CN110472247A - A kind of method of multi-semantic meaning information converged network prediction model response time - Google Patents
A kind of method of multi-semantic meaning information converged network prediction model response time Download PDFInfo
- Publication number
- CN110472247A CN110472247A CN201910770984.4A CN201910770984A CN110472247A CN 110472247 A CN110472247 A CN 110472247A CN 201910770984 A CN201910770984 A CN 201910770984A CN 110472247 A CN110472247 A CN 110472247A
- Authority
- CN
- China
- Prior art keywords
- model
- response time
- neural network
- data characteristics
- vector
- 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.)
- Pending
Links
- 230000004044 response Effects 0.000 title claims abstract description 24
- 238000000034 method Methods 0.000 title claims abstract description 18
- 238000003062 neural network model Methods 0.000 claims abstract description 12
- 239000000284 extract Substances 0.000 claims abstract description 3
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 6
- 210000005036 nerve Anatomy 0.000 claims 1
- 210000002569 neuron Anatomy 0.000 description 9
- 230000006870 function Effects 0.000 description 6
- 238000012545 processing Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- 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
- G06N3/048—Activation functions
-
- 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
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Business, Economics & Management (AREA)
- Neurology (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Economics (AREA)
- Machine Translation (AREA)
Abstract
The invention discloses the methods of multi-semantic meaning information converged network prediction model response time a kind of, include the following steps: to choose N model and extracts the data characteristics of each model, wherein data characteristics the problem of including model description, model title, the label of model, the creation time of model and the week characteristic information of model utilize the response time of the data characteristics prediction model of model；The data characteristics of model is converted into vector using doc2vec model and vector is input to the response time in full Connection Neural Network model using sigmoid function prediction model.This method is during prediction, it is contemplated that and the characteristic informations such as description information, heading message and label of model are more accurate compared to the method prediction for only considering single features.
Description
Technical field
The present invention relates to information data processing technology fields more particularly to a kind of multi-semantic meaning information converged network to predict model
The method of response time.
Background technique
Since the website Stack Overflow starts, the software developer of large quantities of professions is just attracted, has programmed
Fan and computer science personnel's note that they share the problem of they encounter in programming and challenge by the website,
Model by answering others enriches the experience of oneself.Stack Overflow is one and provides the flat of exchange programming thematic knowledge
The question and answer website of platform stores many soft project knowledge, the software systems and third for helping developer's processing increasingly complicated
Square component.But the website Stack Overflow has one disadvantage in that, for user propose the problem of acceptable answer, website
It is upper to be prompted without explicitly estimated response time, certain trouble and problem can be brought to user in this way, if user is eager to know
Road answer has waited for a long time again without answer is obtained, this not only wastes their time, also will affect their work effect
Rate.If user in the problem of proposition, has a specific response time point on Stack Overflow, which can root
The answer for choosing whether to wait the problem according to its response time, is still eager to seek other solutions, so neither
The work that will affect user also increases the convenience of website.Therefore, predict website on model from be created to receive first
Answering the time needed becomes extremely important.
Summary of the invention
According to problem of the existing technology, the invention discloses a kind of multi-semantic meaning information converged network prediction models to answer
The method of time, specifically comprises the following steps:
The problem of N model of selection and data characteristics for extracting each model, wherein data characteristics includes model description, note
Son title, the label of model, the creation time of model and model week characteristic information, utilize model data characteristics prediction
The response time of model；
The data characteristics of model is converted into vector using doc2vec model and vector is input to full connection nerve net
The response time of sigmoid function prediction model is utilized in network model.
Further, vector is input to returning using sigmoid function prediction model in full Connection Neural Network model
When answering the time specifically in the following way:
Full Connection Neural Network model is constructed, including an input layer, two hidden layers and an output layer；
The feature vector of model is input in full Connection Neural Network model, which learn and training updates
Weight and threshold value, the deconditioning when error reaches sets requirement between neuron；
Model parameter is optimized using optimizer during training, utilizes trained full Connection Neural Network
The response time of model prediction model.
By adopting the above-described technical solution, a kind of multi-semantic meaning information converged network prediction model provided by the invention is answered
The method of time, this method are converted to the text feature of model using doc2vec model by the text feature of extraction model
Input information of the vector as full Connection Neural Network, by the multifactor feature of model combine full Connection Neural Network model into
The prediction of row model response time, during prediction, it is contemplated that the spies such as description information, heading message and label of model
Reference breath is more accurate compared to the method prediction for only considering single features.In training neural network, use
AdamOptimizer optimizes model parameter, model can be made to reach better convergence, while using Dropout method
It avoids mind that over-fitting is occurred in network, further improves prediction effect.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The some embodiments recorded in application, for those of ordinary skill in the art, without creative efforts,
It is also possible to obtain other drawings based on these drawings.
Fig. 1 is the flow chart of the method for the present invention.
Specific embodiment
To keep technical solution of the present invention and advantage clearer, with reference to the attached drawing in the embodiment of the present invention, to this
Technical solution in inventive embodiments carries out clear and complete description:
The method of multi-semantic meaning information converged network prediction model response time as shown in Figure 1 a kind of, specifically includes as follows
Step:
S1: 100,000 model data are handled before choosing Stack Overflow 2013, successively reject no answerer
Model data, reject the data that model label is html and other secondary text labels, the response time span for rejecting model is big
In the data of 4e5；
S2: extracting the data characteristics of model from the data after processing, describes (body) the problem of model, the mark of model
It inscribes (title), the label (tags) of model, the creation time (time-rate) of model, the week feature (weekday) of model
Deng, utilize model data characteristics prediction model response time；
Model data characteristics is extracted, mainly considers following aspect:
S21: this research work only considers the text function of user's model, does not consider community's scale, the society ground of publisher
The factors such as position and the achievement obtained, even if these factors are associated with the quality of model.It is assumed that the note after our processing
Son is all good model, has clearly structure, distinct theme and readily comprehensible content.We extract from model
Several data characteristicses below.
S22: describing (body) the problem of model, according to the content of the model of publisher's publication, it can be observed that in problem
The scale of appearance, comprising how many a clause's descriptions, the complexity of semantic relation and the selection of word, are to the detailed of model title
It explains.Problem describes the text feature basic as model, can therefrom excavate much hiding useful information.
S23: the title (title) of model, important feature one of of the title as user's model, the content of text in title
It is the summary that problem is described in model, comprising many crucial important contents, title facilitates answerer faster for model point
Class shortens turnaround time.
S24: the label (tags) of model determines which kind of other label user's model belongs to, get rid of do not need and
Meaningless label.Label has a degree of influence for the turnaround time of model, and label not only can be preferably by model
Classification, and answerer can also rapidly find according to label and oneself be good at and the problem of domain of interest, help to obtain
Good problem answers.
S25: the creation time (time-rate) of model, as a scalar for measuring predicted time, we pass through understanding
The creation time of user's model come determine and predict by how long, just there is first received answer to occur.
S3: the data characteristics of model is converted into vector using doc2vec model, as full Connection Neural Network model
Input；
The main working principle of Doc2vec model is as follows:
S31: by the body of model, title, tags feature regards the set of one group of sentence and word as, for what is given
Sentence sequence d1,d2,d3,...,dTWith word sequence w1,w2,w3,...,wT, doc2vec vector models are the generation of each word
Term vector W is each sentence generation sentence vector D.
S32: sentence vector sum term vector, which is summed or is averaging, obtains feature, predicts next word in sentence；
S33: making the average log maximization of a vector sum term vector by softmax function, i.e., so thatThere is maximum value, i.e.,
S34: wherein yiFor the probability for predicting each word, y is calculatediFormula:
Y=b+Uh (dt-k,...,dt+k,wt-k,...,wt+k；D, W), U and b here is parameter, and h is by Dt-k,...,
Dt+k,wt-k,...,wt+kIt sums or asks and is flat
S4: using the response time of full Connection Neural Network model prediction model, using the feature vector of model as connecting entirely
The input for connecing neural network goes out model using sigmoid function prediction by the continuous study and update of neural network weight
Response time；
Wherein the training step of full Connection Neural Network model is as follows:
The structure of neural network is made of several neurons, and input layer feature vector is hidden by weight computing
The output valve of layer, hidden layer obtain the output valve of neural network by weight computing.
According to the structure chart of neural network: wherein x1,x2,...,xi...,xdFor input layer, neuron number is
D, b11,b12,...,b1h...,b1qFor the neuron of first layer hidden layer, neuron number q, vihFor input layer and hidden layer
Connection weight, b21,b22,...,b2h...,b2sFor the neuron of second layer hidden layer, neuron number s, bihIt is hiding
Connection weight between layer, yjFor output valve, whjConnection weight between hidden layer and output layer, b', b ", b " ' are respectively
The bias of neural network.
S41: the calculating process of neural network is as follows:
The input value of h-th of hidden neuron of first layer are as follows:
The input value of h-th of hidden neuron of the second layer are as follows:
The input value of j-th of output neuron are as follows:
The prediction output valve of neural network isWhereinFor sigmoid activation primitive,
S42: the error of neural network are as follows:
S43: the more new formula of weight is derived according to error function:
η is learning rate, and bias must update ibid.
S5: is predicted to model problem new in Stack Overflow its response time.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
Anyone skilled in the art in the technical scope disclosed by the present invention, according to the technique and scheme of the present invention and its
Inventive concept is subject to equivalent substitution or change, should be covered by the protection scope of the present invention.
Claims (2)
1. a kind of method of multi-semantic meaning information converged network prediction model response time, characterized by comprising:
It chooses N model and extracts the data characteristics of each model, the problem of wherein data characteristics includes model description, model
Title, the label of model, the creation time of model and model week characteristic information, utilize model data characteristics predict model
Response time；
The data characteristics of model is converted into vector using doc2vec model and vector is input to full Connection Neural Network mould
The response time of sigmoid function prediction model is utilized in type.
2. according to the method described in claim 1, it is further characterized in that: vector is input in full Connection Neural Network model sharp
With when response time of sigmoid function prediction model specifically in the following way:
Full Connection Neural Network model is constructed, including an input layer, two hidden layers and an output layer；
The feature vector of model is input in full Connection Neural Network model, which learn and training updates nerve
Weight and threshold value, the deconditioning when error reaches sets requirement between member；
Model parameter is optimized using optimizer during training, utilizes trained full Connection Neural Network model
Predict the response time of model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910770984.4A CN110472247A (en) | 2019-08-20 | 2019-08-20 | A kind of method of multi-semantic meaning information converged network prediction model response time |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910770984.4A CN110472247A (en) | 2019-08-20 | 2019-08-20 | A kind of method of multi-semantic meaning information converged network prediction model response time |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110472247A true CN110472247A (en) | 2019-11-19 |
Family
ID=68513075
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910770984.4A Pending CN110472247A (en) | 2019-08-20 | 2019-08-20 | A kind of method of multi-semantic meaning information converged network prediction model response time |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110472247A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113837457A (en) * | 2021-09-14 | 2021-12-24 | 上海任意门科技有限公司 | Method, computing device and storage medium for predicting interactive behavior state of posts |
-
2019
- 2019-08-20 CN CN201910770984.4A patent/CN110472247A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113837457A (en) * | 2021-09-14 | 2021-12-24 | 上海任意门科技有限公司 | Method, computing device and storage medium for predicting interactive behavior state of posts |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gilbert et al. | Simulating societies: the computer simulation of social phenomena | |
| US11514305B1 (en) | Intelligent control with hierarchical stacked neural networks | |
| CN113033808B (en) | Deep embedded knowledge tracking method based on problem difficulty and student capability | |
| CN103207855B (en) | For the fine granularity sentiment analysis system and method for product review information | |
| CN110428010B (en) | Knowledge tracking method | |
| CN104318340B (en) | Information visualization methods and intelligent visible analysis system based on text resume information | |
| CN113344053B (en) | Knowledge tracking method based on examination question different composition representation and learner embedding | |
| CN108647233A (en) | A kind of answer sort method for question answering system | |
| CN110991645A (en) | Self-adaptive learning method, system and storage medium based on knowledge model | |
| Lu et al. | CMKT: Concept map driven knowledge tracing | |
| CN115545160B (en) | Knowledge tracking method and system for multi-learning behavior collaboration | |
| CN114021722A (en) | Attention knowledge tracking method integrating cognitive portrayal | |
| CN114201684A (en) | Knowledge graph-based adaptive learning resource recommendation method and system | |
| CN113705218A (en) | Event element gridding extraction method based on character embedding, storage medium and electronic device | |
| CN111710428A (en) | Biomedical text representation method for modeling global and local context interaction | |
| Pan et al. | The role of AI assisted socio-cultural frameworks in academic change and higher education growth | |
| CN112905750A (en) | Generation method and device of optimization model | |
| Rouse | The Radical Naturalism of Naturalistic Philosophy of Science | |
| CN110472247A (en) | A kind of method of multi-semantic meaning information converged network prediction model response time | |
| Aldabbagh et al. | Optimal Learning Behavior Prediction System Based on Cognitive Style Using Adaptive Optimization‐Based Neural Network | |
| Zhang | Sharing of teaching resources for English majors based on ubiquitous learning resource sharing platform and neural network | |
| Shen | Recommendation of online educational resources based on neural network | |
| CN115795015A (en) | Comprehensive knowledge tracking method for enhancing test question difficulty | |
| CN114742292A (en) | Knowledge tracking process-oriented two-state co-evolution method for predicting future performance of students | |
| Suddul et al. | A Smart Virtual Tutor with Facial Emotion Recognition for Online Learning |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191119 |
|
| RJ01 | Rejection of invention patent application after publication |