CN109474516A - Instant messaging connection strategy recommended method and system based on convolutional neural networks - Google Patents
Instant messaging connection strategy recommended method and system based on convolutional neural networks Download PDFInfo
- Publication number
- CN109474516A CN109474516A CN201811349684.0A CN201811349684A CN109474516A CN 109474516 A CN109474516 A CN 109474516A CN 201811349684 A CN201811349684 A CN 201811349684A CN 109474516 A CN109474516 A CN 109474516A
- Authority
- CN
- China
- Prior art keywords
- user
- network model
- neural network
- sample
- function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
- H04L51/04—Real-time or near real-time messaging, e.g. instant messaging [IM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/142—Network analysis or design using statistical or mathematical methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/147—Network analysis or design for predicting network behaviour
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
- H04L51/04—Real-time or near real-time messaging, e.g. instant messaging [IM]
- H04L51/043—Real-time or near real-time messaging, e.g. instant messaging [IM] using or handling presence information
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Algebra (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Probability & Statistics with Applications (AREA)
- Pure & Applied Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Machine Translation (AREA)
Abstract
A kind of instant messaging connection strategy recommended method and system based on convolutional neural networks, comprising: obtain the use data that user is directed to target instant communication function on wearable device within a preset period of time, include at least frequency of use using data；Data will be used to input user's disaggregated model；Wherein, which is convolutional neural networks model；Target user's type belonging to user is determined based on the output result of user's disaggregated model, and recommend the corresponding instant messaging connection strategy of target user's type to wearable device, instant messaging connection strategy is used to indicate the rate of connections that wearable device carries out instant messaging connection.Implement the embodiment of the present invention, can reduce the kwh loss of equipment.
Description
Technical field
The present invention relates to fields of communication technology, and in particular to a kind of instant messaging connection strategy based on convolutional neural networks
Recommended method and system.
Background technique
Instant messaging (Instant Messaging, abbreviation IM) is a kind of system for carrying out real time communication through network, should
Real-time communication system allows two people or more people to exchange using the instant transmitting word message of network, file, voice with video, and
The system usually provides service in a manner of website, computer software or mobile applications.
Due to the characteristic that instant messaging has real-time reception, sends information, instant communication function needs in use
Network is in and is continuously connected with state.However, in real life, unstable networks even disconnecting the case where happen occasionally,
So that instant communication function occurs not receiving, sending the failure of information in time, affect people pass through instant communication function into
The usage experience that row is linked up, exchanged.Currently, there is this problem of disconnecting, the company mainly used for instant communication function
Connecing strategy is the rule progress reconnection being incremented by according to the time, once i.e. instant messaging disconnecting, equipment will be incremented by according to the time
Rule carry out for a long time, reconnection incessantly.As it can be seen that existing instant messaging connection strategy, increases the electricity damage of equipment
Consumption.
Summary of the invention
The embodiment of the present invention discloses a kind of instant messaging connection strategy recommended method and system based on convolutional neural networks,
It can reduce the kwh loss of equipment.
First aspect of the embodiment of the present invention discloses a kind of instant messaging connection strategy recommendation based on convolutional neural networks
Method, which comprises
The use data that user is directed to target instant communication function on wearable device within a preset period of time are obtained, it is described
Frequency of use is included at least using data；
User's disaggregated model is inputted using data by described；User's disaggregated model is convolutional neural networks model；
Target user's type belonging to user is determined based on the output result of user's disaggregated model, and is worn to described
The corresponding instant messaging connection strategy of target user's type described in equipment recommendation is worn, the instant messaging connection strategy is used to indicate
The wearable device carries out the rate of connections of instant messaging connection.
As an alternative embodiment, in first aspect of the embodiment of the present invention, the target instant communication function
Including at least the first function, the second function and third function；
It is described by it is described using data input user's disaggregated model before, the method also includes:
It determines user's sample of preset quantity, and obtains each user's sample in preset number of days daily using described
First performance data, the second performance data caused by target instant communication function and third performance data；
Obtain the corresponding first default hyper parameter of first performance data, second performance data corresponding second in advance
If hyper parameter and the corresponding third of the third performance data preset hyper parameter；
It is preset according to first performance data, second performance data, the third performance data, described first super
Parameter, the second default hyper parameter and the third preset hyper parameter and determine one-dimensional vector, and the one-dimensional vector is for anti-
Reflect the service condition that each user's sample uses the target instant communication function daily in the preset number of days；
According to the one-dimensional vector and the preset number of days, determine using the numerical value of the preset number of days as the more of dimension
Dimensional vector, the multi-C vector is for reflecting that each user's sample uses the target instant messaging in the preset number of days
The service condition of function；
The training set of initial neural network model is determined according to the multi-C vector of all user's samples；
Using the training set training initial neural network model, user's disaggregated model is obtained.
As an alternative embodiment, in first aspect of the embodiment of the present invention, it is described to utilize the training training
Practice the initial neural network model, obtain user's disaggregated model, comprising:
The training set is inputted into the initial neural network model, so that the training set successively passes through the initial mind
The superposition convolution operation of convolutional layer through network model and the pondization operation of pond layer, obtain first eigenvector sample；
The full articulamentum that the first eigenvector sample is inputted to the initial neural network model, obtains second feature
Vector sample；
It is trained, is obtained according to object classifiers of the second feature vector sample to the initial neural network model
To user's disaggregated model.
As an alternative embodiment, the training set is tape label in first aspect of the embodiment of the present invention
Training set, the label are used to mark the actual classification value of each user's sample；The actual classification value is according to
What the frequency of use that user's sample actually uses the target instant communication function determined.
As an alternative embodiment, in first aspect of the embodiment of the present invention, it is described according to the second feature
Vector sample is trained the object classifiers of the initial neural network model, obtains user's disaggregated model, comprising:
The object classifiers that the second feature vector sample is inputted to the initial neural network model obtain output point
Class value, and the output category value is determined as to predict classification value；The prediction classification value and the actual classification value are carried out
Compare, obtains comparison result；
The parameter of the initial neural network model is updated according to the comparison result；
Judge whether the loss function of the initial neural network model meets preset condition；The loss function is for anti-
Reflect the error between the output category value of the initial neural network model and the actual classification value；
When the loss function of the initial neural network model meets preset condition, by the initial neural network model
Parameter current be determined as the parameter of user's disaggregated model, and user's disaggregated model is obtained according to the parameter.
Second aspect of the embodiment of the present invention discloses a kind of wearable device, comprising:
Acquiring unit is directed to target instant communication function on wearable device for obtaining user within a preset period of time
It is described to include at least frequency of use using data using data；
Input unit, for inputting user's disaggregated model using data for described；User's disaggregated model is convolution mind
Through network model；
Determination unit, for determining target user's class belonging to user based on the output result of user's disaggregated model
Type；
Recommendation unit, for recommending the corresponding instant messaging of target user's type to connect plan to the wearable device
Slightly, the instant messaging connection strategy is used to indicate the rate of connections that the wearable device carries out instant messaging connection.
As an alternative embodiment, in second aspect of the embodiment of the present invention, the target instant communication function
Including at least the first function, the second function and third function；
The determination unit is also used to before the input unit is by input user's disaggregated model using data,
Determine user's sample of preset quantity；
The acquiring unit, be also used to obtain each user's sample in preset number of days daily using the target i.e.
When communication function caused by the first performance data, the second performance data and third performance data and first function
The corresponding first default hyper parameter of data, the corresponding second default hyper parameter of second performance data and the third function
The corresponding third of data presets hyper parameter；
The determination unit is also used to according to first performance data, second performance data, the third function
Data, the first default hyper parameter, the second default hyper parameter and the third preset hyper parameter and determine one-dimensional vector,
And according to the one-dimensional vector and the preset number of days, determine multidimensional using the numerical value of the preset number of days as dimension to
Amount；Wherein, the one-dimensional vector is for reflecting that each user's sample uses the target daily in the preset number of days
The service condition of instant communication function；The multi-C vector is for reflecting that each user's sample is used in the preset number of days
The service condition of the target instant communication function；
The determination unit is also used to determine initial neural network according to the multi-C vector of all user's samples
The training set of model；
Training unit, for obtaining user's disaggregated model using the training set training initial neural network model.
As an alternative embodiment, in second aspect of the embodiment of the present invention, the training unit, comprising:
First input subelement, for the training set to be inputted the neural network model so that the training set according to
The superposition convolution operation of the secondary convolutional layer by the neural network model and the pondization operation of pond layer, obtain fisrt feature
Vector sample；
Second input subelement, for the first eigenvector sample to be inputted to the full connection of the neural network model
Layer, obtains second feature vector sample；
Training subelement, for the object classifiers according to the second feature vector sample to the neural network model
It is trained, obtains user's disaggregated model.
As an alternative embodiment, the training set is tape label in second aspect of the embodiment of the present invention
Training set, the label are used to mark the actual classification value of each user's sample；The actual classification value is according to
What the frequency of use that user's sample actually uses the target instant communication function determined.
As an alternative embodiment, the trained subelement is according to institute in second aspect of the embodiment of the present invention
It states second feature vector sample to be trained the object classifiers of the neural network model, obtains the side of user's disaggregated model
Formula specifically:
The object classifiers that the second feature vector sample is inputted to the neural network model, obtain output category
Value, and the output category value is determined as to predict classification value；The prediction classification value and the actual classification value are compared
Compared with obtaining comparison result；
The parameter of the initial neural network model is updated according to the comparison result；
Judge whether the loss function of the initial neural network model meets preset condition；The loss function is for anti-
Reflect the error between the output category value of the initial neural network model and the actual classification value；
When the loss function of the initial neural network model meets preset condition, by the initial neural network model
Parameter current be determined as the parameter of user's disaggregated model, and user's disaggregated model is obtained according to the parameter.
The third aspect of the embodiment of the present invention discloses another wearable device, and the wearable device includes:
It is stored with the memory of executable program code；
The processor coupled with the memory；
The processor calls the executable program code stored in the memory, executes the embodiment of the present invention the
On the one hand all or part of the steps in any one disclosed method.
Fourth aspect of the embodiment of the present invention discloses a kind of computer readable storage medium, which is characterized in that it, which is stored, uses
In the computer program of electronic data interchange, wherein the computer program makes computer execute the embodiment of the present invention first
All or part of the steps in any one method disclosed in aspect.
The 5th aspect of the embodiment of the present invention discloses a kind of computer program product, when the computer program product is calculating
When being run on machine, so that the computer executes some or all of any one method of first aspect step.
Compared with prior art, the embodiment of the present invention has the advantages that
In the embodiment of the present invention, obtains user and be directed to target instant communication function on wearable device within a preset period of time
Use data, using data include at least frequency of use；Data will be used to input user's disaggregated model；Wherein, user point
Class model is convolutional neural networks model；Target user's class belonging to user is determined based on the output result of user's disaggregated model
Type, and recommend the corresponding instant messaging connection strategy of target user's type to wearable device, instant messaging connection strategy is used for
Indicate that wearable device carries out the rate of connections of instant messaging connection.As it can be seen that implementing the embodiment of the present invention, user's mistake can be based on
Toward the use habit and user's disaggregated model for instant communication function, the habit of user's future usage instant messaging is carried out
It predicts, and meets the instant messaging connection strategy of user's use habit to wearable device recommendation according to prediction result, so that can
Wearable device in user there are instant messaging connection is just carried out in the case where use demand, compared in the prior art for a long time,
Reconnection incessantly, this programme can reduce the kwh loss of equipment.
Detailed description of the invention
It to describe the technical solutions in the embodiments of the present invention more clearly, below will be to needed in the embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for ability
For the those of ordinary skill of domain, without creative efforts, it can also be obtained according to these attached drawings other attached
Figure.
Fig. 1 is a kind of instant messaging connection strategy recommended method based on convolutional neural networks disclosed by the embodiments of the present invention
Flow diagram；
Fig. 2 is another instant messaging connection strategy recommendation side based on convolutional neural networks disclosed by the embodiments of the present invention
The flow diagram of method；
Fig. 3 is a kind of instant messaging connection strategy recommender system based on convolutional neural networks disclosed by the embodiments of the present invention
Structural schematic diagram；
Fig. 4 is that another instant messaging connection strategy based on convolutional neural networks disclosed by the embodiments of the present invention recommends system
The structural schematic diagram of system.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that the described embodiment is only a part of the embodiment of the present invention, instead of all the embodiments.Based on this
Embodiment in invention, every other reality obtained by those of ordinary skill in the art without making creative efforts
Example is applied, shall fall within the protection scope of the present invention.
It should be noted that term " includes " and " having " and their any changes in the embodiment of the present invention and attached drawing
Shape, it is intended that cover and non-exclusive include.Such as contain the process, method of a series of steps or units, system, product or
Equipment is not limited to listed step or unit, but optionally further comprising the step of not listing or unit or optional
Ground further includes the other step or units intrinsic for these process, methods, product or equipment.
The embodiment of the present invention disclose it is a kind of based on customization corpus voice search topic method and wearable device, can be improved
To the recognition accuracy of the phonetic problem of student, and then improves and search topic accuracy rate.It is described in detail separately below.
Embodiment one
Referring to Fig. 1, Fig. 1 is a kind of instant messaging connection plan based on convolutional neural networks disclosed by the embodiments of the present invention
The slightly flow diagram of recommended method.Wherein, it is pushed away as shown in Figure 1, being somebody's turn to do the instant messaging connection strategy based on convolutional neural networks
The method of recommending may comprise steps of:
101, recommender system obtains user and is directed to target instant communication function on wearable device within a preset period of time
Use data；Wherein, frequency of use is included at least using data.
In the embodiment of the present invention, preset time period be can be five days, seven days or 15 days；Target instant communication function can
To include micro- merely function, positioning function and video call function etc.；It the use of data can also include using time point, the present invention
Embodiment is without limitation.
102, recommender system will use data to input user's disaggregated model；Wherein, which is convolutional Neural
Network model.
In the embodiment of the present invention, recommender system will use data input user type model when, can to use data into
Line number Data preprocess, and the use data after data prediction are input to user's disaggregated model；Wherein, user's disaggregated model
For preparatory trained convolutional neural networks model.
103, recommender system determines target user's type belonging to user based on the output result of user's disaggregated model.
In the embodiment of the present invention, recommender system can be according to different user for the service condition of target instant communication function
User is determined for the real-time demand of target instant communication function, and according to user for the reality of target instant communication function
The desirability of when property divides user type.Specifically, user type can be according to user for target instant communication function
The desirability of real-time user is divided into a variety of user types from high to low, wherein the higher expression use of desirability
Family is higher using the frequency of target instant communication function, i.e., user wants the real-time of the connection of target instant communication function
Ask higher.For example, 8 kinds of user types can be divided, wherein may include that high-frequency uses type, middle frequency usage type and low frequency
Rate is using type etc., and the embodiment of the present invention is without limitation.As it can be seen that the embodiment of the present invention, it can be based on user for target Instant Messenger
The desirability of the connection real-time of telecommunication function divides user type, defers to the routine use habit of user, both realized for
The classifying rationally of different user group, and improve the usage experience of user.
104, recommender system recommends the corresponding instant messaging connection strategy of target user's type to wearable device, this is immediately
Communication connection strategy is used to indicate the rate of connections that wearable device carries out instant messaging connection.
In the embodiment of the present invention, instant messaging connection strategy can serve to indicate that wearable device carries out instant messaging connection
Connection Time point and rate of connections.
As it can be seen that by method described in Fig. 1, can based on the passing use habit for instant communication function of user with
And user's disaggregated model, the habit of user's future usage instant messaging is predicted, and is set according to prediction result to wearable
The standby instant messaging connection strategy for recommending to meet user's use habit, so that wearable device is in user, there are the feelings of use demand
Instant messaging connection is just carried out under condition, having compared long-time, reconnection incessantly, this programme in the prior art can reduce equipment
Kwh loss；It is used furthermore it is possible to be divided based on desirability of the user for the connection real-time of target instant communication function
Family type defers to the routine use habit of user, had not only realized the classifying rationally for different user group, but also improve user
Usage experience.
Embodiment two
Referring to Fig. 2, Fig. 2 is another instant messaging connection based on convolutional neural networks disclosed by the embodiments of the present invention
The flow diagram of policy recommendation method.It wherein, as shown in Fig. 2, should the instant messaging connection strategy based on convolutional neural networks
Recommended method may comprise steps of:
201, recommender system determines user's sample of preset quantity, and it is daily in preset number of days to obtain each user's sample
Use the first performance data, the second performance data caused by target instant communication function and third performance data.
In the embodiment of the present invention, optionally, it is contemplated that regional otherness determines preset quantity in more just mode
User's sample, it can determine target provinces, cities and autonomous regions, and according to preset ratio in each target provinces, cities and autonomous regions with
Machine chooses user's sample, and user's sample of preset quantity is determined by user's sample of each target provinces, cities and autonomous regions.For example,
Recommender system can determine 20 target provinces, cities and autonomous regions, and random in each target provinces, cities and autonomous regions according to identical ratio
250 user's samples are chosen, determine 5000 user's samples altogether by 250 user's samples of each provinces, cities and autonomous regions.As it can be seen that
The embodiment of the present invention can avoid the sampling error as caused by regional disparity, and then effectively by the way of mostly sampling
It ensure that the accuracy and reliability that obtained training result is trained using sampled data.
In embodiments of the present invention, target instant communication function at least may include the first function, the second function and third
Function；Correspondingly, it is the first function that each user's sample uses data caused by the first function daily in preset number of days
Data, it is the second performance data that each user's sample uses data caused by the second function daily in preset number of days, each
It is third performance data that user's sample uses data caused by third function daily in preset number of days.
202, recommender system obtains the corresponding first default hyper parameter of the first performance data, the second performance data corresponding the
Two default hyper parameters and the corresponding third of third performance data preset hyper parameter.
203, recommender system is according to the first performance data, the second performance data, third performance data, the first default super ginseng
Number, the second default hyper parameter and third preset hyper parameter and determine one-dimensional vector；Wherein, the one-dimensional vector is for reflecting each use
Family sample uses the service condition of target instant communication function daily in preset number of days.
204, recommender system is determined according to above-mentioned one-dimensional vector and preset number of days using the numerical value of preset number of days as dimension
Multi-C vector；Wherein, the multi-C vector is for reflecting that each user's sample uses target instant communication function in preset number of days
Service condition.
205, recommender system determines the training set of initial neural network model according to the multi-C vector of all user's samples.
For step 201~205, for example, the available 5000 user's samples of recommender system, and obtain each use
Family sample is in 30 days daily using video call function data V caused by video call functiont, produced using positioning function
Raw positioning function data LtAnd micro- merely performance data W caused by function is chatted using micro-t；Due to each instant communication function
The requirement of rate real-time for instant messaging is different, therefore can introduce three hyper parameters a, β and γ, considers further that the dilute of the frequency
Property is dredged, the function for reflecting the daily network liveness of each user's sample is finally obtained, which is denoted as F, it may be assumed that
F=[α log (1+Vt)+βlog(1+Lt)+γlog(1+Wt)]
As it can be seen that above-mentioned function F is to indicate that each user's sample uses the use of instant communication function daily in 30 days
The one-dimensional vector of situation；Further, it is known that preset number of days is 30 days, therefore, available using 30 days multidimensional as dimension
Vector FvFor (F1, F2, F3..., F29, F30)；Wherein, the multi-C vector is for reflecting that each user's sample uses mesh in 30 days
Mark the service condition of instant communication function；It further, can be according to the multi-C vector F of all user's samplesvIt determines initial
The training set and test set of neural network model, the matrix of respectively 4000 1*30 and the matrix of 000 1*30；Wherein,
The partial data form of training set are as follows:
206, recommender system obtains user's disaggregated model using the initial neural network model of training set training.
As an alternative embodiment, recommender system is used using the initial neural network model of training set training
Family disaggregated model may include:
Training set is inputted into initial neural network model, so that training set successively passes through the convolution of initial neural network model
The superposition convolution operation of layer and the pondization operation of pond layer, obtain first eigenvector sample；
The full articulamentum that first eigenvector sample is inputted to initial neural network model obtains second feature vector sample
This；
It is trained according to object classifiers of the second feature vector sample to initial neural network model, obtains user point
Class model.
In the embodiment of the present invention, initial neural network model can be by input layer, hidden layer, full articulamentum and output layer
Structure composition, each layer can be made of single or multiple neurons, wherein the neuron of the initial neural network model is such as
Under:
f(∑ixiwi+b)；
Recommender system can be input to neuron for training set as the input data of the initial neural network model, by mind
The input data of the neuron is mapped to output end by the activation primitive through running in member, obtains the output data of the neuron,
And using the output data as the input data of next layer of neuron；Wherein, in order to preferably be fitted actual function, Ke Yixuan
Nonlinear activation function, such as ReLu function are selected, i.e. activation primitive f is indicated are as follows:
Wherein, the hidden layer of initial neural network model can be convolutional layer and pond layer；It will enter into initial nerve net
The input data of the input layer of network model successively passes through the superposition convolution operation of convolutional layer and the pondization operation of pond layer, with reality
Now to the Fusion Features of input data, first eigenvector sample is obtained；Further first eigenvector sample is inputted initial
The full articulamentum of neural network model obtains global characteristics vector sample, i.e. second feature vector sample；Finally, according to second
Feature vector sample is trained object classifiers, obtains user's disaggregated model.Wherein, object classifiers can be
Softmax function, is defined as follows:
Still optionally further, above-mentioned training set is the training set of tape label, and the label is for marking each user's sample
Actual classification value；And actual classification value is determined according to the frequency of use of user's sample actual use target instant communication function
's.Specifically, acquisition user's sample and the corresponding target instant communication function of each user's sample use data it
It afterwards, can be according to using the corresponding actual classification value of the artificially defined each user's sample of data, and to user's sample and the use
Family sample is corresponding to stamp the label comprising the actual classification value using data；Wherein, each classification value can correspond to difference
User type.For example, user type may include that high-frequency uses type, middle frequency usage type and low frequency to use type,
In, it is 9 that high-frequency, which can correspond to classification value using type,；It is 5 that middle frequency usage type, which can correspond to classification value,；Low frequency can using type
To correspond to classification value as 2.
Further optional, according to above-described embodiment content, recommender system is according to second feature vector sample to nerve
The object classifiers of network model are trained, and are obtained user's disaggregated model and be may include:
By the object classifiers of second feature vector sample input neural network model, output category value is obtained, and will be defeated
Classification value is determined as predicting classification value out；Prediction classification value is compared with actual classification value, obtains comparison result；
The parameter of initial neural network model is updated according to comparison result；
Judge whether the loss function of initial neural network model meets preset condition；The loss function is initial for reflecting
Error between the output category value and actual classification value of neural network model；
When the loss function of initial neural network model meets preset condition, by the current ginseng of initial neural network model
Number is determined as the parameter of user's disaggregated model, and obtains user's disaggregated model according to parameter.
In the embodiment of the present invention, second feature vector sample can be passed through into softmax function (object classifiers), to obtain
Output category probability distribution, then compared with model answer (actual classification value corresponding probability distribution), find out intersection
Entropy obtains loss function；Wherein, loss function can be used to indicate that the output category value and reality point of initial neural network model
Error condition between class value, i.e. hypothesis output category value (prediction classification value) are y, and actual classification value is y, then loss function
Loss can carry out error calculation using cross entropy, it may be assumed that
Loss (y_, y)=- Σ y_*logy
It should be noted that during updating the parameter of initial neural network model according to comparison result, Ke Yitong
The speed of overfitting rate adjustment undated parameter.Specifically, learning rate (learning_rate) illustrates what every subparameter updated
Amplitude size；Learning rate is excessive, will lead to parameter to be optimized and fluctuates near minimum value, does not restrain；Learning rate is too small, can lead
Parameter to be optimized is caused to restrain slow.In the training process of parameter for updating initial neural network model according to comparison result,
The direction that the update of parameter declines towards loss function gradient.Due to consideration that learning rate is conducive to greatly quickly repeatedly at first
In generation, finds optimal solution, and the later period slowly adjusts ginseng, therefore exponential damping learning rate can be used, i.e., learning rate is with the variation of exercise wheel number
And dynamic updates, learning rate calculation formula is as follows:
Wherein, LearningRateBase is learning rate initial value, and LearningRateDecay is learning rate attenuation rate,
GlobalStep has recorded current exercise wheel number, for that can not train shape parameter, the decaying of learning rate step type.It changes eventually by more wheels
In generation, reduces penalty values by back-propagation algorithm (BP), finally obtains training pattern function, i.e. user type model.
Wherein, being somebody's turn to do the instant messaging connection strategy recommended method based on convolutional neural networks further includes step 207~210,
For the description of step 207~210, the detailed description that step 101~104 are directed in embodiment one is please referred to, the present invention is implemented
Example repeats no more.
As it can be seen that by method described in Fig. 2, can based on the passing use habit for instant communication function of user with
And user's disaggregated model, the habit of user's future usage instant messaging is predicted, and is set according to prediction result to wearable
The standby instant messaging connection strategy for recommending to meet user's use habit, so that wearable device is in user, there are the feelings of use demand
Instant messaging connection is just carried out under condition, having compared long-time, reconnection incessantly, this programme in the prior art can reduce equipment
Kwh loss；And it can be divided and be used based on desirability of the user for the connection real-time of target instant communication function
Family type defers to the routine use habit of user, had not only realized the classifying rationally for different user group, but also improve user
Usage experience；In addition, avoiding the sampling error as caused by regional disparity, and then effectively by the way of mostly sampling
It ensure that the accuracy and reliability that obtained training result is trained using sampled data.
Embodiment three
Referring to Fig. 3, Fig. 3 is a kind of instant messaging connection plan based on convolutional neural networks disclosed by the embodiments of the present invention
The slightly structural schematic diagram of recommender system.As shown in figure 3, the recommender system may include:
Acquiring unit 301 is directed to target instant messaging function on wearable device for obtaining user within a preset period of time
The use data of energy, and this is supplied to input unit 302 using data.
In the embodiment of the present invention, preset time period be can be five days, seven days or 15 days；Target instant communication function can
To include micro- merely function, positioning function and video call function etc.；When may include frequency of use and use using data
Between point, the embodiment of the present invention is without limitation.
Input unit 302 for that data will be used to input user's disaggregated model, and triggers the starting of determination unit 303；Its
In, which is convolutional neural networks model.
It, can be to using data when input unit 302 will use data input user type model in the embodiment of the present invention
Data prediction is carried out, and the use data after data prediction are input to user's disaggregated model；Wherein, user classification mould
Type is preparatory trained convolutional neural networks model.
Determination unit 303 determines target user's type belonging to user for the output result based on user's disaggregated model,
And target user's type is supplied to recommendation unit 304.
Recommendation unit 304 should for recommending the corresponding instant messaging connection strategy of target user's type to wearable device
Instant messaging connection strategy is used to indicate the rate of connections that wearable device carries out instant messaging connection.
In the embodiment of the present invention, instant messaging connection strategy can serve to indicate that wearable device carries out instant messaging connection
Connection Time point and rate of connections.
As it can be seen that can be practised based on the passing use for instant communication function of user by recommender system described in Fig. 3
Used and user's disaggregated model, predicts the habit of user's future usage instant messaging, and according to prediction result to can wear
The instant messaging connection strategy that equipment recommendation meets user's use habit is worn, there are use demands so that wearable device is in user
In the case where just carry out instant messaging connection, compared in the prior art for a long time, reconnection incessantly, this programme can reduce
The kwh loss of equipment；Furthermore it is possible to be drawn based on desirability of the user for the connection real-time of target instant communication function
Divide user type, defers to the routine use habit of user, not only realized the classifying rationally for different user group, but also improve
The usage experience of user.
Example IV
Referring to Fig. 4, Fig. 4 is another instant messaging connection based on convolutional neural networks provided in an embodiment of the present invention
The structural schematic diagram of policy recommendation system, wherein recommender system shown in Fig. 4 be recommender system as shown in Figure 4 further into
Row optimization obtains.Compared with recommender system shown in Fig. 3, in recommender system shown in Fig. 4:
Above-mentioned determination unit 303, be also used to above-mentioned input unit 302 will use data input user's disaggregated model it
Before, determine user's sample of preset quantity, and be supplied to acquiring unit 301.
In the embodiment of the present invention, target instant communication function at least may include the first function, the second function and third function
Energy.
In the embodiment of the present invention, optionally, it is contemplated that regional otherness determines preset quantity in more just mode
User's sample, it can determine target provinces, cities and autonomous regions, and according to preset ratio in each target provinces, cities and autonomous regions with
Machine chooses user's sample, and user's sample of preset quantity is determined by user's sample of each target provinces, cities and autonomous regions.For example,
Determination unit 303 can determine 20 target provinces, cities and autonomous regions, and according to identical ratio in each target provinces, cities and autonomous regions
250 user's samples are randomly selected, determine 5000 user's samples altogether by 250 user's samples of each provinces, cities and autonomous regions.
As it can be seen that the embodiment of the present invention, can avoid the sampling error as caused by regional disparity by the way of mostly sampling, into
And the accuracy and reliability that obtained training result is trained using sampled data is effectively ensured.
In embodiments of the present invention, target instant communication function at least may include the first function, the second function and third
Function；Correspondingly, it is the first function that each user's sample uses data caused by the first function daily in preset number of days
Data, it is the second performance data that each user's sample uses data caused by the second function daily in preset number of days, each
It is third performance data that user's sample uses data caused by third function daily in preset number of days.
Above-mentioned acquiring unit 301 is also used to obtain each user's sample and uses target Instant Messenger daily in preset number of days
First performance data, the second performance data caused by telecommunication function and third performance data and the first performance data are corresponding
The first default hyper parameter, the corresponding second default hyper parameter of the second performance data and the corresponding third of third performance data it is pre-
If hyper parameter.
Above-mentioned determination unit 303 is also used to according to the first performance data, the second performance data, third performance data, first
Default hyper parameter, the second default hyper parameter and third preset hyper parameter and determine one-dimensional vector, and according to one-dimensional vector and
Preset number of days determines the multi-C vector using the numerical value of preset number of days as dimension.
Wherein, one-dimensional vector is for reflecting that each user's sample uses target instant communication function daily in preset number of days
Service condition；Multi-C vector is used to reflect the use feelings that each user's sample uses target instant communication function in preset number of days
Condition.
Above-mentioned determination unit 303 is also used to determine initial neural network model according to the multi-C vector of all user's samples
Training set, and the training set is supplied to training unit 305.
Training unit 305, for obtaining user's disaggregated model using the initial neural network model of training set training.
In the embodiment of the present invention, user's disaggregated model can be supplied to input unit 302 by training unit 305.
As an alternative embodiment, as shown in figure 4, above-mentioned training unit 305 may include:
First input subelement 3051, for training set to be inputted neural network model, so that training set successively passes through mind
The superposition convolution operation of convolutional layer through network model and the pondization operation of pond layer, obtain first eigenvector sample, and
The first eigenvector sample is supplied to the second input subelement 3052.
Second input subelement 3052, for first eigenvector sample to be inputted to the full articulamentum of neural network model,
Second feature vector sample is obtained, and the second feature vector sample is supplied to trained subelement 3053.
Training subelement 3053, for being carried out according to object classifiers of the second feature vector sample to neural network model
Training, obtains user's disaggregated model.
As another optional embodiment, as shown in figure 4, training subelement 3053 is according to second feature vector sample
The object classifiers of neural network model are trained, the mode for obtaining user's disaggregated model is specifically as follows:
By the object classifiers of second feature vector sample input neural network model, output category value is obtained, and will be defeated
Classification value is determined as predicting classification value out；Prediction classification value is compared with actual classification value, obtains comparison result；
The parameter of initial neural network model is updated according to comparison result；
Judge whether the loss function of initial neural network model meets preset condition；Loss function is for reflecting initial mind
Error between output category value through network model and actual classification value；
When the loss function of initial neural network model meets preset condition, by the current ginseng of initial neural network model
Number is determined as the parameter of user's disaggregated model, and obtains user's disaggregated model according to parameter.
In the embodiment of the present invention, training set is the training set of tape label, which is used to mark the reality of each user's sample
Border classification value；The actual classification value is determined according to the frequency of use of user's sample actual use target instant communication function.
As it can be seen that can be practised based on the passing use for instant communication function of user by recommender system described in Fig. 4
Used and user's disaggregated model, predicts the habit of user's future usage instant messaging, and according to prediction result to can wear
The instant messaging connection strategy that equipment recommendation meets user's use habit is worn, there are use demands so that wearable device is in user
In the case where just carry out instant messaging connection, compared in the prior art for a long time, reconnection incessantly, this programme can reduce
The kwh loss of equipment；And it can be drawn based on desirability of the user for the connection real-time of target instant communication function
Divide user type, defers to the routine use habit of user, not only realized the classifying rationally for different user group, but also improve
The usage experience of user；In addition, avoiding the sampling error as caused by regional disparity, in turn by the way of mostly sampling
The accuracy and reliability that obtained training result is trained using sampled data has been effectively ensured.
Those of ordinary skill in the art will appreciate that all or part of the steps in the various methods of above-described embodiment is can
It is completed with instructing relevant hardware by program, which can be stored in a computer readable storage medium, storage
Medium include read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory,
RAM), programmable read only memory (Programmable Read-only Memory, PROM), erasable programmable is read-only deposits
Reservoir (Erasable Programmable Read Only Memory, EPROM), disposable programmable read-only memory (One-
Time Programmable Read-Only Memory, OTPROM), the electronics formula of erasing can make carbon copies read-only memory
(Electrically-Erasable Programmable Read-Only Memory, EEPROM), CD-ROM (Compact
Disc Read-Only Memory, CD-ROM) or other disc memories, magnetic disk storage, magnetic tape storage or can
For carrying or any other computer-readable medium of storing data.
Above to a kind of instant messaging connection strategy recommendation side based on convolutional neural networks disclosed by the embodiments of the present invention
Method and system are described in detail, and specific case used herein explains the principle of the present invention and embodiment
It states, the above description of the embodiment is only used to help understand the method for the present invention and its core ideas；Meanwhile for this field
Those skilled in the art, according to the thought of the present invention, there will be changes in the specific implementation manner and application range, to sum up institute
It states, the contents of this specification are not to be construed as limiting the invention.
Claims (10)
1. a kind of instant messaging connection strategy recommended method based on convolutional neural networks, which is characterized in that the described method includes:
Obtain the use data that user is directed to target instant communication function on wearable device within a preset period of time, the use
Data include at least frequency of use；
User's disaggregated model is inputted using data by described；User's disaggregated model is convolutional neural networks model；
Target user's type belonging to user is determined based on the output result of user's disaggregated model, and wearable is set to described
Standby to recommend the corresponding instant messaging connection strategy of target user's type, the instant messaging connection strategy is used to indicate described
The rate of connections of wearable device progress instant messaging connection.
2. the method according to claim 1, wherein the target instant communication function includes at least the first function
Energy, the second function and third function；
It is described by it is described using data input user's disaggregated model before, the method also includes:
It determines user's sample of preset quantity, and obtains each user's sample and use the target daily in preset number of days
First performance data, the second performance data caused by instant communication function and third performance data；
It obtains the corresponding first default hyper parameter of first performance data, second performance data corresponding second and presets and surpass
Parameter and the corresponding third of the third performance data preset hyper parameter；
According to first performance data, second performance data, the third performance data, the first default super ginseng
Several, the second default hyper parameter and the third preset hyper parameter and determine one-dimensional vector, and the one-dimensional vector is for reflecting
Each user's sample uses the service condition of the target instant communication function daily in the preset number of days；
According to the one-dimensional vector and the preset number of days, determine multidimensional using the numerical value of the preset number of days as dimension to
Amount, the multi-C vector is for reflecting that each user's sample uses the target instant communication function in the preset number of days
Service condition；
The training set of initial neural network model is determined according to the multi-C vector of all user's samples；
Using the training set training initial neural network model, user's disaggregated model is obtained.
3. according to the method described in claim 2, it is characterized in that, described utilize the training set training initial mind
Through network model, user's disaggregated model is obtained, comprising:
The training set is inputted into the initial neural network model, so that the training set successively passes through the initial nerve net
The superposition convolution operation of the convolutional layer of network model and the pondization operation of pond layer, obtain first eigenvector sample；
The full articulamentum that the first eigenvector sample is inputted to the initial neural network model, obtains second feature vector
Sample；
It is trained, is used according to object classifiers of the second feature vector sample to the initial neural network model
Family disaggregated model.
4. according to the method described in claim 3, it is characterized in that, the training set is the training set of tape label, the label
For marking the actual classification value of each user's sample；The actual classification value is actually used according to user's sample
What the frequency of use of the target instant communication function determined.
5. according to the method described in claim 4, it is characterized in that, it is described according to the second feature vector sample to it is described just
The object classifiers of beginning neural network model are trained, and obtain user's disaggregated model, comprising:
The object classifiers that the second feature vector sample is inputted to the initial neural network model, obtain output category
Value, and the output category value is determined as to predict classification value；The prediction classification value and the actual classification value are compared
Compared with obtaining comparison result；
The parameter of the initial neural network model is updated according to the comparison result；
Judge whether the loss function of the initial neural network model meets preset condition；The loss function is for reflecting institute
State the error between the output category value of initial neural network model and the actual classification value；
When the loss function of the initial neural network model meets preset condition, by working as the initial neural network model
Preceding parameter is determined as the parameter of user's disaggregated model, and obtains user's disaggregated model according to the parameter.
6. a kind of instant messaging connection strategy recommender system based on convolutional neural networks characterized by comprising
Acquiring unit, the use for being directed to target instant communication function on wearable device within a preset period of time for obtaining user
Data, it is described to include at least frequency of use using data；
Input unit, for inputting user's disaggregated model using data for described；User's disaggregated model is convolutional Neural net
Network model；
Determination unit, for determining target user's type belonging to user based on the output result of user's disaggregated model；
Recommendation unit, for recommending the corresponding instant messaging connection strategy of target user's type to the wearable device,
The instant messaging connection strategy is used to indicate the rate of connections that the wearable device carries out instant messaging connection.
7. recommender system according to claim 6, which is characterized in that the target instant communication function includes at least first
Function, the second function and third function；
The determination unit is also used to before the input unit is by input user's disaggregated model using data, determines
User's sample of preset quantity；
The acquiring unit is also used to obtain each user's sample and uses the target Instant Messenger daily in preset number of days
First performance data, the second performance data caused by telecommunication function and third performance data and first performance data
The corresponding second default hyper parameter of corresponding first default hyper parameter, second performance data and the third performance data
Corresponding third presets hyper parameter；
The determination unit is also used to according to first performance data, second performance data, the third function number
According to, the first default hyper parameter, the second default hyper parameter and the third preset hyper parameter and determine one-dimensional vector, with
And according to the one-dimensional vector and the preset number of days, determine multidimensional using the numerical value of the preset number of days as dimension to
Amount；Wherein, the one-dimensional vector is for reflecting that each user's sample uses the target daily in the preset number of days
The service condition of instant communication function；The multi-C vector is for reflecting that each user's sample is used in the preset number of days
The service condition of the target instant communication function；
The determination unit is also used to determine initial neural network model according to the multi-C vector of all user's samples
Training set；
Training unit, for obtaining user's disaggregated model using the training set training initial neural network model.
8. recommender system according to claim 7, which is characterized in that the training unit, comprising:
First input subelement, for the training set to be inputted the neural network model, so that the training set successively passes through
The superposition convolution operation of the convolutional layer of the neural network model and the pondization operation of pond layer are crossed, first eigenvector is obtained
Sample；
Second input subelement, for the first eigenvector sample to be inputted to the full articulamentum of the neural network model,
Obtain second feature vector sample；
Training subelement, for being carried out according to object classifiers of the second feature vector sample to the neural network model
Training, obtains user's disaggregated model.
9. recommender system according to claim 8, which is characterized in that the training set is the training set of tape label, described
Label is used to mark the actual classification value of each user's sample；The actual classification value is practical according to user's sample
It is determined using the frequency of use of the target instant communication function.
10. recommender system according to claim 9, which is characterized in that the trained subelement is according to the second feature
Vector sample is trained the object classifiers of the neural network model, obtains the mode of user's disaggregated model specifically:
The object classifiers that the second feature vector sample is inputted to the neural network model obtain output category value, and
The output category value is determined as to predict classification value；The prediction classification value is compared with the actual classification value, is obtained
To comparison result；
The parameter of the initial neural network model is updated according to the comparison result；
Judge whether the loss function of the initial neural network model meets preset condition；The loss function is for reflecting institute
State the error between the output category value of initial neural network model and the actual classification value；
When the loss function of the initial neural network model meets preset condition, by working as the initial neural network model
Preceding parameter is determined as the parameter of user's disaggregated model, and obtains user's disaggregated model according to the parameter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811349684.0A CN109474516B (en) | 2018-11-13 | 2018-11-13 | Method and system for recommending instant messaging connection strategy based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811349684.0A CN109474516B (en) | 2018-11-13 | 2018-11-13 | Method and system for recommending instant messaging connection strategy based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109474516A true CN109474516A (en) | 2019-03-15 |
| CN109474516B CN109474516B (en) | 2021-06-01 |
Family
ID=65672548
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811349684.0A Active CN109474516B (en) | 2018-11-13 | 2018-11-13 | Method and system for recommending instant messaging connection strategy based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109474516B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112580713A (en) * | 2020-12-15 | 2021-03-30 | 广州芯德通信科技股份有限公司 | Routing equipment control method and system based on pattern recognition |
| CN113469206A (en) * | 2020-03-31 | 2021-10-01 | 华为技术有限公司 | Method, device, equipment and storage medium for obtaining artificial intelligence model |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1852355A (en) * | 2005-10-17 | 2006-10-25 | 华为技术有限公司 | Method and device for collecting user communication characteristic information |
| CN1968527A (en) * | 2006-09-27 | 2007-05-23 | 蒋晓海 | Method for intelligent connection between mobile phone client and server |
| US8514717B2 (en) * | 2011-01-21 | 2013-08-20 | T-Mobile Usa, Inc. | Smart connection manager |
| CN105681157A (en) * | 2014-11-20 | 2016-06-15 | 腾讯科技（深圳）有限公司 | Long connection performance prompting method and apparatus based on instant messaging |
| CN107368718A (en) * | 2017-07-06 | 2017-11-21 | 同济大学 | A kind of user browsing behavior authentication method and system |
| CN108513344A (en) * | 2018-03-16 | 2018-09-07 | 广东小天才科技有限公司 | Suspension reconnection method, device, wearable device and the medium of wearable device |
-
2018
- 2018-11-13 CN CN201811349684.0A patent/CN109474516B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1852355A (en) * | 2005-10-17 | 2006-10-25 | 华为技术有限公司 | Method and device for collecting user communication characteristic information |
| CN1968527A (en) * | 2006-09-27 | 2007-05-23 | 蒋晓海 | Method for intelligent connection between mobile phone client and server |
| US8514717B2 (en) * | 2011-01-21 | 2013-08-20 | T-Mobile Usa, Inc. | Smart connection manager |
| US9107146B2 (en) * | 2011-01-21 | 2015-08-11 | T-Mobile Usa, Inc. | Smart connection manager |
| CN105681157A (en) * | 2014-11-20 | 2016-06-15 | 腾讯科技（深圳）有限公司 | Long connection performance prompting method and apparatus based on instant messaging |
| CN107368718A (en) * | 2017-07-06 | 2017-11-21 | 同济大学 | A kind of user browsing behavior authentication method and system |
| CN108513344A (en) * | 2018-03-16 | 2018-09-07 | 广东小天才科技有限公司 | Suspension reconnection method, device, wearable device and the medium of wearable device |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113469206A (en) * | 2020-03-31 | 2021-10-01 | 华为技术有限公司 | Method, device, equipment and storage medium for obtaining artificial intelligence model |
| CN112580713A (en) * | 2020-12-15 | 2021-03-30 | 广州芯德通信科技股份有限公司 | Routing equipment control method and system based on pattern recognition |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109474516B (en) | 2021-06-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107423442A (en) | Method and system, storage medium and computer equipment are recommended in application based on user's portrait behavioural analysis | |
| CN111353392B (en) | Face change detection method, device, equipment and storage medium | |
| CN108665366A (en) | Determine method, terminal device and the computer readable storage medium of consumer's risk grade | |
| CN108665064A (en) | Neural network model training, object recommendation method and device | |
| CN108510280A (en) | A kind of financial fraud behavior prediction method based on mobile device behavioral data | |
| CN111798244B (en) | Transaction fraud monitoring method and device | |
| CN106097043A (en) | The processing method of a kind of credit data and server | |
| CN114896899B (en) | Multi-agent distributed decision method and system based on information interaction | |
| CN110349012A (en) | Data predication method and computer readable storage medium | |
| CN110472798A (en) | Prediction technique, device and the computer readable storage medium of time series data | |
| CN110225368A (en) | A kind of video locating method, device and electronic equipment | |
| CN109474516A (en) | Instant messaging connection strategy recommended method and system based on convolutional neural networks | |
| CN109670927A (en) | The method of adjustment and its device of credit line, equipment, storage medium | |
| CN116992151A (en) | Online course recommendation method based on double-tower graph convolution neural network | |
| CN110443382A (en) | A kind of user interaction approach based on intelligent customer service platform | |
| CN113381888B (en) | Control method, device, equipment and storage medium | |
| CN113191880A (en) | Bank teller terminal cash adding suggestion determination method and device | |
| CN111368060B (en) | Self-learning method, device and system for conversation robot, electronic equipment and medium | |
| CN116452322A (en) | Credit card recommendation method and device | |
| CN110399617A (en) | Audit data processing method, system and readable storage medium storing program for executing | |
| CN109741172A (en) | Credit method for early warning, device, system and storage medium | |
| CN113656560B (en) | Emotion category prediction method and device, storage medium and electronic equipment | |
| CN112926952A (en) | Cloud computing-combined big data office business processing method and big data server | |
| CN114996435A (en) | Information recommendation method, device, equipment and storage medium based on artificial intelligence | |
| CN112184056B (en) | Data feature extraction method and system based on convolutional neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |