CN117119535A

CN117119535A - Data distribution method and system for mobile terminal cluster hot spot sharing

Info

Publication number: CN117119535A
Application number: CN202311192198.3A
Authority: CN
Inventors: 唐乳蜂; 蔡进; 胡文勇
Original assignee: Shenzhen Huiyou Yipin Technology Co ltd
Current assignee: Shenzhen Huiyou Yipin Technology Co ltd
Priority date: 2023-09-15
Filing date: 2023-09-15
Publication date: 2023-11-24

Abstract

The invention relates to the technical field of data distribution, and discloses a data distribution method for mobile terminal cluster hot spot sharing, which comprises the following steps: cleaning the hot spot sharing log set data of the mobile terminal cluster into a standard sharing log set; training a preset sharing decision model into a sharing and shunting model according to a standard sharing log set; generating a behavior feature sequence set according to the standard shared log set and the shared behavior feature set, and training a preset behavior time sequence model into a behavior analysis model according to the behavior feature sequence set; the method comprises the steps that real-time logs of equipment in a mobile terminal cluster are used for calculating secondary behavior characteristics corresponding to the real-time logs of the equipment in the mobile terminal cluster by using a behavior analysis model, and secondary sharing strategies corresponding to the secondary behavior characteristics are calculated according to a sharing shunt model; and carrying out shared transmission on the encrypted real-time transmission data according to the secondary sharing strategy. The invention also provides a data distribution system for sharing the cluster hot spots of the mobile terminal. The invention can improve the data distribution efficiency.

Description

Data distribution method and system for mobile terminal cluster hot spot sharing

Technical Field

The invention relates to the technical field of data distribution, in particular to a data distribution method and a system for mobile terminal cluster hot spot sharing.

Background

With the continuous rising of mobile social networks and the continuous development of intelligent terminal equipment, the data traffic of the mobile terminal is in explosive growth, and particularly under the condition of lacking of a Wi-Fi network or unstable signals, a hotspot sharing method of the mobile terminal is required to be used for sharing internet connection for a cluster, so that huge traffic load and network congestion are caused to the mobile network of the hotspot equipment, and therefore, data distribution is required to be carried out on the hotspot sharing of the mobile terminal cluster.

The existing data distribution method for hot spot sharing of the mobile terminal cluster is mostly a data distribution method based on a fixed communication protocol, for example, webSocket communication or D2D communication is mainly used for realizing direct sharing transmission data among devices in the mobile terminal cluster, in practical application, the data distribution method based on the fixed communication protocol cannot dynamically adjust a policy of data distribution according to the working state and load of each cluster device, which may cause data congestion and traffic load of a single mobile device, and the data distribution method based on the fixed communication protocol may encrypt and transmit the whole data in the data transmission process, and may cause cycle growth of data transmission and lower efficiency in data distribution because of integral data decryption at a receiving terminal.

Disclosure of Invention

The invention provides a data distribution method and a system for mobile terminal cluster hot spot sharing, which mainly aim to solve the problem of lower efficiency in data distribution.

In order to achieve the above object, the present invention provides a data splitting method for mobile terminal cluster hot spot sharing, including:

acquiring a hot spot sharing log set of a mobile terminal cluster, cleaning the hot spot sharing log set data into a standard sharing log set, and splitting the standard sharing log set time sequence into a hot spot log sequence set;

extracting a sharing behavior feature set, a sharing policy feature set and a sharing efficiency feature set from the standard sharing log set respectively, training a preset sharing decision model into a sharing shunt model according to the sharing behavior feature set, the sharing policy feature set and the sharing efficiency feature set, wherein training the preset sharing decision model into the sharing shunt model according to the sharing behavior feature set, the sharing policy feature set and the sharing efficiency feature set comprises: combining the sharing behavior feature set and the sharing strategy feature set into a sharing parameter feature set, and calculating a feature gain set of the sharing parameter feature set according to a preset sharing decision model by using the following gain algorithm:

Wherein G is _i For the ith feature gain in the feature gain set, i and N are feature group numbers, N is the total number of shared parameter feature groups in the shared parameter feature group set, the total number of shared parameter feature groups in the shared parameter feature group set is equal to the total number of feature gains in the feature gain set, gamma and psi are preset constants, and G _i,L For the sum of the first partial derivatives of the data contained in the left node of the root node when the input parameter of the shared decision model is the ith shared parameter feature group in the shared parameter feature group, H _i,L For the sum of the second partial derivatives of the data contained in the left node of the root node when the input parameters of the shared decision model are the ith shared parameter feature group in the shared parameter feature group, G _i,R For the sum of the first partial derivatives of the data contained in the right node of the root node when the input parameter is the ith shared parameter feature group in the shared parameter feature group of the shared decision model, H _i,R For the right node of the root node of the shared decision model when the input parameter is the ith shared parameter feature group in the shared parameter feature group setSum of second partial derivatives of the data contained in the points, G _n,L For the sum of the first partial derivatives of the data contained in the left node of the root node when the input parameter of the shared decision model is the nth shared parameter feature group in the shared parameter feature group, H _n,L For the sum of the second partial derivatives of the data contained in the left node of the root node when the input parameters of the shared decision model are the nth shared parameter feature group in the shared parameter feature group, G _n,R For the sum of the first partial derivatives of the data contained in the right node of the root node when the input parameter is the nth shared parameter feature group in the shared parameter feature group of the shared decision model, H _n,R Adding the second partial derivatives of the data contained in the right node of the root node when the input parameter of the sharing decision model is the nth sharing parameter feature group in the sharing parameter feature group, wherein max is a maximum value function symbol, and min is a minimum value function symbol; performing iterative node updating on the shared decision model according to the characteristic gain set to obtain a secondary decision model; calculating a prediction efficiency characteristic set corresponding to the shared parameter characteristic set by using the secondary decision model; calculating an efficiency loss value between the predicted efficiency feature set and the shared efficiency feature set, and carrying out iterative updating on model parameters of the secondary decision model according to the efficiency loss value to obtain a shared shunt model;

Generating a behavior feature sequence set according to the hot spot log sequence set and the shared behavior feature set, and training a preset behavior time sequence model into a behavior analysis model according to the behavior feature sequence set;

selecting devices in the mobile terminal cluster one by one as target mobile devices, acquiring real-time logs of the target mobile devices, calculating secondary behavior characteristics corresponding to the real-time logs by using the behavior analysis model, and calculating secondary sharing strategies corresponding to the secondary behavior characteristics according to the sharing shunt model;

and acquiring real-time shunt data of the target mobile equipment, identifying privacy data from the real-time shunt data, privacy encrypting the real-time shunt data into real-time transmission data according to the privacy data, and carrying out sharing transmission on the transmission data according to the secondary sharing strategy.

Optionally, the cleaning the hotspot sharing log set data into a standard sharing log set includes:

screening out repeated shared logs from the hot spot shared log set to obtain a duplicate-removed shared log set;

respectively extracting an offside shared log and a messy code shared log from the duplicate removal shared log set;

Carrying out log feature clustering on the duplicate removal shared log set to obtain a clustered log class set;

and replacing the offside shared log and the messy code shared log in the duplicate removal shared log set according to the clustering log class set to obtain a standard shared log.

Optionally, the splitting the standard shared log set into the hotspot log sequence set sequentially includes:

selecting standard shared logs in the standard shared log set one by one as target standard shared logs, and performing time sequence sequencing on each log in the target standard shared logs to obtain target sequencing shared logs;

splitting each log in the target ordering shared log according to a preset time domain period to obtain a hot spot log sequence;

and collecting all the hot spot log sequences into a hot spot log sequence set.

Optionally, the extracting the sharing behavior feature set, the sharing policy feature set and the sharing efficiency feature set from the standard sharing log set respectively includes:

selecting standard shared logs in the standard shared log set one by one as target shared logs, and performing format transcoding on the target shared logs to obtain target shared data;

Respectively extracting a transmission time stamp, a transmission data quantity, a transmission direction and a transmission data type from the target shared data, and encoding the transmission time stamp, the transmission data quantity, the transmission direction and the transmission data type into a shared behavior characteristic;

extracting bandwidth parameters, frequency parameters, compression parameters and segmentation parameters from the target shared data respectively, and encoding the bandwidth parameters, the frequency parameters, the compression parameters and the segmentation parameters into sharing strategy characteristics;

respectively extracting packet loss rate and retransmission times from the target shared data, and calculating a shared efficiency characteristic according to the packet loss rate, the retransmission times, the bandwidth parameter, the transmission data type, the transmission time stamp and the transmission data;

all the sharing behavior features are collected into a sharing behavior feature set, all the sharing strategy features are collected into a sharing strategy feature set, and all the sharing efficiency features are collected into a sharing efficiency feature set.

Optionally, the calculating the sharing efficiency feature according to the packet loss rate, the retransmission times, the bandwidth parameter, the transmission data type, the transmission time stamp, and the transmission data includes:

Calculating a sharing efficiency characteristic according to the packet loss rate, the retransmission times, the bandwidth parameter, the transmission data type, the transmission time stamp and the transmission data by using the following sharing efficiency formula:

wherein X refers to the sharing efficiency characteristic, alpha refers to the bandwidth parameter, s refers to the transmission data amount, d refers to the packet loss rate, c refers to the retransmission times, Y ₂ Refers to the end timestamp in the transmission timestamps, Y ₁ Refers to the start timestamp of the transmission timestamps, and beta refers to the transmission data type.

Optionally, training a preset behavior time sequence model according to the behavior feature sequence set into a behavior analysis model, including:

performing sliding convolution on the behavior characteristic sequence set by using a preset behavior time sequence model to obtain a behavior time sequence characteristic set;

calculating a short-term behavioral feature set of the behavioral time series feature set by using the following short-term behavioral feature algorithm:

r _i,t ＝σ(x _i,t ·W ₁ +h _i,t-1 ·W ₂ +b ₁ )

u _i,t ＝σ(x _i,t ·W ₃ +h _i,t-1 ·W ₄ +b ₂ )

wherein r is _i,t Means that the component update feature of the ith short-term behavior feature in the short-term behavior feature set at the t moment is shown as sigma, the sigma is a sigmoid activation function, t is a time sequence number, and x _i,t Is the behavior time sequence characteristic input at the t moment of the ith behavior time sequence characteristic in the behavior time sequence characteristic set, W ₁ 、W ₂ Respectively a first weight and a second weight of the behavior time sequence model, h _i,t-1 Refers to the short-term behavioral characteristics of the ith short-term behavioral characteristic in the short-term behavioral characteristics set at time t-1, b ₁ Is the update weight of the behavior time sequence model, u _i,t Means that the component reset feature, W, of the ith short-term behavioral feature at the t-th moment in the short-term behavioral feature set ₃ 、W ₄ The third weight and the fourth weight of the behavior time sequence model are respectively b ₂ Is the reset weight of the behavior time sequence model, c _i,t Means that the component memory feature of the ith short-term behavior feature in the short-term behavior feature set at the t moment is RELU as an activation function, W ₅ 、W ₆ A fifth weight and a sixth weight of the behavioral timing model,is the product of elements, b ₃ Is the memory weight of the behavior time sequence model, h _t Means the short-term behavioral characteristics of the ith short-term behavioral characteristic in the short-term behavioral characteristic set at time t;

calculating a long-term behavioral feature set of the behavioral time series feature set by using the following long-term behavioral feature algorithm:

wherein,means that the i-th long-term behavior feature in the set of long-term behavior features is a jump update feature at time t,/for the time point>Means the long-term behavior feature of the ith long-term behavior feature in the long-term behavior feature set at the t-p time, p being the number of hidden units in the jump-home layer of the behavior time sequence model,/ >Means that the i-th long-term behavior feature in the set of long-term behavior features is a skip reset feature at time t,/for>Means that the i-th long-term behavior feature in the long-term behavior feature set is a jump memory feature at the t-th moment,/for>Means the long-term behavioral characteristics of the ith long-term behavioral characteristic in the long-term behavioral characteristics set at time t;

the short-term behavior feature set and the long-term behavior feature set are fully connected and fused into a standard time sequence behavior feature set;

calculating a time sequence loss value of the behavior characteristic sequence set and the standard time sequence behavior characteristic set, and carrying out iterative updating on model parameters of the behavior time sequence model according to the time sequence loss value to obtain a behavior analysis model.

Optionally, the calculating, according to the sharing shunt model, a secondary sharing policy corresponding to the secondary behavior feature includes:

generating random strategy features by using a preset random number algorithm, and generating a random parameter feature group according to the random strategy features and the secondary behavior features;

calculating random efficiency characteristics corresponding to the random parameter characteristic groups according to the sharing shunt model;

carrying out iterative updating on the random efficiency characteristic by using a simulated annealing algorithm to obtain a standard efficiency characteristic;

And carrying out policy parameter configuration on the target mobile equipment according to the standard efficiency characteristics to obtain a secondary sharing policy.

Optionally, the identifying privacy data from the real-time streaming data includes:

performing data classification on the real-time shunt data according to a preset data type to obtain text shunt data, form shunt data and picture shunt data;

extracting a form structure from the form split data, and performing privacy matching on the form structure to obtain a privacy structure group;

form privacy data corresponding to the privacy structure group are screened from the form split data;

sequentially performing text word segmentation and stop word screening operation on the text split data to obtain a standard text split word set;

vectorizing the standard text split word set into a text word vector set, and carrying out privacy word matching on the text word vector set to obtain a privacy word vector set;

mapping the standard text splitting word set by using the privacy word vector set to obtain text privacy data;

carrying out picture convolution, position coding and attention coding on the picture shunt data in sequence to obtain picture feature codes;

Performing privacy decoding on the picture feature codes by using a self-attention mechanism to obtain picture privacy data;

and collecting the form privacy data, the text privacy data and the picture privacy data into a sensitive data set.

Optionally, the privacy encrypting the real-time shunt data into real-time transmission data according to the privacy data includes:

selecting data in the privacy data one by one as target data, and sequentially performing data blocking and data expansion operation on the target data to obtain target expanded data blocks;

sequentially performing column confusion and round key encryption operation on the target extended data block by using a preset encryption key to obtain target encrypted data;

adding preset mark placeholders at the front and rear positions of the target encrypted data to obtain target replacement data;

and replacing target data in the real-time shunt data by using the target replacement data until the target data is the last data in the privacy data, and taking the updated real-time shunt data as real-time encryption data.

In order to solve the above problem, the present invention further provides a data offloading system for mobile-end cluster hotspot sharing, where the system includes:

The data splitting module is used for acquiring a hot spot sharing log set of the mobile terminal cluster, cleaning the hot spot sharing log set data into a standard sharing log set, and splitting the standard sharing log set time sequence into a hot spot log sequence set;

the split training module is configured to extract a sharing behavior feature set, a sharing policy feature set, and a sharing efficiency feature set from the standard sharing log set, respectively, train a preset sharing decision model into a sharing split model according to the sharing behavior feature set, the sharing policy feature set, and the sharing efficiency feature set, and train the preset sharing decision model into a sharing split model according to the sharing behavior feature set, the sharing policy feature set, and the sharing efficiency feature set, where the training includes: combining the sharing behavior feature set and the sharing strategy feature set into a sharing parameter feature set, and calculating a feature gain set of the sharing parameter feature set according to a preset sharing decision model by using the following gain algorithm:

wherein G is _i For the ith feature gain in the feature gain set, i and N are feature group numbers, N is the total number of shared parameter feature groups in the shared parameter feature group set, the total number of shared parameter feature groups in the shared parameter feature group set is equal to the total number of feature gains in the feature gain set, gamma and psi are preset constants, and G _i,L For the sum of the first partial derivatives of the data contained in the left node of the root node when the input parameter of the shared decision model is the ith shared parameter feature group in the shared parameter feature group, H _i,L For the sum of the second partial derivatives of the data contained in the left node of the root node when the input parameters of the shared decision model are the ith shared parameter feature group in the shared parameter feature group, G _i,R Root for the shared decision model when input parameters are the ith shared parameter feature set of the shared parameter feature setSum of first partial derivatives of data contained in right node of nodes, H _i,R For the sum of the second partial derivatives of the data contained in the right node of the root node when the input parameters of the shared decision model are the ith shared parameter feature group in the shared parameter feature group, G _n,L For the sum of the first partial derivatives of the data contained in the left node of the root node when the input parameter of the shared decision model is the nth shared parameter feature group in the shared parameter feature group, H _n,L For the sum of the second partial derivatives of the data contained in the left node of the root node when the input parameters of the shared decision model are the nth shared parameter feature group in the shared parameter feature group, G _n,R For the sum of the first partial derivatives of the data contained in the right node of the root node when the input parameter is the nth shared parameter feature group in the shared parameter feature group of the shared decision model, H _n,R Adding the second partial derivatives of the data contained in the right node of the root node when the input parameter of the sharing decision model is the nth sharing parameter feature group in the sharing parameter feature group, wherein max is a maximum value function symbol, and min is a minimum value function symbol; performing iterative node updating on the shared decision model according to the characteristic gain set to obtain a secondary decision model; calculating a prediction efficiency characteristic set corresponding to the shared parameter characteristic set by using the secondary decision model; calculating an efficiency loss value between the predicted efficiency feature set and the shared efficiency feature set, and carrying out iterative updating on model parameters of the secondary decision model according to the efficiency loss value to obtain a shared shunt model;

the behavior training module is used for generating a behavior feature sequence set according to the hot spot log sequence set and the shared behavior feature set, and training a preset behavior time sequence model into a behavior analysis model according to the behavior feature sequence set;

The strategy configuration module is used for selecting devices in the mobile terminal cluster one by one as target mobile devices, acquiring real-time logs of the target mobile devices, calculating secondary behavior characteristics corresponding to the real-time logs by using the behavior analysis model, and calculating secondary sharing strategies corresponding to the secondary behavior characteristics according to the sharing shunt model;

and the encryption transmission module is used for acquiring the real-time shunt data of the target mobile equipment, identifying privacy data from the real-time shunt data, carrying out privacy encryption on the real-time shunt data into real-time transmission data according to the privacy data, and carrying out sharing transmission on the transmission data according to the secondary sharing strategy.

According to the method, the hotspot sharing log set data of the mobile terminal cluster are acquired, the hotspot sharing log set data are cleaned to form the standard sharing log set, the standard sharing log set time sequence is split to form the hotspot log sequence set, the accuracy of the data can be improved, the log data are rearranged to be grouped according to the equipment number and the time sequence, the subsequent extraction of time sequence features is facilitated, the sharing behavior feature set, the sharing strategy feature set and the sharing efficiency feature set are respectively extracted from the standard sharing log set, the preset sharing decision model is trained to form the sharing shunt model according to the sharing behavior feature set, the sharing strategy feature set and the sharing efficiency feature set, the influence relation of the sharing behavior feature of the user and the sharing strategy feature on the sharing efficiency in the data shunt process can be determined, the behavior feature sequence set is generated according to the hotspot log sequence set and the sharing behavior feature set, the preset behavior time sequence model is trained to form the analysis model according to the behavior feature sequence set, the behavior feature of each equipment in the mobile terminal cluster can be extracted, and the future behavior feature change of the equipment in the time period can be conveniently predicted in the future time period.

The method comprises the steps of selecting devices in a mobile terminal cluster one by one as target mobile devices, obtaining real-time logs of the target mobile devices, calculating secondary behavior characteristics corresponding to the real-time logs by using the behavior analysis model, calculating secondary sharing strategies corresponding to the secondary behavior characteristics according to the sharing distribution model, predicting the sharing behavior characteristics of the target mobile devices in a future time period, selecting the sharing strategy with the maximum transmission efficiency for data transmission, saving the waiting time of a system, improving the data distribution efficiency, recognizing privacy data from the real-time distribution data by obtaining the real-time distribution data of the target mobile devices, privacy encrypting the real-time distribution data into real-time transmission data according to the privacy data, and carrying out sharing transmission on the transmission data according to the secondary sharing strategies, so that the privacy data in the distribution data can be flexibly recognized and encrypted for transmission, the whole encryption transmission method in the data distribution is avoided, the time loss caused by encryption is reduced, the safety in the data distribution is ensured, and the data distribution efficiency is improved. Therefore, the data distribution method and the system for mobile terminal cluster hot spot sharing can solve the problem of lower efficiency in data distribution.

Drawings

Fig. 1 is a flow chart of a data splitting method for mobile end cluster hot spot sharing according to an embodiment of the present application;

FIG. 2 is a flow chart of extracting a sharing behavior feature set, a sharing policy feature set and a sharing efficiency feature set according to an embodiment of the present application;

FIG. 3 is a flow chart of generating real-time transmission data according to an embodiment of the present application;

FIG. 4 is a functional block diagram of a data distribution system for mobile end cluster hot spot sharing according to an embodiment of the present application;

the achievement of the objects, functional features and advantages of the present application will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

The embodiment of the application provides a data distribution method for mobile terminal cluster hot spot sharing. The execution main body of the data distribution method for mobile terminal cluster hot spot sharing includes, but is not limited to, at least one of a server, a terminal and the like capable of being configured to execute the method provided by the embodiment of the application. In other words, the data offloading method of mobile terminal cluster hotspot sharing may be performed by software or hardware installed in a terminal device or a server device, where the software may be a blockchain platform. The service end includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like. The server may be an independent server, or may be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and basic cloud computing services such as big data and artificial intelligence platforms.

Referring to fig. 1, a flow chart of a data splitting method for mobile end cluster hot spot sharing according to an embodiment of the invention is shown. In this embodiment, the data splitting method for mobile terminal cluster hotspot sharing includes:

s1, acquiring a hot spot sharing log set of a mobile terminal cluster, cleaning data of the hot spot sharing log set into a standard sharing log set, and splitting a time sequence of the standard sharing log set into a hot spot log sequence set.

In the embodiment of the present invention, the mobile terminal cluster refers to a cluster formed by a group of mobile devices, where the mobile devices may be smart phones, tablet computers, notebook computers, and the like, the mobile devices have a wireless network connection function, each of the hotspot sharing logs in the hotspot sharing log set corresponds to a hotspot sharing log of one mobile terminal device in the mobile terminal cluster, the hotspot sharing log is a log generated during hotspot sharing, and the hotspot sharing refers to a situation that a plurality of users or devices can share one wireless network hotspot in a specific area or location to access an internet or a local area network resource, and the hotspot is usually one wireless local area network hotspot, so that a plurality of users are allowed to share the same network connection through the wireless connection device.

In detail, the hot spot sharing log refers to a log file for recording and tracking hot spot sharing activities, and the log file comprises a connection log, a traffic log, an error log, a user activity log, a performance log, a security log and the like.

In the embodiment of the present invention, the cleaning the hotspot sharing log set data to a standard sharing log set includes:

In detail, the repeated shared log refers to a log in which repeated data appear in the hot spot shared log set, the repeated shared log can be screened out from the hot spot shared log set by using a hash code detection method to obtain a duplicate removal shared log set, the offside shared log refers to a log in which error value domain data appear in the hot spot shared log set, the messy code shared log refers to a log in which messy code data appear in the hot spot shared log set, and the offside shared log and the messy code shared log can be respectively extracted from the duplicate removal shared log set by using a value domain matching or key character detection method.

Specifically, a k-means clustering method or a hierarchical clustering algorithm can be utilized to perform log feature clustering on the primary shared log set to obtain a clustered log class set; and replacing the offside sharing logs and the messy code sharing logs in the duplicate removal sharing log set according to the clustering log class set to obtain a standard sharing log, wherein the step of selecting the offside sharing logs or the messy code sharing logs in the duplicate removal sharing log set one by one as target logs, taking the clustering log class corresponding to the target log in the clustering log class set as target clustering log class, taking the clustering center of the target clustering log class as target clustering center, taking the logs corresponding to the target clustering center as target center logs, and replacing the target logs by utilizing the target center logs to obtain the standard sharing log.

In detail, the splitting the standard shared log set time sequence into a hot log sequence set includes:

and collecting all the hot spot log sequences into a hot spot log sequence set.

Specifically, the time domain period may be one hour or half an hour, and splitting each log in the target ordering shared log according to a preset time domain period refers to taking a corresponding log in each time domain period in the target ordering shared log as a hot spot log, and collecting all the hot spot logs into a hot spot log sequence according to a time sequence, for example, taking a corresponding log in 14:00 to 15:00 of 22 years, 4 months and 5 days as a hot spot log in one time domain period when the time domain period is one hour.

In the embodiment of the invention, the hotspot sharing log set of the mobile terminal cluster is obtained, the hotspot sharing log set data is cleaned into the standard sharing log set, the standard sharing log set time sequence is split into the hotspot log sequence set, the accuracy of the data can be improved, and the log data is reprogrammed into groups according to the equipment number and the time sequence, so that the subsequent extraction of the time sequence characteristics is convenient.

S2, respectively extracting a sharing behavior feature set, a sharing strategy feature set and a sharing efficiency feature set from the standard sharing log set, and training a preset sharing decision model into a sharing shunt model according to the sharing behavior feature set, the sharing strategy feature set and the sharing efficiency feature set.

In the embodiment of the present invention, each shared behavior feature in the shared behavior feature set corresponds to a behavior feature of each standard shared log in the standard shared log set, for example, a behavior timestamp of receiving and sending data, a size of transmitted data, and the like during hot spot sharing; each sharing policy feature in the sharing policy feature set corresponds to a policy feature of each standard sharing log in the standard sharing log set, for example, a bandwidth resource size, a processor operating frequency, whether to perform data compression, data segment transmission, and the like; each sharing efficiency characteristic in the sharing efficiency characteristic set corresponds to an efficiency characteristic of each standard sharing log in the standard sharing log set, such as a transmission speed, a success rate, a delay, and the like at each transmission.

In the embodiment of the present invention, referring to fig. 2, the extracting a sharing behavior feature set, a sharing policy feature set and a sharing efficiency feature set from the standard sharing log set includes:

s21, selecting standard shared logs in the standard shared log set one by one as target shared logs, and performing format transcoding on the target shared logs to obtain target shared data;

S22, respectively extracting a transmission time stamp, a transmission data amount, a transmission direction and a transmission data type from the target shared data, and encoding the transmission time stamp, the transmission data amount, the transmission direction and the transmission data type into a shared behavior characteristic;

s23, respectively extracting bandwidth parameters, frequency parameters, compression parameters and segmentation parameters from the target shared data, and encoding the bandwidth parameters, the frequency parameters, the compression parameters and the segmentation parameters into sharing strategy characteristics;

s24, respectively extracting packet loss rate and retransmission times from the target shared data, and calculating a shared efficiency characteristic according to the packet loss rate, the retransmission times, the bandwidth parameter, the transmission data type, the transmission time stamp and the transmission data;

s25, integrating all the sharing behavior features into a sharing behavior feature set, integrating all the sharing strategy features into a sharing strategy feature set, and integrating all the sharing efficiency features into a sharing efficiency feature set.

Specifically, the performing format transcoding on the target shared log to obtain target shared data includes extracting a log format of the target shared log, and performing log analysis on the target shared log according to the log format to obtain target shared data, where the log format may be a log format such as XML, JSON, etc., and log analysis may be performed on the target shared log according to the log format by using a log analysis tool such as Logstash, fluentd, flume, etc., to obtain target shared data.

In detail, the transmission time stamp refers to a start time stamp and a corresponding end time stamp of data transmission, the transmission data amount refers to a capacity size of all data transmitted in a data transmission process, the transmission direction refers to a direction of data flow in a corresponding data transmission process, the transmission data type refers to a data type of each transmission data in a corresponding data transmission process, and the encoding of the transmission time stamp, the transmission data amount, the transmission direction and the transmission data type into a sharing behavior feature refers to sequential vectorization encoding of the transmission time stamp, the transmission data amount, the transmission direction and the transmission data type and splicing into the sharing behavior feature according to a fixed arrangement sequence.

In detail, the bandwidth parameter refers to the bandwidth capacity in the data transmission process, the frequency parameter refers to the working frequency of a processor of the mobile terminal device in the data transmission process, the compression parameter refers to the algorithm type of whether data compression and data compression are performed in the transmission process, and the segmentation parameter refers to whether data segmentation and each segment size of data segmentation are performed in the data transmission process.

Specifically, the extracting the transmission time stamp, the transmission data amount, the transmission direction and the transmission data type from the target shared data respectively includes:

extracting transmission data quantity and transmission data type from the target shared data by using a keyword matching method;

respectively extracting a start time stamp and an end time stamp from the target shared data, and generating a transmission time stamp according to the start time stamp and the end time stamp;

and respectively extracting a start address and an end address from the target shared data, and generating a transmission direction according to the start address and the end address.

In detail, the start timestamp is a timestamp corresponding to when data transmission starts, the end timestamp is a timestamp corresponding to when data transmission ends, the generating of the transmission timestamp according to the start timestamp and the end timestamp is to arrange the end timestamp and the start timestamp into the transmission timestamp according to a fixed sequence, the start address is an address of a data sender, the end address is an address of a data receiver, the generating of the transmission direction according to the start address and the end address is to arrange and encode the start address and the end address according to a fixed sequence, the packet loss rate is a ratio of the number of lost data packets to the total number of transmitted data packets in a data transmission process, and the retransmission times are the times of retransmission in a transmission process.

In detail, the calculating the sharing efficiency characteristic according to the packet loss rate, the retransmission times, the bandwidth parameter, the transmission data type, the transmission time stamp, and the transmission data includes:

wherein X refers to the sharing efficiency characteristic, alpha refers to the bandwidth parameter, s refers to the transmission data amount, d refers to the packet loss rate, c refers to the retransmission times, Y ₂ Refers to the end timestamp in the transmission timestamps, Y ₁ Refers to the start time in the transmission time stampAnd the timestamp, beta refers to the transmission data type.

In detail, by calculating the sharing efficiency characteristic according to the packet loss rate, the retransmission times, the bandwidth parameter, the transmission data type, the transmission time stamp and the transmission data type by using the sharing efficiency formula, the transmission efficiency can be determined by combining the success rate in data transmission and the transmission speed and the data type of each group of data, and the method has wider applicability than the single packet loss rate calculation.

In particular, the sharing decision model may be a decision tree model or an XGBoost model, the inputs of the sharing decision model are the sharing behavior features in the sharing behavior feature set and the sharing policy features in the sharing policy feature set, and the outputs of the sharing decision model are the sharing efficiency features in the sharing efficiency feature set.

In detail, training a preset sharing decision model into a sharing shunt model according to the sharing behavior feature set, the sharing policy feature set and the sharing efficiency feature set includes:

combining the sharing behavior feature set and the sharing strategy feature set into a sharing parameter feature set, and calculating a feature gain set of the sharing parameter feature set according to a preset sharing decision model by using the following gain algorithm:

wherein G is _i For the ith feature gain in the feature gain set, i and N are feature group numbers, N is the total number of shared parameter feature groups in the shared parameter feature group set, the total number of shared parameter feature groups in the shared parameter feature group set is equal to the total number of feature gains in the feature gain set, gamma and psi are preset constants, and G _i,L For the sum of the first partial derivatives of the data contained in the left node of the root node when the input parameter of the shared decision model is the ith shared parameter feature group in the shared parameter feature group, H _i,L For the sum of the second partial derivatives of the data contained in the left node of the root node when the input parameters of the shared decision model are the ith shared parameter feature group in the shared parameter feature group, G _i,R For the sum of the first partial derivatives of the data contained in the right node of the root node when the input parameter is the ith shared parameter feature group in the shared parameter feature group of the shared decision model, H _i,R For the sum of the second partial derivatives of the data contained in the right node of the root node when the input parameters of the shared decision model are the ith shared parameter feature group in the shared parameter feature group, G _n,L For the sum of the first partial derivatives of the data contained in the left node of the root node when the input parameter of the shared decision model is the nth shared parameter feature group in the shared parameter feature group, H _n,L For the sum of the second partial derivatives of the data contained in the left node of the root node when the input parameters of the shared decision model are the nth shared parameter feature group in the shared parameter feature group, G _n,R For the sum of the first partial derivatives of the data contained in the right node of the root node when the input parameter is the nth shared parameter feature group in the shared parameter feature group of the shared decision model, H _n,R Adding the second partial derivatives of the data contained in the right node of the root node when the input parameter of the sharing decision model is the nth sharing parameter feature group in the sharing parameter feature group, wherein max is a maximum value function symbol, and min is a minimum value function symbol;

performing iterative node updating on the shared decision model according to the characteristic gain set to obtain a secondary decision model;

calculating a prediction efficiency characteristic set corresponding to the shared parameter characteristic set by using the secondary decision model;

and calculating an efficiency loss value between the predicted efficiency characteristic set and the shared efficiency characteristic set, and carrying out iterative updating on model parameters of the secondary decision model according to the efficiency loss value to obtain a shared shunt model.

In detail, the merging the sharing behavior feature set and the sharing policy feature set into a sharing parameter feature set refers to merging each sharing feature in the sharing behavior feature set and a corresponding sharing policy feature in the sharing policy feature set into a sharing parameter feature set, and collecting all sharing parameter feature sets into a sharing parameter feature set.

Specifically, the step of updating the sharing decision model by iteration nodes according to the feature gain set to obtain a secondary decision model refers to the step of iteratively splitting sub-nodes of the sharing decision model according to features in a sharing parameter feature group corresponding to the feature gain with the largest value in the feature gain set to obtain the secondary decision model.

In detail, an efficiency loss value between the prediction efficiency feature set and the shared efficiency feature set may be calculated by using a cross entropy loss value function, and model parameters of the secondary decision model may be iteratively updated according to the efficiency loss value by using a random gradient descent algorithm to obtain a shared shunt model.

In the embodiment of the invention, the sharing behavior feature set, the sharing strategy feature set and the sharing efficiency feature set are respectively extracted from the standard sharing log set, and a preset sharing decision model is trained into a sharing distribution model according to the sharing behavior feature set, the sharing strategy feature set and the sharing efficiency feature set, so that the influence relationship of the sharing behavior feature and the sharing strategy feature of a user on the sharing efficiency in the data distribution process can be determined.

S3, generating a behavior feature sequence set according to the hot spot log sequence set and the shared behavior feature set, and training a preset behavior time sequence model into a behavior analysis model according to the behavior feature sequence set.

In the embodiment of the present invention, each behavior feature sequence in the behavior feature sequence set corresponds to each hotspot log sequence in the hotspot log sequence set, each behavior feature in the behavior feature sequence corresponds to one hotspot log in the hotspot log sequence, and generating the behavior feature sequence set according to the hotspot log sequence set and the shared behavior feature set refers to mapping each shared behavior feature in the shared behavior feature set to each hotspot log sequence in the hotspot log sequence set, so as to obtain a behavior feature sequence set, where the behavior time sequence model may be an LSTM model or a time sequence circulation neural network model.

In the embodiment of the present invention, training a preset behavior time sequence model according to the behavior feature sequence set into a behavior analysis model includes:

r _i,t ＝σ(x _i,t ·W ₁ +h _i,t-1 ·W ₂ +b ₁ )

u _i,t ＝σ(x _i,t ·W ₃ +h _i,t-1 ·W ₄ +b ₂ )

wherein r is _i,t Means that the component update feature of the ith short-term behavior feature in the short-term behavior feature set at the t moment is shown as sigma, the sigma is a sigmoid activation function, t is a time sequence number, and x _i,t Is the behavior time sequence characteristic input at the t moment of the ith behavior time sequence characteristic in the behavior time sequence characteristic set, W ₁ 、W ₂ Respectively a first weight and a second weight of the behavior time sequence model, h _i,t-1 Refers to whatThe short-term behavioral characteristics at time t-1 of the ith short-term behavioral characteristics in the set of short-term behavioral characteristics, b ₁ Is the update weight of the behavior time sequence model, u _i,t Means that the component reset feature, W, of the ith short-term behavioral feature at the t-th moment in the short-term behavioral feature set ₃ 、W ₄ The third weight and the fourth weight of the behavior time sequence model are respectively b ₂ Is the reset weight of the behavior time sequence model, c _i,t Means that the component memory feature of the ith short-term behavior feature in the short-term behavior feature set at the t moment is RELU as an activation function, W ₅ 、W ₆ A fifth weight and a sixth weight of the behavioral timing model,is the product of elements, b ₃ Is the memory weight of the behavior time sequence model, h _t Means the short-term behavioral characteristics of the ith short-term behavioral characteristic in the short-term behavioral characteristic set at time t;

wherein,means that the i-th long-term behavior feature in the set of long-term behavior features is a jump update feature at time t,/for the time point>Means the long-term behavior feature of the ith long-term behavior feature in the long-term behavior feature set at the t-p time, p being the number of hidden units in the jump-home layer of the behavior time sequence model,/>Means that the i-th long-term behavior feature in the set of long-term behavior features is a skip reset feature at time t,/for>Means that the i-th long-term behavior feature in the long-term behavior feature set is a jump memory feature at the t-th moment,/for>Means the long-term behavioral characteristics of the ith long-term behavioral characteristic in the long-term behavioral characteristics set at time t;

In detail, the short-term behavior feature set of the behavior time sequence feature set is calculated by using the short-term behavior feature algorithm, the long-term behavior feature set of the behavior time sequence feature set is calculated by using the long-term behavior feature algorithm, and the time sequence features in the behavior time sequence feature set can be extracted by combining a gating recursion unit and a gating reset unit in the behavior time sequence model, so that the accuracy of behavior feature prediction is improved.

Specifically, the calculating the time sequence loss values of the behavior feature sequence set and the standard time sequence behavior feature set, performing iterative updating on the model parameters of the behavior time sequence model according to the time sequence loss values, and performing iterative updating on the model parameters of the secondary decision model according to the efficiency loss values, wherein the method for obtaining the behavior analysis model is consistent with the method for obtaining the efficiency loss values between the prediction efficiency feature set and the sharing efficiency feature set calculated in the step S2, and the method for obtaining the sharing shunt model is not repeated here.

According to the embodiment of the invention, the behavior characteristic sequence set is generated according to the hot spot log sequence set and the shared behavior characteristic set, and the preset behavior time sequence model is trained into the behavior analysis model according to the behavior characteristic sequence set, so that the change rule of the behavior characteristics of each device in the mobile terminal cluster along with time can be extracted, and further, the behavior characteristic change in the future time period of the device can be conveniently predicted.

S4, selecting devices in the mobile terminal cluster one by one as target mobile devices, acquiring real-time logs of the target mobile devices, calculating secondary behavior characteristics corresponding to the real-time logs by using the behavior analysis model, and calculating secondary sharing strategies corresponding to the secondary behavior characteristics according to the sharing shunt model.

In the embodiment of the invention, the real-time log refers to a device log generated by the target mobile device in real time, wherein the real-time log of the target mobile device can be obtained by using a buried point method.

In the embodiment of the present invention, the calculating, according to the sharing shunt model, a secondary sharing policy corresponding to the secondary behavior feature includes:

Specifically, the random number algorithm may be a linear congruence method, a random number library function or a gaussian distribution random number generation method, and the method for generating the random parameter feature set according to the random policy feature and the secondary behavior feature is consistent with the method for combining the shared behavior feature set and the shared policy feature set into the shared parameter feature set in the step S2, which is not described herein.

In detail, the step of iteratively updating the random efficiency features by using a simulated annealing algorithm to obtain standard efficiency features refers to selecting a random efficiency feature with the largest numerical value in the random efficiency features as the standard efficiency feature in iteration times corresponding to the simulated annealing algorithm.

In the embodiment of the invention, the devices in the mobile terminal cluster are selected one by one as the target mobile devices, the real-time logs of the target mobile devices are obtained, the secondary behavior characteristics corresponding to the real-time logs are calculated by using the behavior analysis model, the secondary sharing strategy corresponding to the secondary behavior characteristics is calculated according to the sharing shunt model, the sharing behavior characteristics of the target mobile devices in the future time period can be predicted, and the sharing strategy with the maximum transmission efficiency is selected for data transmission, so that the waiting time of the system is saved, and meanwhile, the data shunting efficiency is improved.

S5, acquiring real-time shunt data of the target mobile device, identifying privacy data from the real-time shunt data, privacy encrypting the real-time shunt data into real-time transmission data according to the privacy data, and carrying out sharing transmission on the transmission data according to the secondary sharing strategy.

In the embodiment of the present invention, the real-time streaming data refers to data that needs to be transmitted by the target mobile device in real time, and the private data refers to data related to user privacy in the real-time streaming data, such as user identity information, user address information, and the like.

In an embodiment of the present invention, the identifying the privacy data from the real-time streaming data includes:

In detail, the form structure refers to a composition structure of a form and a definition structure of each field, and the form structure can be extracted from the form log by using a database model or an API interface; privacy matching is performed on the list structure to obtain a privacy structure group, namely privacy matching is performed on attribute items of various fields of the list, all the privacy items are collected into a sensitive structure group, for example, attribute items corresponding to an address are used as privacy items in the list structure, wherein privacy matching can be performed on the list structure by using a regular expression or a character string searching method to obtain the privacy structure group.

In detail, text word segmentation can be performed on the text split data by using a bidirectional maximum matching algorithm or a forward maximum matching algorithm, stop word screening operation can be performed on the text split data by using a stop word list matching method, the standard text split word set can be vectorized into a text word vector set by using word2Vec, glove and other coding modes, and privacy word matching can be performed on the text word vector set by using an European feature distance algorithm to obtain a privacy word vector set.

Specifically, the VGG-16 model can be utilized to carry out picture convolution on the picture shunt data, the coding layer of the transform model can be utilized to carry out position coding and attention coding to obtain picture feature coding, and the multistage linear residual network can be utilized to carry out privacy decoding on the picture feature coding by utilizing a self-attention mechanism to obtain picture privacy data.

In detail, referring to fig. 3, the privacy encrypting the real-time streaming data into real-time transmission data according to the privacy data includes:

s31, selecting data in the privacy data one by one as target data, and sequentially performing data blocking and data expansion operation on the target data to obtain a target expansion data block;

S32, sequentially performing column confusion and round key encryption operation on the target expanded data block by using a preset encryption key to obtain target encrypted data;

s33, adding a preset mark placeholder at the front and rear positions of the target encrypted data to obtain target replacement data;

s34, replacing target data in the real-time shunt data by using the target replacement data until the target data is the last data in the privacy data, and taking the updated real-time shunt data as real-time encryption data.

In detail, the data partitioning refers to dividing the target data into a plurality of data blocks according to a fixed byte length, the data expansion refers to adding 0 characters at the end of the last surplus data of the partitioning to reach the fixed byte length, the column confusion (column permutation) is a data processing technology for randomly arranging or reordering columns in a data set, the round key addition (AddRoundKey) refers to performing exclusive or operation on the output data blocks of the round function and corresponding round keys in an expanded key array, and the flag placeholder is used for prompting a receiver that the data is encrypted data and needs privacy decryption.

In the embodiment of the invention, the privacy data is identified from the real-time shunt data by acquiring the real-time shunt data of the target mobile device, the real-time shunt data is encrypted into the real-time transmission data according to the privacy data, and the transmission data is shared and transmitted according to the secondary sharing strategy, so that the privacy data in the shunt data can be flexibly identified and encrypted for transmission, the method of integral encryption transmission in data shunt is avoided, the time loss caused by encryption is reduced, the safety in data shunt is ensured, and the data shunt efficiency is improved.

The method comprises the steps of selecting devices in a mobile terminal cluster one by one as target mobile devices, obtaining real-time logs of the target mobile devices, calculating secondary behavior characteristics corresponding to the real-time logs by using the behavior analysis model, calculating secondary sharing strategies corresponding to the secondary behavior characteristics according to the sharing distribution model, predicting the sharing behavior characteristics of the target mobile devices in a future time period, selecting the sharing strategy with the maximum transmission efficiency for data transmission, saving the waiting time of a system, improving the data distribution efficiency, recognizing privacy data from the real-time distribution data by obtaining the real-time distribution data of the target mobile devices, privacy encrypting the real-time distribution data into real-time transmission data according to the privacy data, and carrying out sharing transmission on the transmission data according to the secondary sharing strategies, so that the privacy data in the distribution data can be flexibly recognized and encrypted for transmission, the whole encryption transmission method in the data distribution is avoided, the time loss caused by encryption is reduced, the safety in the data distribution is ensured, and the data distribution efficiency is improved. Therefore, the data distribution method for mobile terminal cluster hot spot sharing can solve the problem of lower efficiency in data distribution.

Fig. 4 is a functional block diagram of a data distribution system for mobile end cluster hot spot sharing according to an embodiment of the present invention.

The data distribution system 100 for mobile terminal cluster hotspot sharing in the present invention may be installed in an electronic device. Depending on the implementation function, the data splitting system 100 shared by the mobile end cluster hotspots may include a data splitting module 101, a splitting training module 102, a behavior training module 103, a policy configuration module 104, and an encryption transmission module 105. The module of the invention, which may also be referred to as a unit, refers to a series of computer program segments, which are stored in the memory of the electronic device, capable of being executed by the processor of the electronic device and of performing a fixed function.

In the present embodiment, the functions concerning the respective modules/units are as follows:

the data splitting module 101 is configured to obtain a hotspot sharing log set of the mobile terminal cluster, clean the hotspot sharing log set data into a standard sharing log set, and split the standard sharing log set time sequence into a hotspot log sequence set;

the split training module 102 is configured to extract a sharing behavior feature set, a sharing policy feature set, and a sharing efficiency feature set from the standard sharing log set, respectively, train a preset sharing decision model into a sharing split model according to the sharing behavior feature set, the sharing policy feature set, and the sharing efficiency feature set, and train the preset sharing decision model into a sharing split model according to the sharing behavior feature set, the sharing policy feature set, and the sharing efficiency feature set, where the training includes: combining the sharing behavior feature set and the sharing strategy feature set into a sharing parameter feature set, and calculating a feature gain set of the sharing parameter feature set according to a preset sharing decision model by using the following gain algorithm:

Wherein G is _i For the ith feature gain in the feature gain set, i and N are feature group numbers, N is the total number of shared parameter feature groups in the shared parameter feature group set, the total number of shared parameter feature groups in the shared parameter feature group set is equal to the total number of feature gains in the feature gain set, gamma and psi are preset constants, and G _i,L For the sum of the first partial derivatives of the data contained in the left node of the root node when the input parameter of the shared decision model is the ith shared parameter feature group in the shared parameter feature group, H _i,L For the sum of the second partial derivatives of the data contained in the left node of the root node when the input parameters of the shared decision model are the ith shared parameter feature group in the shared parameter feature group, G _i,R The ith common node in the feature set with the input parameters being the shared parameters for the shared decision modelThe sum of the first partial derivatives of the data contained in the right node of the root node when sharing the parameter feature set, H _i,R For the sum of the second partial derivatives of the data contained in the right node of the root node when the input parameters of the shared decision model are the ith shared parameter feature group in the shared parameter feature group, G _n,L For the sum of the first partial derivatives of the data contained in the left node of the root node when the input parameter of the shared decision model is the nth shared parameter feature group in the shared parameter feature group, H _n,L For the sum of the second partial derivatives of the data contained in the left node of the root node when the input parameters of the shared decision model are the nth shared parameter feature group in the shared parameter feature group, G _n,R For the sum of the first partial derivatives of the data contained in the right node of the root node when the input parameter is the nth shared parameter feature group in the shared parameter feature group of the shared decision model, H _n,R Adding the second partial derivatives of the data contained in the right node of the root node when the input parameter of the sharing decision model is the nth sharing parameter feature group in the sharing parameter feature group, wherein max is a maximum value function symbol, and min is a minimum value function symbol; performing iterative node updating on the shared decision model according to the characteristic gain set to obtain a secondary decision model; calculating a prediction efficiency characteristic set corresponding to the shared parameter characteristic set by using the secondary decision model; calculating an efficiency loss value between the predicted efficiency feature set and the shared efficiency feature set, and carrying out iterative updating on model parameters of the secondary decision model according to the efficiency loss value to obtain a shared shunt model;

The behavior training module 103 is configured to generate a behavior feature sequence set according to the hotspot log sequence set and the shared behavior feature set, and train a preset behavior time sequence model to a behavior analysis model according to the behavior feature sequence set;

the policy configuration module 104 is configured to select devices in the mobile terminal cluster one by one as target mobile devices, obtain a real-time log of the target mobile devices, calculate secondary behavior features corresponding to the real-time log by using the behavior analysis model, and calculate a secondary sharing policy corresponding to the secondary behavior features according to the sharing and shunting model;

the encryption transmission module 105 is configured to obtain real-time distribution data of the target mobile device, identify privacy data from the real-time distribution data, encrypt the real-time distribution data into real-time transmission data according to the privacy data, and perform sharing transmission on the transmission data according to the secondary sharing policy.

In detail, each module in the data splitting system 100 for sharing the mobile end cluster hot spot in the embodiment of the present invention adopts the same technical means as the data splitting method for sharing the mobile end cluster hot spot in the foregoing fig. 1 to 3, and can generate the same technical effects, which is not repeated here.

In the several embodiments provided in the present invention, it should be understood that the disclosed apparatus, system and method may be implemented in other manners. For example, the system embodiments described above are merely illustrative, e.g., the division of the modules is merely a logical function division, and other manners of division may be implemented in practice.

The modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical units, may be located in one place, or may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional module in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units can be realized in a form of hardware or a form of hardware and a form of software functional modules.

It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims shall not be construed as limiting the claim concerned.

The embodiment of the application can acquire and process the related data based on the artificial intelligence technology. Among these, artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use knowledge to obtain optimal results.

Furthermore, it is evident that the word "comprising" does not exclude other elements or steps, and that the singular does not exclude a plurality. Multiple units or systems set forth in the system embodiments may also be implemented by one unit or system in software or hardware. The terms first, second, etc. are used to denote a name, but not any particular order.

Finally, it should be noted that the above-mentioned embodiments are merely for illustrating the technical solution of the present application and not for limiting the same, and although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made to the technical solution of the present application without departing from the spirit and scope of the technical solution of the present application.

Claims

1. The data distribution method for mobile terminal cluster hot spot sharing is characterized by comprising the following steps:

s1: acquiring a hot spot sharing log set of a mobile terminal cluster, cleaning the hot spot sharing log set data into a standard sharing log set, and splitting the standard sharing log set time sequence into a hot spot log sequence set;

s2: extracting a sharing behavior feature set, a sharing policy feature set and a sharing efficiency feature set from the standard sharing log set respectively, training a preset sharing decision model into a sharing shunt model according to the sharing behavior feature set, the sharing policy feature set and the sharing efficiency feature set, wherein training the preset sharing decision model into the sharing shunt model according to the sharing behavior feature set, the sharing policy feature set and the sharing efficiency feature set comprises:

s21: combining the sharing behavior feature set and the sharing strategy feature set into a sharing parameter feature set, and calculating a feature gain set of the sharing parameter feature set according to a preset sharing decision model by using the following gain algorithm:

s22: performing iterative node updating on the shared decision model according to the characteristic gain set to obtain a secondary decision model;

s23: calculating a prediction efficiency characteristic set corresponding to the shared parameter characteristic set by using the secondary decision model;

s24: calculating an efficiency loss value between the predicted efficiency feature set and the shared efficiency feature set, and carrying out iterative updating on model parameters of the secondary decision model according to the efficiency loss value to obtain a shared shunt model;

s3: generating a behavior feature sequence set according to the hot spot log sequence set and the shared behavior feature set, and training a preset behavior time sequence model into a behavior analysis model according to the behavior feature sequence set;

S4: selecting devices in the mobile terminal cluster one by one as target mobile devices, acquiring real-time logs of the target mobile devices, calculating secondary behavior characteristics corresponding to the real-time logs by using the behavior analysis model, and calculating secondary sharing strategies corresponding to the secondary behavior characteristics according to the sharing shunt model;

s5: and acquiring real-time shunt data of the target mobile equipment, identifying privacy data from the real-time shunt data, privacy encrypting the real-time shunt data into real-time transmission data according to the privacy data, and carrying out sharing transmission on the transmission data according to the secondary sharing strategy.

2. The method for offloading data for hotspot sharing of a mobile terminal cluster according to claim 1, wherein the cleaning the hotspot sharing log set data into a standard sharing log set includes:

3. The method for data splitting for mobile end cluster hot spot sharing according to claim 1, wherein the time sequence of splitting the standard sharing log set into a hot spot log sequence set comprises:

and collecting all the hot spot log sequences into a hot spot log sequence set.

4. The method for data offloading in hot spot sharing of a mobile terminal cluster according to claim 1, wherein the extracting the sharing behavior feature set, the sharing policy feature set, and the sharing efficiency feature set from the standard sharing log set respectively includes:

5. The method for offloading data in a mobile cluster hotspot sharing as set forth in claim 4, wherein the calculating the sharing efficiency characteristic according to the packet loss rate, the retransmission number, the bandwidth parameter, the transmission data type, the transmission timestamp, and the transmission data includes:

6. The method for data offloading in hot spot sharing by a mobile terminal cluster according to claim 1, wherein training a preset behavior time sequence model according to the behavior feature sequence set comprises:

r _i,t ＝o(x _i，t ·W ₁ +h _i,t-1 ·W ₂ +b ₁ )

u _i，t ＝σ(x _i，t ·W ₃ +h _i，t-1 ·W ₄ +b ₂ )

wherein r is _i,t Means that the component update feature of the ith short-term behavior feature in the short-term behavior feature set at the t moment is shown as sigma, the sigma is a sigmoid activation function, t is a time sequence number, and x _i,t Is the behavior time sequence characteristic input at the t moment of the ith behavior time sequence characteristic in the behavior time sequence characteristic set, W ₁ 、W ₂ The behavior time sequence modelAnd a second weight, h _i,t-1 Refers to the short-term behavioral characteristics of the ith short-term behavioral characteristic in the short-term behavioral characteristics set at time t-1, b ₁ Is the update weight of the behavior time sequence model, u _i,t Means that the component reset feature, W, of the ith short-term behavioral feature at the t-th moment in the short-term behavioral feature set ₃ 、W ₄ The third weight and the fourth weight of the behavior time sequence model are respectively b ₂ Is the reset weight of the behavior time sequence model, c _i,t Means that the component memory feature of the ith short-term behavior feature in the short-term behavior feature set at the t moment is RELU as an activation function, W ₅ 、W ₆ A fifth weight and a sixth weight of the behavioral timing model,is the product of elements, b ₃ Is the memory weight of the behavior time sequence model, h _t Means the short-term behavioral characteristics of the ith short-term behavioral characteristic in the short-term behavioral characteristic set at time t;

wherein the method comprises the steps of，Refers to the jump update feature of the ith long-term behavioral feature at time t in the long-term behavioral feature set, Means the long-term behavior feature of the ith long-term behavior feature in the long-term behavior feature set at the t-p time, p being the number of hidden units in the jump-home layer of the behavior time sequence model,/>Means that the i-th long-term behavior feature in the set of long-term behavior features is a skip reset feature at time t,/for>Means that the i-th long-term behavior feature in the long-term behavior feature set is a jump memory feature at the t-th moment,/for>Means the long-term behavioral characteristics of the ith long-term behavioral characteristic in the long-term behavioral characteristics set at time t;

7. The method for sharing data by a mobile terminal cluster hotspot according to claim 1, wherein the calculating the secondary sharing policy corresponding to the secondary behavior feature according to the sharing splitting model includes:

8. The method for data offloading of mobile-end cluster hotspot sharing of claim 1, wherein identifying private data from the real-time offload data comprises:

9. The method for data offloading in a mobile terminal cluster hotspot sharing as claimed in claim 1, wherein privacy encrypting the real-time offload data into real-time transmission data according to the privacy data comprises:

10. A data offloading system for mobile-end cluster hotspot sharing, the system comprising:

Wherein G is _i For the ith feature gain in the feature gain set, i and N are feature group numbers, N is the total number of shared parameter feature groups in the shared parameter feature group set, the total number of shared parameter feature groups in the shared parameter feature group set is equal to the total number of feature gains in the feature gain set, gamma and psi are preset constants, and G _i,L For the sum of the first partial derivatives of the data contained in the left node of the root node when the input parameter of the shared decision model is the ith shared parameter feature group in the shared parameter feature group, H _i,L For the sum of the second partial derivatives of the data contained in the left node of the root node when the input parameters of the shared decision model are the ith shared parameter feature group in the shared parameter feature group, G _i,R For the sum of the first partial derivatives of the data contained in the right node of the root node when the input parameter is the ith shared parameter feature group in the shared parameter feature group of the shared decision model, H _i,R For the sum of the second partial derivatives of the data contained in the right node of the root node when the input parameters of the shared decision model are the ith shared parameter feature group in the shared parameter feature group, G _n,L For the sum of the first partial derivatives of the data contained in the left node of the root node when the input parameter of the shared decision model is the nth shared parameter feature group in the shared parameter feature group, H _n,L For the sum of the second partial derivatives of the data contained in the left node of the root node when the input parameters of the shared decision model are the nth shared parameter feature group in the shared parameter feature group, G _n,R The input parameters of the shared decision model are the characteristics of the shared parametersSum of first partial derivatives of data contained in right node of root node at nth shared parameter feature group in group, H _n,R Adding the second partial derivatives of the data contained in the right node of the root node when the input parameter of the sharing decision model is the nth sharing parameter feature group in the sharing parameter feature group, wherein max is a maximum value function symbol, and min is a minimum value function symbol; performing iterative node updating on the shared decision model according to the characteristic gain set to obtain a secondary decision model; calculating a prediction efficiency characteristic set corresponding to the shared parameter characteristic set by using the secondary decision model; calculating an efficiency loss value between the predicted efficiency feature set and the shared efficiency feature set, and carrying out iterative updating on model parameters of the secondary decision model according to the efficiency loss value to obtain a shared shunt model;