that even if an attacker can obtain some data, it is impossible to infer the real data of a specific user. Such privacy-protected data is used for subsequent analysis to ensure that the user's personal privacy is not infringed even in the case of multi-party data sharing.

[0018] The historical communication behavior data and the multiple differential privacy communication behavior data are subjected to time sequence mode communication law capture to construct a periodic communication time window sequence.

[0019] Based on the preset time slicing granularity, the target user and the multiple similar users' historical communication data are segmented into multiple time periods to obtain multiple communication behavior slices. A flow-gray nonlinear mapping rule is used to perform spatio-temporal gray coding on each communication behavior slice. This is a complex processing process aimed at mapping communication data into gray pattern data suitable for analysis. By merging and analogizing the gray coding of multiple communication behavior slices, a reference and associated gray band is constructed to obtain a behavior gray matrix. Through image-based spatio-temporal law capture of the behavior gray matrix, such as Gaussian mixture model clustering and morphological analysis, periodic communication time window sequences are identified. These time windows represent the peak and trough periods of user communication activities.

[0020] In the process of controlling the portable wireless CPE communication on-off of the target user using the periodic communication time window sequence, the confidence action optimization of the communication wake-up timing is performed according to the multiple short-term communication event streams of the multiple similar users.

[0021] The short-term communication event streams of multiple similar users are monitored, which are processed by a differential privacy protection mechanism to ensure that user privacy is not disclosed. The short-term communication event streams are incrementally spatio-temporally aggregated to construct a group communication state observation value, which more comprehensively displays the user's behavior pattern. The real-time communication state of the target user is retrieved, and a multi-dimensional difference vector between the target user and the group communication state is calculated, which describes the behavior difference between the target user and other similar users. The difference vector is used for action mapping, and a wake-up decision is output according to the mapping result. Finally, based on the wake-up decision, the timing drift compensation of the periodic communication time window sequence is optimized to ensure that the device is woken up for communication at the most appropriate time.

[0022] Further, the historical communication behavior data and the plurality of differential privacy communication behavior data are captured for timing mode communication rules to construct a periodic communication time window sequence, including: The historical communication behavior data is time-sequentially segmented based on a preset time slicing granularity to obtain a plurality of communication behavior slices. A traffic-gray nonlinear mapping rule is used to traverse the plurality of communication behavior slices for spatio-temporal gray coding processing to output a reference gray strip. A plurality of associated gray strips are constructed based on the plurality of differential privacy communication behavior data by analogy. The reference gray strip and the plurality of associated gray strips are stacked to construct a behavior gray matrix. The behavior gray matrix is used for image-based spatio-temporal rule capture to construct the periodic communication time window sequence.

[0023] The time slicing granularity, i.e., the length of each time slice, can be adjusted according to the characteristics of the communication behavior. For example, it can be minutes, hours, days, etc., depending on the time scale of the communication pattern being analyzed. The historical communication behavior data is time-sequentially segmented according to the preset time slicing granularity, and the data is divided into a plurality of communication behavior slices, each slice containing communication behavior information in that time period.

[0024] The traffic-gray nonlinear mapping rule is a nonlinear mapping rule that maps the characteristics (such as traffic, duration, etc.) of each time slice (communication behavior slice) to a gray value. The gray value represents the communication intensity at a certain time or the intensity of a certain behavior pattern. Through nonlinear mapping, the complexity and variability of communication behavior can be better captured. This mapping rule is applied to each communication behavior slice to convert it to a reference gray strip, which represents the communication behavior in that time period. The gray value of the strip reflects the communication intensity or pattern characteristics of that time period.

[0025] For multiple differential privacy communication behavior data, the same traffic-gray nonlinear mapping rule as the benchmark gray strip is adopted to generate multiple associated gray strips. Although these data are protected by privacy, they still reflect similar communication behavior patterns. By spatiotemporal coding of the differential privacy communication behavior data, the communication behavior patterns of these users can be inferred, thereby constructing an associated gray strip similar to the benchmark data. In this way, even under differential privacy protection, the communication behavior rules of similar users can still be analyzed.

[0026] The benchmark gray strip and the multiple associated gray strips are stacked together according to time or other dimensions to form a behavior gray matrix. The behavior gray matrix contains communication behavior information of multiple users in different time periods. The gray value of each matrix element represents the communication intensity or pattern of a specific time and user.

[0027] The behavior gray matrix is essentially an image, and each element is a gray value representing a certain feature of time and communication behavior. Through image processing techniques such as Gaussian mixture model, morphological analysis, and clustering analysis, regular patterns of communication behavior can be extracted from the matrix. According to the image analysis, the captured rules reflect specific periodic communication activities, which can be peak communication periods, low periods, or other regularly occurring communication behaviors of users. Through these rules, a periodic communication time window sequence is constructed to represent the user's high communication demand in certain time periods and low demand in other time periods.

[0028] Further, the image-based spatiotemporal rule capture based on the behavior gray matrix to construct the periodic communication time window sequence includes: Based on the behavior gray matrix, longitudinal gray integral projection is performed to identify multiple communication density peak intervals. Gaussian mixture model clustering hot zone identification is performed on the behavior gray matrix to locate multiple high-frequency communication hot zones. Morphological horizontal connected domain detection is performed on the behavior gray matrix to locate multiple cross-user synchronous communication event windows. The multiple communication density peak intervals, multiple high-frequency communication hot zones, and multiple cross-user synchronous communication event windows are time-series cross-stitched to output the periodic communication time window sequence.

[0029] The behavior gray matrix is subjected to longitudinal gray integral projection. The longitudinal direction refers to processing in the time dimension, i.e., summing or averaging the gray values of each column to obtain the total communication density of each time period. Through this projection, the communication intensity trend in the time period is obtained. If the communication density of certain time periods is significantly higher than that of other time periods, these time periods are communication density peak intervals, which correspond to the user's high communication demand periods and can also be considered as potential communication peaks.

[0030] A Gaussian mixture model is a probabilistic model that assumes data is composed of multiple Gaussian distributions (i.e., normal distributions). By performing Gaussian mixture model clustering on the behavior grayscale matrix, the communication behavior data is divided into multiple different clusters or regions. Based on clustering, high-frequency hot regions of communication behavior are identified, i.e., regions with higher frequency and more concentrated communication activities. These high-frequency communication hot regions correspond to the hot spot communication demand intervals of the user, i.e., the user has higher frequency of communication activities in these time intervals.

[0031] Morphology is an image processing technique used to process structural features in images. By detecting horizontal connected components, regions with similar grayscale values and adjacent in time are identified in the behavior grayscale matrix. In the behavior grayscale matrix, some communication events span multiple users and are synchronized in time. These cross-user synchronized event windows correspond to periods of multi-user collaboration or communication traffic sharing. By detecting these horizontal connected components, regions where multiple users communicate in the same or similar time intervals are found.

[0032] The extracted communication density peak interval, high-frequency hot region, and cross-user synchronized communication event window are cross-spliced in chronological order to merge spatiotemporal patterns from different sources to construct a comprehensive periodic communication time window sequence. This sequence reflects the typical communication period of the target user and similar users, including the communication activity patterns of different user groups in different time intervals, as well as their peak and trough periods.

[0033] Further, the multiple communication density peak intervals, multiple high-frequency communication hot regions, and multiple cross-user synchronized communication event windows are cross-spliced in chronological order to output the periodic communication time window sequence, including: The multiple communication density peak intervals, multiple high-frequency communication hot regions, and multiple cross-user synchronized communication event windows are cross-spliced in chronological order to obtain an initial candidate time window sequence. The initial candidate time window sequence is traversed to locate multiple groups of spatiotemporal conflict overlapping windows. The multiple groups of spatiotemporal conflict overlapping windows are subjected to window weight overlapping resolution to obtain multiple conflict-free fusion time windows. The multiple conflict-free fusion time windows are used to perform coverage-based topological reconstruction of the initial candidate time window sequence, and the periodic communication time window sequence is output.

[0034] The multiple communication density peak intervals, high-frequency communication hot regions, and cross-user synchronized communication event windows identified are merged in chronological order. Each interval and event window represents a different type of communication activity, and the spliced sequence contains multiple possible communication time windows. The cross-spliced sequence is an initial candidate time window sequence that contains all possible high communication demand periods. These time window candidates may have overlaps and conflicts, so further optimization is needed.

[0035] The initial candidate time window sequence is traversed one by one to check the overlap of each time window with other time windows, which may overlap in time, causing different communication events to be unable to proceed simultaneously. By analyzing the time start and end points of each time window, it is determined which time windows overlap, and overlapping time windows will cause resource contention or communication interference, so these space-time conflicts need to be resolved.

[0036] For each set of overlapping space-time windows, weight resolution is performed according to their communication density, priority, etc. Weight resolution can be achieved in various ways, such as selecting high-priority windows, merging adjacent time windows, or selecting a best time window in the overlapping area to avoid resource conflicts. The time windows after weight resolution no longer have conflicts. After fusion, multiple conflict-free fusion time windows are obtained, which represent the optimized communication opportunities of the device in different time periods.

[0037] The initial candidate time window sequence is covered by multiple conflict-free fusion time windows for topology reconstruction. Topology reconstruction can be understood as rearrangement and combination according to conflict-free time windows, ensuring that the entire communication time window sequence covers the user's periodic communication needs while avoiding resource conflicts and interference. The final periodic communication time window sequence is an optimized, non-overlapping time window sequence that reflects the user's communication needs in different periods. These time windows will serve as the basis for communication scheduling for wireless CPE devices, ensuring that the device can wake up at the most appropriate time, reducing energy consumption, and improving communication efficiency.

[0038] Further, in the process of controlling the portable wireless CPE communication on-off of the target user using the periodic communication time window sequence, the confidence action optimization of the communication wake-up time is performed according to the multiple short-term communication event streams of the multiple similar users, including: The event stream of the multiple similar users is monitored for differential privacy protection, obtaining the multiple short-term communication event streams. The multiple short-term communication event streams are incrementally aggregated in space and time to construct a group communication state observation value. The real-time communication state of the target user is retrieved and a multi-dimensional difference vector between the real-time communication state and the group communication state observation value is calculated. The multi-dimensional difference vector is used for action mapping to output a wake-up decision action. The timing drift compensation of the periodic communication time window sequence is performed according to the wake-up decision action.

[0039] In order to protect the privacy of multiple similar users, differential privacy technology is used to protect their communication behavior data, which means that even if a third party analyzes these data, it cannot determine the behavior of a specific user. Monitoring the communication behavior of these similar users, especially short-term communication events such as communication switching, traffic fluctuations, protocol changes, etc. within a short period of time, these short-term event streams reflect the user's communication state and demand at a specific time. Through differential privacy monitoring, multiple short-term communication event streams are obtained, providing timing information for subsequent analysis.

[0040] For the collected multiple short-term communication event streams, an incremental method is used for spatio-temporal aggregation, which means that the current aggregation state is updated in real time whenever new event stream data arrives. This way can efficiently process real-time data streams and avoid processing large data sets. The data of multiple short-term communication event streams are merged according to time (timing) and space (user group), forming group communication state observations, which reflect the communication state changes of multiple similar users in different time periods.

[0041] The real-time communication state of the target user is obtained from the device, including the user's current communication connection status, data transmission volume, connection quality, etc. By comparing the target user's real-time communication state and the group communication state observation, the difference between them is calculated, which is multi-dimensional, such as communication quality, traffic pattern, delay, etc. The multi-dimensional difference vector describes the deviation between the target user's current state and the group state.

[0042] Using a pre-set mapping mechanism, the multi-dimensional difference vector is mapped to a specific operation action, which is achieved through a rule engine or other decision algorithm. The purpose of mapping is to determine the most suitable wake-up time according to the value of the difference vector. Based on the mapping result, the specific wake-up decision action is output, i.e. indicating whether the device needs to wake up, when to wake up, and how to communicate, etc. This wake-up decision action directly affects the power management and communication efficiency of the device.

[0043] The periodic communication time window sequence will be affected by various factors, resulting in actual communication window drift or change. According to the wake-up decision action, the periodic communication time window sequence is adjusted to compensate for the timing drift. Specifically, if there is a deviation between the actual timing of a communication event and the predetermined time window, the position or length of the time window can be adjusted to compensate for this drift. By compensating for the timing drift, the final communication time window sequence can better match the real-time needs of the target user and the group state, thereby optimizing the communication wake-up time of the device and reducing energy waste.

[0044] Further, historical communication behavior data of the target user is retrieved, and a baseline communication behavior feature is constructed based on the historical communication behavior data, including: From the historical communication behavior data, a communication timestamp sequence, a communication duration sequence, an uplink data volume sequence, a downlink data volume sequence, and a communication protocol type sequence are extracted; multi-container structured feature extraction is performed on the communication timestamp sequence, the communication duration sequence, the uplink and downlink data volume sequences, and the communication protocol type sequence to obtain a communication time distribution feature, a communication traffic pattern feature, and a service association feature; and the communication time distribution feature, the communication traffic pattern feature, and the service association feature are merged to construct the baseline communication behavior feature.

[0045] The communication timestamp sequence records the time when the user starts each communication, and each timestamp represents the time when the user initiates a communication, which reflects the time distribution of the user's communication activity; the communication duration sequence records the duration of each communication, i.e., the length of time elapsed from the start to the end of the communication, which reflects the length of each communication of the user and is an important feature for analyzing communication needs and behavior patterns; the uplink data volume sequence records the amount of data transmitted in each communication, which is used to represent the amount of data sent by the user, and the size of the uplink traffic can reflect the user's upload demand; the downlink data volume sequence records the amount of data transmitted in each communication, which represents the amount of data received by the user, and the downlink traffic reflects the user's download demand; the communication protocol type sequence records the protocol type used by the user in each communication, such as HTTP, TCP, UDP, etc., and different protocol types are related to different communication scenarios and service needs.

[0046] Multi-container structured feature extraction means that the data is split into multiple containers for organization and processing, for example, the communication timestamp sequence is divided into different time periods such as weekdays and weekends, daytime and nighttime, in order to analyze the communication patterns in different time periods, or the communication duration, uplink and downlink data volume are grouped by interval to identify different intensities and patterns of user communication.

[0047] Through structured analysis of the communication timestamp sequence, the communication time distribution feature is obtained, such as the user's peak communication period, trough period, or the user's active period, which reflects the user's communication activity in different time periods.

[0048] Through analysis of the uplink and downlink data volume sequences, the communication traffic pattern feature is extracted, such as whether there is an abnormal increase in data traffic in certain time periods, or whether the user is mainly uploading data or downloading data, and the traffic pattern feature indicates the type of user's communication demand.

[0049] By analyzing the sequence of communication protocol types, features related to different services are extracted, such as the high frequency of a certain user using a certain protocol in a specific period. These features help identify the user's service demand type.

[0050] The communication time distribution feature, communication traffic pattern feature, and service association feature are combined. This combination process can be a simple feature splicing, or different features can be fused through weighted averaging, principal component analysis, etc. The final benchmark communication behavior feature is a comprehensive index representing the typical pattern of the user in historical communication behavior.

[0051] Further, the service identity information of the target user and the benchmark communication behavior feature are used as hierarchical clustering retrieval conditions to retrieve a plurality of similar users, including: The CPE loading device type, industry attribute, and service level are extracted from the service identity information. One-hot encoding is used for digital vector conversion of the CPE loading device type, industry attribute, and service level to generate an identity encoding vector. The identity encoding vector is used to perform service identity screening in the service identity layer of the communication user library to obtain a plurality of service-associated users. The plurality of service communication behavior features of the plurality of service-associated users are retrieved in the device feature behavior layer of the communication user library, and the benchmark communication behavior feature is used for similarity quantification to finely screen out the plurality of similar users.

[0052] Communication equipment (CPE) supports different types of devices, such as smartphones, tablets, laptops, routers, etc. Different device types affect communication behavior, such as device processing power, network support, and power consumption requirements. The industry in which the user is located, such as education, healthcare, finance, manufacturing, etc., has a significant impact on communication needs and patterns. For example, the healthcare industry uses more real-time video streaming, and the education industry focuses on data download or access to online education platforms. Service level is related to the user's communication service package, such as ordinary users, advanced users, or VIP users. Different levels of users have different communication priorities and traffic requirements.

[0053] One-hot encoding is a way to convert categorical variables into digital vectors. For each discrete attribute, such as device type, industry attribute, and service level, it is converted into a vector containing multiple binary elements. For example, if the device type has "smartphone" and "laptop", "smartphone" can be represented as [1, 0], and "laptop" as [0, 1]. For each user, the device type, industry attribute, and service level are encoded using one-hot encoding to obtain an identity encoding vector containing multiple dimensions. This vector provides user identity information features for subsequent similarity calculations.

[0054] In the communication user library, there are different user groups, hierarchical according to business identity, for example, according to identity information such as device type, industry attribute, and service level, users can be divided into different business identity groups. By comparing the identity code vector of the target user with the identity code vector of other users in the communication user library, those user groups with similar business identity are screened out. These user groups are similar in device type, industry attribute, and service level, so their communication behavior has a certain degree of similarity.

[0055] In the communication user library, in addition to the business identity layer, there is also a device feature behavior layer, which describes the specific communication behavior characteristics of the user, including communication time distribution, traffic pattern, protocol usage, and other information. For the screened business-related users, their specific business communication behavior characteristics are retrieved from the device feature behavior layer, including the user's communication duration, data transmission volume, uplink and downlink traffic, etc.

[0056] The business communication behavior characteristics of these users are compared with the reference communication behavior characteristics of the target user, and the similarity measurement method includes Euclidean distance, cosine similarity, Manhattan distance, etc. Through this similarity quantification, those users with the closest behavior characteristics are further screened out. Through this fine similarity quantification, more accurate similar users are obtained, which are very similar to the target user in communication behavior and are suitable for communication demand prediction, traffic optimization, etc.

[0057] Further, it also includes: The reference communication behavior characteristics are subjected to Laplace mechanism driven differentiated weight disturbance allocation to generate multiple disturbed communication behavior characteristics. After projecting the multiple disturbed communication behavior characteristics and multiple business communication behavior characteristics into a multi-dimensional heterogeneous feature space, the user behavior similarity topology is constructed by cross-quantifying the behavior similarity between users using covariance weighted Mahalanobis distance. The behavior high-correlation topology is obtained by traversing the user behavior similarity topology using a pre-set similarity threshold to prune the correlation strength. The multiple similar users are obtained by extracting the associated business user nodes in the behavior high-correlation topology.

[0058] Laplace mechanism is a method in differential privacy. By adding noise subject to Laplace distribution to the original data to disturb the data, different degrees of disturbance weight are given to different features according to the importance or sensitivity of the reference communication behavior characteristics, for example, the timestamp feature needs small disturbance to retain periodicity, while some business traffic features can be disturbed more. After Laplace disturbance processing, multiple disturbed communication behavior characteristics are obtained, which not only retain the overall mode of user behavior, but also protect privacy.

[0059] The plurality of disturbance communication behavior features and the plurality of service communication behavior features are uniformly mapped to a multi-dimensional heterogeneous feature space, which allows fusion of different types of data such as time distribution, traffic pattern, service association, etc., to form a multi-dimensional heterogeneous feature representation. Mahalanobis distance comprehensively considers the covariance relationship between features, and can more accurately measure the behavior difference between different users. The covariance weighting method can give different weights to different feature dimensions, so that important features have greater influence in similarity calculation. The Mahalanobis distance of each pair of users in the multi-dimensional heterogeneous feature space is calculated to obtain a behavior similarity matrix between users, and the similarity matrix is converted into a topology to obtain a user behavior similarity topology, wherein the nodes represent users and the weights of the edges represent the behavior similarity between users.

[0060] According to experience or system requirements, a similarity threshold is set, only the user pairs with a similarity higher than the threshold are retained, and the edges with a similarity lower than the threshold are removed in the user behavior similarity topology, and the high-similarity edges are retained. After pruning, only the highly related user nodes and their edges are retained in the topology, and the connections between these nodes can better reflect the real behavior similarity.

[0061] In the behavior high-correlation topology, according to the adjacency relationship of the target user node, user nodes with high similarity to the target user behavior are selected, and finally a plurality of similar users with high similarity to the target user communication behavior are obtained. These similar users will be used as a reference for subsequent communication time window optimization, wake-up time prediction, etc.

[0062] Further, each communication control time window in the sequence of periodic communication time windows includes a window time start and end point, a window active power level, and a window service priority weight.

[0063] The window time start and end point is dynamically generated based on the time boundary of the communication density peak interval, which represents the active communication period of the user, usually when the data transmission demand is high. By analyzing the time boundary in the communication density peak interval, such as the start time and end time of the communication peak, the start and end time of each time window is dynamically set to ensure that these time windows can effectively cover the communication peak of the user and reduce the idle time.

[0064] The window active power level is mapped to a preset power level according to the gray mean value of the corresponding period. The gray value is encoded according to the intensity and frequency of communication activity, and the higher the value, the more frequent the communication activity and the greater the demand. According to the gray mean value of each period, it is mapped to a preset power level, such as P1, P2, P3, P4. These power levels are set by the power consumption management strategy of the system, and P1 is the lowest power and P4 is the highest power. Through the mapping relationship, the communication power is dynamically adjusted according to the actual communication demand to ensure that the peak period can meet the demand and reduce energy waste in low demand periods.

[0065] The window service priority weight is obtained by normalizing the transverse connected domain area of the user-synchronous communication event window. In a multi-user system, some communication events can be performed simultaneously by multiple users, such as multiple users simultaneously starting a video call, which can cause great competition for network resources, and thus priority management is needed. The transverse connected domain refers to the synchronous communication activity region of multiple users in the same time period, indicating a set of multi-user communication events, and the connected domain area indicates the intensity or demand of synchronous communication between multiple users in the same time window. The transverse connected domain area of the user-synchronous communication event window is normalized to obtain a relative priority weight, which is compared with the priority of other time windows to determine the service priority of the time window. The final service priority weight will be assigned to each communication time window, and time windows with high demand and great impact will obtain higher priority to ensure that they are processed first in resource scheduling.

[0066] In the second embodiment, based on the same inventive concept as the low-power communication control method of the portable wireless CPE in the foregoing embodiments, as shown in the following table, the application embodiment provides a low-power communication control device for a portable wireless CPE, which comprises: Figure 2 In the second embodiment, based on the same inventive concept as the low-power communication control method of the portable wireless CPE in the foregoing embodiments, as shown in the following table, the application embodiment provides a low-power communication control device for a portable wireless CPE, which comprises: The behavior feature construction module 10 is configured to call historical communication behavior data of a target user and construct a baseline communication behavior feature based on the historical communication behavior data. The similar user retrieval module 20 is configured to use the service identity information of the target user and the baseline communication behavior feature as hierarchical clustering retrieval conditions to retrieve a plurality of similar users. The privacy calling module 30 is configured to perform joint differential privacy calling on the communication behavior of the plurality of similar users to obtain a plurality of differential privacy communication behavior data. The communication rule capturing module 40 is configured to capture the timing mode communication rule of the historical communication behavior data and the plurality of differential privacy communication behavior data to construct a periodic communication time window sequence. The confidence action optimization module 50 is configured to optimize the confidence action of the communication wake-up time in the process of controlling the communication on-off of the portable wireless CPE of the target user using the periodic communication time window sequence according to a plurality of short-term communication event streams of the plurality of similar users.

[0067] Further, the communication rule capturing module 40 is configured to perform the following operation steps: The historical communication behavior data is segmented based on a preset time slicing granularity to obtain a plurality of communication behavior slices; a flow-gray nonlinear mapping rule is adopted to traverse the plurality of communication behavior slices for spatio-temporal gray coding processing, and a reference gray strip is output; a plurality of associated gray strips are constructed based on the plurality of differential privacy communication behavior data by analogy; the reference gray strip and the plurality of associated gray strips are stacked to construct a behavior gray matrix; and image spatio-temporal rule capturing is performed based on the behavior gray matrix to construct the periodic communication time window sequence.

[0068] Further, the communication rule capturing module 40 is configured to perform the following operation steps: Longitudinal gray integral projection is performed based on the behavior gray matrix to identify a plurality of communication density peak intervals; a plurality of high-frequency communication hot areas are located by performing Gaussian mixture model clustering hot area identification on the behavior gray matrix; a plurality of cross-user synchronous communication event windows are located by performing morphological transverse connected domain detection on the behavior gray matrix; and the plurality of communication density peak intervals, the plurality of high-frequency communication hot areas, and the plurality of cross-user synchronous communication event windows are time-series cross-stitched to output the periodic communication time window sequence.

[0069] Further, the communication rule capturing module 40 is configured to perform the following operation steps: The plurality of communication density peak intervals, the plurality of high-frequency communication hot areas, and the plurality of cross-user synchronous communication event windows are time-series cross-stitched to obtain an initial candidate time window sequence; a plurality of groups of spatio-temporal conflict overlapping windows are located by traversing the initial candidate time window sequence; window weight overlapping is eliminated for the plurality of groups of spatio-temporal conflict overlapping windows to obtain a plurality of non-conflict fusion time windows; and the initial candidate time window sequence is covered and topologically reconstructed using the plurality of non-conflict fusion time windows to output the periodic communication time window sequence.

[0070] Further, the confidence action optimization module 50 is configured to perform the following operation steps: A plurality of short-term communication event streams are obtained by performing differential privacy protection event stream monitoring on the plurality of similar users; a group communication state observation value is constructed by incrementally aggregating the plurality of short-term communication event streams; a real-time communication state of the target user is called and a multi-dimensional difference vector of the real-time communication state and the group communication state observation value is calculated; an action mapping is performed using the multi-dimensional difference vector to output a wake-up decision action; and a time-series drift compensation of the periodic communication time window sequence is performed according to the wake-up decision action.

[0071] Further, the behavior feature construction module 10 is configured to perform the following operation steps: extracting a communication timestamp sequence, a communication duration sequence, an uplink data volume sequence, a downlink data volume sequence and a communication protocol type sequence from the historical communication behavior data; performing multi-container structured feature extraction on the communication timestamp sequence, the communication duration sequence, the uplink and downlink data volume sequences and the communication protocol type sequence to obtain communication time distribution features, communication traffic pattern features and service association features; merging the communication time distribution features, the communication traffic pattern features and the service association features to construct the baseline communication behavior features.

[0072] Further, the similar user retrieval module 20 is configured to perform the following operation steps: extracting a CPE loading device type, an industry attribute and a service level from the service identity information; performing digital vector conversion on the CPE loading device type, the industry attribute and the service level using one-hot encoding to generate an identity encoding vector; performing service identity coarse screening on the service identity layer of the communication user library using the identity encoding vector to obtain a plurality of service-associated users; and calling a plurality of service communication behavior features of the plurality of service-associated users on the device feature behavior layer of the communication user library, and performing similarity quantization with the baseline communication behavior features to finely screen out the plurality of similar users.

[0073] Further, the similar user retrieval module 20 is configured to perform the following operation steps: performing Laplace mechanism driven differentiated weight disturbance allocation on the baseline communication behavior features to generate a plurality of disturbed communication behavior features; after projecting the plurality of disturbed communication behavior features and a plurality of service communication behavior features into a multi-dimensional heterogeneous feature space, performing cross-quantization of user behavior similarity using covariance weighted Mahalanobis distance to construct a user behavior similarity topology; traversing the user behavior similarity topology using a preset similarity threshold to perform correlation strength pruning to obtain a high correlation topology; and extracting associated service user nodes in the high correlation topology to obtain the plurality of similar users.

[0074] Further, each communication control time window in the periodic communication time window sequence includes a window time start and end point, a window active power level and a window service priority weight.

[0075] Through the foregoing detailed description of the portable wireless CPE low-power communication control method, those skilled in the art can clearly understand the portable wireless CPE low-power communication control device in the embodiment. Since the device corresponds to the disclosed method, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0076] The above merely describes the preferred embodiments of the present application, and is not intended to limit the present application in any form. Although the present application has been disclosed with the preferred embodiments as above, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content to obtain equivalent embodiments with equivalent changes, as long as the changes or modifications do not deviate from the technical solutions of the present application. Any modification, equivalent change and modification made to the above embodiments based on the technical essence of the present application still falls within the scope of the technical solutions of the present application.

Claims

1. A low power consumption communication control method for a portable wireless CPE, characterized by, The method comprises: retrieve the historical communication behavior data of the target user, and construct a baseline communication behavior feature based on the historical communication behavior data; use the service identity information of the target user and the baseline communication behavior feature as hierarchical clustering retrieval conditions to retrieve a plurality of similar users; perform communication behavior joint differential privacy calling on the plurality of similar users to obtain a plurality of differential privacy communication behavior data; capture the timing mode communication rule of the historical communication behavior data and the plurality of differential privacy communication behavior data to construct a periodic communication time window sequence; in the process of controlling the portable wireless CPE communication on-off of the target user by using the periodic communication time window sequence, perform confidence action optimization of the communication wake-up timing according to a plurality of short-term communication event streams of the plurality of similar users.

2. The portable wireless CPE low power consumption communication control method of claim 1, wherein, The method comprises: time-slice the historical communication behavior data based on a preset time-slice granularity to obtain a plurality of communication behavior slices; perform spatio-temporal grayscale coding processing on the plurality of communication behavior slices by using a traffic-gray nonlinear mapping rule to output a baseline grayscale strip; construct a plurality of associated grayscale strips based on the plurality of differential privacy communication behavior data by analogy; stack the baseline grayscale strip and the plurality of associated grayscale strips to construct a behavior grayscale matrix; perform image-based spatio-temporal rule capturing based on the behavior grayscale matrix to construct the periodic communication time window sequence.

3. The portable wireless CPE low power consumption communication control method of claim 2, wherein, The method comprises: perform longitudinal grayscale integral projection based on the behavior grayscale matrix to identify a plurality of communication density peak intervals; perform Gaussian mixture model clustering hot area identification on the behavior grayscale matrix to locate a plurality of high-frequency communication hot areas; locate a plurality of cross-user synchronous communication event windows by performing morphological transverse connected domain detection on the behavior grayscale matrix; time-cross splice the plurality of communication density peak intervals, the plurality of high-frequency communication hot areas, and the plurality of cross-user synchronous communication event windows to output the periodic communication time window sequence.

4. The portable wireless CPE low power consumption communication control method of claim 3, wherein, The method comprises: time-cross splice the plurality of communication density peak intervals, the plurality of high-frequency communication hot areas, and the plurality of cross-user synchronous communication event windows to obtain an initial candidate time window sequence; locate a plurality of groups of spatio-temporal conflict overlapping windows by traversing the initial candidate time window sequence; perform window weight overlapping elimination on the plurality of groups of spatio-temporal conflict overlapping windows to obtain a plurality of non-conflict fusion time windows; perform coverage-based topological reconstruction of the initial candidate time window sequence by using the plurality of non-conflict fusion time windows to output the periodic communication time window sequence.

5. The portable wireless CPE low power consumption communication control method of claim 1, wherein, The method comprises: performing differential privacy protection on event stream monitoring of the plurality of similar users to obtain a plurality of short-term communication event streams; incrementally performing spatio-temporal aggregation on the plurality of short-term communication event streams to construct a group communication state observation value; calling a real-time communication state of the target user and calculating a multi-dimensional difference vector between the real-time communication state and the group communication state observation value; performing action mapping using the multi-dimensional difference vector to output a wake-up decision action; performing timing drift compensation on the sequence of periodic communication time windows according to the wake-up decision action.

6. The portable wireless CPE low power consumption communication control method of claim 1, wherein, calling historical communication behavior data of a target user and constructing a baseline communication behavior feature based on the historical communication behavior data, including: extracting a communication timestamp sequence, a communication duration sequence, an uplink data volume sequence, a downlink data volume sequence, and a communication protocol type sequence from the historical communication behavior data; performing multi-container structured feature extraction on the communication timestamp sequence, the communication duration sequence, the uplink and downlink data volume sequences, and the communication protocol type sequence to obtain a communication time distribution feature, a communication traffic pattern feature, and a service association feature; merging the communication time distribution feature, the communication traffic pattern feature, and the service association feature to construct the baseline communication behavior feature.

7. The portable wireless CPE low power consumption communication control method of claim 6, wherein, using the service identity information of the target user and the baseline communication behavior feature as hierarchical clustering retrieval conditions to retrieve a plurality of similar users, including: extracting a CPE-loaded device type, an industry attribute, and a service level from the service identity information; performing digital vector conversion on the CPE-loaded device type, the industry attribute, and the service level using one-hot encoding to generate an identity encoding vector; performing service identity coarse screening in the service identity layer of the communication user library using the identity encoding vector to obtain a plurality of service-associated users; in the device feature behavior layer of the communication user library, calling a plurality of service communication behavior features of the plurality of service-associated users and performing similarity quantization with the baseline communication behavior feature to finely screen out the plurality of similar users.

8. The portable wireless CPE low power consumption communication control method of claim 7, wherein, Further comprising: performing differential weight disturbance distribution driven by a Laplace mechanism on the baseline communication behavior feature to generate a plurality of perturbed communication behavior features; after projecting the plurality of perturbed communication behavior features and the plurality of service communication behavior features into a multi-dimensional heterogeneous feature space, performing cross-quantization of user behavior similarity using a covariance-weighted Mahalanobis distance to construct a user behavior similarity topology; performing correlation strength pruning by traversing the user behavior similarity topology using a pre-set similarity threshold to obtain a behavior high-correlation topology; extracting associated service user nodes in the behavior high-correlation topology to obtain the plurality of similar users.

9. The portable wireless CPE low power consumption communication control method of claim 4, wherein, Each communication control time window in the sequence of periodic communication time windows includes a window time start and end point, a window activation power level, and a window service priority weight.

10. A portable wireless CPE low power consumption communication control device, characterized by, A device for implementing the portable wireless CPE low-power communication control method of any one of claims 1-9, the device comprising: a behavior feature construction module for calling historical communication behavior data of a target user and constructing a baseline communication behavior feature based on the historical communication behavior data; The similar user retrieval module is configured to take the service identity information of the target user and the reference communication behavior feature as hierarchical clustering retrieval conditions to retrieve a plurality of similar users; The privacy calling module is configured to perform joint differential privacy calling on the plurality of similar users to obtain a plurality of differential privacy communication behavior data; The communication law capturing module is configured to capture time sequence mode communication laws of the historical communication behavior data and the plurality of differential privacy communication behavior data to construct a periodic communication time window sequence. The confidence action optimization module is configured to perform confidence action optimization of a communication wake-up timing according to a plurality of short-term communication event streams of the plurality of similar users in a process of controlling the portable wireless CPE communication on-off of the target user by using the periodic communication time window sequence.

Images