CN107392664A - Stock Price Fluctuation forecasting system and method based on media information tensor supervised learning - Google Patents

Stock Price Fluctuation forecasting system and method based on media information tensor supervised learning Download PDF

Info

Publication number
CN107392664A
CN107392664A CN201710596662.3A CN201710596662A CN107392664A CN 107392664 A CN107392664 A CN 107392664A CN 201710596662 A CN201710596662 A CN 201710596662A CN 107392664 A CN107392664 A CN 107392664A
Authority
CN
China
Prior art keywords
information
tensor
dimension
stock
media
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
Application number
CN201710596662.3A
Other languages
Chinese (zh)
Inventor
李庆
蒋李灵
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southwestern University Of Finance And Economics
Original Assignee
Southwestern University Of Finance And Economics
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Southwestern University Of Finance And Economics filed Critical Southwestern University Of Finance And Economics
Priority to CN201710596662.3A priority Critical patent/CN107392664A/en
Publication of CN107392664A publication Critical patent/CN107392664A/en
Pending legal-status Critical Current

Links

Classifications

    • 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
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0278—Product appraisal
    • 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
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange

Landscapes

  • Business, Economics & Management (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Theoretical Computer Science (AREA)
  • Marketing (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Economics (AREA)
  • Technology Law (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a kind of Stock Price Fluctuation forecasting system and method based on media information tensor supervised learning, according to media information, builds tensor, the prediction of Stock Price Fluctuation is carried out using supervised learning algorithm.The present invention has advantages below：（1）From three marketing information, stock invester's emotion information, Media News information dimensional attribute information, including qualitatively and quantitatively both sides information source, including official's news information and social media information, so as to more fully analyze the Internet media and the relation of stock market；（2）With tensor come presentation medium information space, so that the correlation between different dimensions information be recorded；（3）Tensor supervised learning algorithm realizes expansion of the computer learning algorithm from vector to tensor.

Description

Stock Price Fluctuation forecasting system and method based on media information tensor supervised learning
Technical field
The present invention relates to a kind of Stock Price Fluctuation forecasting system and method, more particularly to one kind to be based on media information tensor The Stock Price Fluctuation forecasting system and method for supervised learning.
Background technology
With the development of information technology, the Internet media is increasingly becoming the media format of main flow.Particularly with blog, micro- The rise of social media based on rich, socialization news, wikipedia and network forum, its influence of media power are increasingly sharpened. Magnanimity information and fission formula, which are propagated, makes the Internet media generate very important influence to stock market.
In the prior art, the quantitative information in the Internet media is mostly considered when being predicted to Stock Price Fluctuation, such as Transaction value of stock etc., and ignore qualitatively information, the news of such as company, cause Stock Price Fluctuation prediction inaccurate.
When combining different latitude information prediction Stock Price Fluctuation, prior art is by the information characteristics value of different dimensions A super characteristic vector is spliced into, then removes shadow of the detection the Internet media information to stock market with based on vector forecasting model Ring.But because the information of different dimensions is reciprocal effect, and is closely related, complements one another, by the information of different dimensions it Between association cut-off after direct splicing into one be a super characteristic vector, so easily there is dimension disaster.Dimension disaster, i.e., The attribute vector length inputted in machine learning algorithm is long, causes algorithm performance no longer to be improved with increasing for information content, Reduce algorithm effect on the contrary.And when super characteristic vector is spliced, it is believed that the information characteristics of different dimensions are mutual Independent, the interaction between different dimensions information characteristics is reduced, even have ignored the contact between them.
The content of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of based on media information tensor supervised learning Stock Price Fluctuation forecasting system and method, to solve the problems, such as that existing Stock Price Fluctuation precision of prediction is low.
To reach above-mentioned purpose, embodiments of the invention adopt the following technical scheme that：
First aspect, there is provided a kind of Stock Price Fluctuation Forecasting Methodology based on media information tensor supervised learning, it includes：
S1, collect marketing information, Media News information, stock invester's emotion information；
Further, marketing information comes from the master data of company, the scale of transaction value, company such as stock, hands over The information such as easy quantity, can be from the current function situation of the angle reaction company of data.News media's information comes from daily stock News, the content of company's basic condition is contained, investor can be allowed to obtain abundant information, fully understand the situation of company, bag Containing passive or positive content, the irrational investment of investor is easily influenceed.Stock invester's emotion information comes from the societies such as stock forum Media are handed over, social mood is caught by the emotion word in social media.
S2, the information to collection pre-process；
Alternatively, pretreatment includes carrying out text-processing to the Media News information of collection, and stock invester's emotion information of collection is entered Row sentiment analysis.
S3, based on pretreated information architecture tensor；
The present invention information of tensor representation different dimensions, including marketing information, Media News information, stock invester's emotion Information.
Wherein, marketing information dimension to marketing information combination by carrying out vectorization.Media News information is tieed up By constructing news dictionary, news dictionary is made up of degree the noun in news and emotion word.News with dictionary by contrasting, then adds The weight of upper equivalent, newsletter archive information is finally converted to vector.Stock invester's emotion information dimension is by calculating two emotions The factor represents stock invester's emotion information, i.e. emotional factor of the public to the emotional factor of single branch stock and the public to whole market. Three rank tensors are built based on marketing information dimension, Media News information dimension, stock invester's emotion information dimension, to retain difference The potential information of relation and dimension interphase interaction between dimension.
Tensor XtRepresent the investment bad border residing for investor during time t.Tensor Xt∈ℝI 1 ×I 2 ×I 3Presentation medium information is empty Between, its I1、I2、I3Dimension of the tensor per single order is represented, i.e. there is I stock market information1Individual property value, Media News information have I2It is individual Property value, stock invester's emotion information have I3Individual property value.T represents t-th of tensor sample, and the time can be represented corresponding to stock market T investors after the media information that goes, i.e., the subscript different samples set.ai1,i2,i3Some element value in tensor X is represented, It corresponds to be designated as i under the first rank dimension1, i is designated as under second-order dimension2, i is designated as under the 3rd rank dimension3Element.On t Individual tensor sample XtConstruction it is as follows：
ai1,1,1,1≤i1≤I1Marketing information dimension is represented, the property value number of the dimension hasI 1It is individual.
a2, i2,2,1≤i2≤I2Presentation medium news information dimension, the property value number of the dimension haveI 2It is individual.
a3,3, i3,1≤i3≤I3Stock invester's emotion information dimension is represented, the property value number of the dimension hasI 3It is individual.
Tensor X after filling is completetRepresent a training sample, to be predicted share price y corresponding with the sampletRepresent. If N number of sample is trained, it is expressed as in training sample, corresponding desired value y values are expressed as。
S4, the tensor to construction carry out tensor resolution and reconstruct；
Decomposed using Tucker and tensor is decomposed, same order does not carry out mould exhibition to tensor media information space according to tensor first Matrix is opened, using the length of corresponding rank as expansion matrix column, the length sum of remaining rank is as row, tensor inner element value root Structural plan is deployed successively accordingly, and the Single cell fusion of tensor not same order information is just completed while expansion.Again by mould Deploy the decomposition of matrix, merge the information of not same order again, and obtain the factor matrix of corresponding dimensional information.
Tensor after decomposition, it is necessary to be reconstructed again to it.It is multiplied by core tensor with factor matrix, allows different dimensions It is complementary to one another between information, potential information will show in tensor after reconstitution, as final training sample, last handle The tensor that reconstruct is completed, which is put into tensor supervised learning algorithm, carries out model training and model prediction.
S5, based on marketing information and reconstruct after tensor, using tensor supervised learning algorithm to volatility carry out Training and prediction.
Specifically, supervised learning algorithm is expanded from vector to tensor, based on the tensor of reconstruct to the market environment Under volatility be trained and predict.
On the other hand, there is provided a kind of Stock Price Fluctuation forecasting system based on media information tensor supervised learning, it is wrapped Include：
Information collection module, for collecting marketing information, Media News information, stock invester's emotion information；
Further, marketing information comes from the master data of company, the scale of transaction value, company such as stock, hands over The information such as easy quantity, can be from the current function situation of the angle reaction company of data.News media's information comes from daily stock News, the content of company's basic condition is contained, investor can be allowed to obtain abundant information, fully understand the situation of company, bag Containing passive or positive content, the irrational investment of investor is easily influenceed.Stock invester's emotion information comes from the societies such as stock forum Media are handed over, social mood is caught by the emotion word in social media.
Pretreatment module, for being pre-processed to the information of collection；
Alternatively, pretreatment includes carrying out text-processing to the Media News information of collection, and stock invester's emotion information of collection is entered Row sentiment analysis.
Tensor builds module, for based on pretreated information architecture tensor；
The present invention information of tensor representation different dimensions, including marketing information, Media News information, stock invester's emotion Information.
Wherein, marketing information dimension to marketing information combination by carrying out vectorization.Media News information is tieed up By constructing news dictionary, news dictionary is made up of degree the noun in news and emotion word.News with dictionary by contrasting, then adds The weight of upper equivalent, newsletter archive information is finally converted to vector.Stock invester's emotion information dimension is by calculating two emotions The factor represents stock invester's emotion information, i.e. emotional factor of the public to the emotional factor of single branch stock and the public to whole market. Three rank tensors are built based on marketing information dimension, Media News information dimension, stock invester's emotion information dimension, to retain difference The potential information of relation and dimension interphase interaction between dimension.
Tensor XtRepresent the investment bad border residing for investor during time t.Tensor Xt∈ℝI 1 ×I 2 ×I 3Presentation medium information is empty Between, its I1、I2、I3Dimension of the tensor per single order is represented, i.e. there is I stock market information1Individual property value, Media News information have I2It is individual Property value, stock invester's emotion information have I3Individual property value.T represents t-th of tensor sample, and the time can be represented corresponding to stock market T investors after the media information that goes, i.e., the subscript different samples set.ai1,i2,i3Some element value in tensor X is represented, It corresponds to be designated as i under the first rank dimension1, i is designated as under second-order dimension2, i is designated as under the 3rd rank dimension3Element.On t Individual tensor sample XtConstruction it is as follows：
ai1,1,1,1≤i1≤I1Marketing information dimension is represented, the property value number of the dimension hasI 1It is individual.
a2, i2,2,1≤i2≤I2Presentation medium news information dimension, the property value number of the dimension haveI 2It is individual.
a3,3, i3,1≤i3≤I3Stock invester's emotion information dimension is represented, the property value number of the dimension hasI 3It is individual.
Tensor X after filling is completetRepresent a training sample, to be predicted share price y corresponding with the sampletRepresent. If N number of sample is trained, it is expressed as in training sample, corresponding desired value y values are expressed as。
Tensor processing module, for carrying out tensor resolution and reconstruct to the tensor of construction；
Decomposed using Tucker and tensor is decomposed, same order does not carry out mould exhibition to tensor media information space according to tensor first Matrix is opened, using the length of corresponding rank as expansion matrix column, the length sum of remaining rank is as row, tensor inner element value root Structural plan is deployed successively accordingly, and the Single cell fusion of tensor not same order information is just completed while expansion.Again by mould Deploy the decomposition of matrix, merge the information of not same order again, and obtain the factor matrix of corresponding dimensional information.
Tensor after decomposition, it is necessary to be reconstructed again to it.It is multiplied by core tensor with factor matrix, allows different dimensions It is complementary to one another between information, potential information will show in tensor after reconstitution, as final training sample, last handle The tensor that reconstruct is completed, which is put into tensor supervised learning algorithm, carries out model training and model prediction.
Training and prediction module, for based on the tensor after marketing information and reconstruct, being calculated using tensor supervised learning Method is trained and predicted to volatility.
Specifically, supervised learning algorithm is expanded from vector to tensor, based on the tensor of reconstruct to the market environment Under volatility be trained and predict.
The present invention proposes a kind of Stock Price Fluctuation forecasting system and method tool based on media information tensor supervised learning There is following advantage：（1）From three marketing information, stock invester's emotion information, Media News information dimensional attribute information, bag Qualitatively and quantitatively both sides information source, including official's news information and social media information is included, so as to more fully analyze The Internet media and the relation of stock market；（2）With tensor come presentation medium information space, so that by between different dimensions information Correlation record；（3）Tensor supervised learning algorithm realizes expansion of the computer learning algorithm from vector to tensor； To reach more preferable Prediction of Stock Price effect, decision references are provided for stock invester and investment company.
Brief description of the drawings
In order to illustrate the technical solution of the embodiments of the present invention more clearly, required use in being described below to embodiment Accompanying drawing be briefly described, it should be apparent that, drawings in the following description are only some embodiments of the present invention, for this For the those of ordinary skill of field, on the premise of not paying creative work, it can also be obtained according to these accompanying drawings other Accompanying drawing.
Fig. 1 is the prediction steps block diagram of the present invention；
Fig. 2 is the forecasting system figure of the present invention.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete Site preparation describes, it is clear that described embodiment is only part of the embodiment of the present invention, rather than whole embodiments.It is based on Embodiment in the present invention, those of ordinary skill in the art are obtained every other under the premise of creative work is not made Embodiment, belong to the scope of protection of the invention.
The present invention is based on media information, is particularly based on media information tensor supervised learning algorithm and Stock Price Fluctuation is carried out Prediction, decision-making is provided for financial market.
With reference to the accompanying drawings and examples, the embodiment of the present invention is described in further detail.Implement below Example is used to illustrate the present invention, but is not limited to the scope of the present invention.
The embodiment of the present invention provides a kind of Stock Price Fluctuation Forecasting Methodology based on media information tensor supervised learning, such as Shown in Fig. 1, specifically include：
S1, collect marketing information, Media News information, stock invester's emotion information；
Further, marketing information comes from the master data of company, the scale of transaction value, company such as stock, hands over The information such as easy quantity, can be from the current function situation of the angle reaction company of data.News media's information comes from daily stock News, the content of company's basic condition is contained, investor can be allowed to obtain abundant information, fully understand the situation of company, bag Containing passive or positive content, the irrational investment of investor is easily influenceed.Stock invester's emotion information comes from the societies such as stock forum Media are handed over, social mood is caught by the emotion word in social media.
S2, the information to collection pre-process；
Alternatively, pretreatment includes carrying out text-processing to the Media News information of collection, and stock invester's emotion information of collection is entered Row sentiment analysis.
S3, based on pretreated information architecture tensor；
The present invention information of tensor representation different dimensions, including marketing information, Media News information, stock invester's emotion Information.
Wherein, marketing information dimension to marketing information combination by carrying out vectorization.Media News information is tieed up By constructing news dictionary, news dictionary is made up of degree the noun in news and emotion word.News with dictionary by contrasting, then adds The weight of upper equivalent, newsletter archive information is finally converted to vector.Stock invester's emotion information dimension is by calculating two emotions The factor represents stock invester's emotion information, i.e. emotional factor of the public to the emotional factor of single branch stock and the public to whole market. Three rank tensors are built based on marketing information dimension, Media News information dimension, stock invester's emotion information dimension, to retain difference The potential information of relation and dimension interphase interaction between dimension.
Tensor XtRepresent the investment bad border residing for investor during time t.Tensor Xt∈ℝI 1 ×I 2 ×I 3Presentation medium information is empty Between, its I1、I2、I3Dimension of the tensor per single order is represented, i.e. there is I stock market information1Individual property value, Media News information have I2It is individual Property value, stock invester's emotion information have I3Individual property value.T represents t-th of tensor sample, and the time can be represented corresponding to stock market T investors after the media information that goes, i.e., the subscript different samples set.ai1,i2,i3Some element value in tensor X is represented, It corresponds to be designated as i under the first rank dimension1, i is designated as under second-order dimension2, i is designated as under the 3rd rank dimension3Element.On t Individual tensor sample XtConstruction it is as follows：
ai1,1,1,1≤i1≤I1Marketing information dimension is represented, the property value number of the dimension hasI 1It is individual.
a2, i2,2,1≤i2≤I2Presentation medium news information dimension, the property value number of the dimension haveI 2It is individual.
a3,3, i3,1≤i3≤I3Stock invester's emotion information dimension is represented, the property value number of the dimension hasI 3It is individual.
Tensor X after filling is completetRepresent a training sample, to be predicted share price y corresponding with the sampletRepresent. If N number of sample is trained, it is expressed as in training sample, corresponding desired value y values are expressed as。
S4, the tensor to construction carry out tensor resolution and reconstruct；
Decomposed using Tucker and tensor is decomposed, same order does not carry out mould exhibition to tensor media information space according to tensor first Matrix is opened, using the length of corresponding rank as expansion matrix column, the length sum of remaining rank is as row, tensor inner element value root Structural plan is deployed successively accordingly, and the Single cell fusion of tensor not same order information is just completed while expansion.Again by mould Deploy the decomposition of matrix, merge the information of not same order again, and obtain the factor matrix of corresponding dimensional information.
Tensor after decomposition, it is necessary to be reconstructed again to it.It is multiplied by core tensor with factor matrix, allows different dimensions It is complementary to one another between information, potential information will show in tensor after reconstitution, as final training sample, last handle The tensor that reconstruct is completed, which is put into tensor supervised learning algorithm, carries out model training and model prediction.
S5, based on marketing information and reconstruct after tensor, using tensor supervised learning algorithm to volatility carry out Training and prediction.
Specifically, supervised learning algorithm is expanded from vector to tensor, based on the tensor of reconstruct to the market environment Under volatility be trained and predict.
Another embodiment of the present invention provides a kind of Stock Linkage evaluation system based on enterprise network, as shown in Fig. 2 tool Body includes：
Information collection module, for collecting marketing information, Media News information, stock invester's emotion information；
Further, marketing information comes from the master data of company, the scale of transaction value, company such as stock, hands over The information such as easy quantity, can be from the current function situation of the angle reaction company of data.News media's information comes from daily stock News, the content of company's basic condition is contained, investor can be allowed to obtain abundant information, fully understand the situation of company, bag Containing passive or positive content, the irrational investment of investor is easily influenceed.Stock invester's emotion information comes from the societies such as stock forum Media are handed over, social mood is caught by the emotion word in social media.
Pretreatment module, for being pre-processed to the information of collection；
Alternatively, pretreatment includes carrying out text-processing to the Media News information of collection, and stock invester's emotion information of collection is entered Row sentiment analysis.
Tensor builds module, for based on pretreated information architecture tensor；
The present invention information of tensor representation different dimensions, including marketing information, Media News information, stock invester's emotion Information.
Wherein, marketing information dimension to marketing information combination by carrying out vectorization.Media News information is tieed up By constructing news dictionary, news dictionary is made up of degree the noun in news and emotion word.News with dictionary by contrasting, then adds The weight of upper equivalent, newsletter archive information is finally converted to vector.Stock invester's emotion information dimension is by calculating two emotions The factor represents stock invester's emotion information, i.e. emotional factor of the public to the emotional factor of single branch stock and the public to whole market. Three rank tensors are built based on marketing information dimension, Media News information dimension, stock invester's emotion information dimension, to retain difference The potential information of relation and dimension interphase interaction between dimension.
Tensor XtRepresent the investment bad border residing for investor during time t.Tensor Xt∈ℝI 1 ×I 2 ×I 3Presentation medium information is empty Between, its I1、I2、I3Dimension of the tensor per single order is represented, i.e. there is I stock market information1Individual property value, Media News information have I2It is individual Property value, stock invester's emotion information have I3Individual property value.T represents t-th of tensor sample, and the time can be represented corresponding to stock market T investors after the media information that goes, i.e., the subscript different samples set.ai1,i2,i3Some element value in tensor X is represented, It corresponds to be designated as i under the first rank dimension1, i is designated as under second-order dimension2, i is designated as under the 3rd rank dimension3Element.On t Individual tensor sample XtConstruction it is as follows：
ai1,1,1,1≤i1≤I1Marketing information dimension is represented, the property value number of the dimension hasI 1It is individual.
a2, i2,2,1≤i2≤I2Presentation medium news information dimension, the property value number of the dimension haveI 2It is individual.
a3,3, i3,1≤i3≤I3Stock invester's emotion information dimension is represented, the property value number of the dimension hasI 3It is individual.
Tensor X after filling is completetRepresent a training sample, to be predicted share price y corresponding with the sampletRepresent. If N number of sample is trained, it is expressed as in training sample, corresponding desired value y values are expressed as。
Tensor processing module, for carrying out tensor resolution and reconstruct to the tensor of construction；
Decomposed using Tucker and tensor is decomposed, same order does not carry out mould exhibition to tensor media information space according to tensor first Matrix is opened, using the length of corresponding rank as expansion matrix column, the length sum of remaining rank is as row, tensor inner element value root Structural plan is deployed successively accordingly, and the Single cell fusion of tensor not same order information is just completed while expansion.Again by mould Deploy the decomposition of matrix, merge the information of not same order again, and obtain the factor matrix of corresponding dimensional information.
Tensor after decomposition, it is necessary to be reconstructed again to it.It is multiplied by core tensor with factor matrix, allows different dimensions It is complementary to one another between information, potential information will show in tensor after reconstitution, as final training sample, last handle The tensor that reconstruct is completed, which is put into tensor supervised learning algorithm, carries out model training and model prediction.
Training and prediction module, for based on the tensor after marketing information and reconstruct, being calculated using tensor supervised learning Method is trained and predicted to volatility.
Specifically, supervised learning algorithm is expanded from vector to tensor, based on the tensor of reconstruct to the market environment Under volatility be trained and predict.
The foregoing is only a specific embodiment of the invention, but protection scope of the present invention is not limited thereto, any Those familiar with the art the invention discloses technical scope in, change or replacement can be readily occurred in, should all be contained Cover within protection scope of the present invention.Therefore, protection scope of the present invention described should be defined by scope of the claims.

Claims (10)

  1. A kind of 1. Stock Price Fluctuation Forecasting Methodology based on media information tensor supervised learning, it is characterised in that：It is included such as Lower step：
    S1, collect marketing information, Media News information, stock invester's emotion information；
    S2, the information to collection pre-process；
    S3, based on pretreated information architecture tensor；
    S4, the tensor to construction carry out tensor resolution and reconstruct；
    S5, based on marketing information and reconstruct after tensor, volatility is trained using tensor supervised learning algorithm And prediction.
  2. 2. Stock Price Fluctuation Forecasting Methodology according to claim 1, it is characterised in that：The pretreatment is included to collecting The Media News information carry out text-processing, sentiment analysis is carried out to stock invester's emotion information of collection.
  3. 3. Stock Price Fluctuation Forecasting Methodology according to claim 1, it is characterised in that：The step of structure tensor, wraps Include：Marketing information dimension, Media News information dimension, stock invester's emotion information dimension are built, is tieed up based on marketing information Degree, Media News information dimension, stock invester's emotion information dimension build three rank tensors.
  4. 4. Stock Price Fluctuation Forecasting Methodology according to claim 1, it is characterised in that：The tensor resolution method is Tucker is decomposed, and the reconstructing method is that core tensor is multiplied with factor matrix.
  5. 5. Stock Price Fluctuation Forecasting Methodology according to claim 3, it is characterised in that：The marketing information dimension To carry out vectorization to marketing information combination；The Media News information dimension is to contrast news and news dictionary, is added The weight of upper equivalent, is converted to vector；Stock invester's emotion information dimension by calculate the public to the emotion of single branch stock because The emotional factor of son and the public to whole market；
    Tensor XtRepresent the investment bad border residing for investor, tensor X during time tt∈ℝI 1 ×I 2 ×I 3Presentation medium information space, its I1、I2、I3Dimension of the tensor per single order is represented, i.e. there is I stock market information1Individual property value, Media News information have I2Individual attribute Value, stock invester's emotion information have I3Individual property value；T represents t-th of tensor sample, can represent that time t is thrown corresponding to stock market Money person after the media information that goes, i.e., the subscript different samples set；ai1,i2,i3Some element value in tensor X is represented, its Correspond to be designated as i under the first rank dimension1, i is designated as under second-order dimension2, i is designated as under the 3rd rank dimension3Element, on t-th Tensor sample XtConstruction it is as follows：
    ai1,1,1,1≤i1≤I1Marketing information dimension is represented, the property value number of the dimension hasI 1It is individual；
    a2, i2,2,1≤i2≤I2Presentation medium news information dimension, the property value number of the dimension haveI 2It is individual；
    a3,3, i3,1≤i3≤I3Stock invester's emotion information dimension is represented, the property value number of the dimension hasI 3It is individual.
  6. A kind of 6. Stock Price Fluctuation forecasting system based on media information tensor supervised learning, it is characterised in that：It includes：
    Information collection module, for collecting marketing information, Media News information, stock invester's emotion information；
    Pretreatment module, for being pre-processed to the information of collection；
    Tensor builds module, for based on pretreated information architecture tensor；
    Tensor processing module, tensor resolution and reconstruct are carried out to the tensor of construction；
    Training and prediction module, based on the tensor after marketing information and reconstruct, using tensor supervised learning algorithm to share price Fluctuation is trained and predicted.
  7. 7. the Stock Price Fluctuation forecasting system according to claim 6 based on media information tensor supervised learning, it is special Sign is：The pretreatment includes carrying out text-processing to the Media News information of collection, to stock invester's feelings of collection Feel information and carry out sentiment analysis.
  8. 8. the Stock Price Fluctuation forecasting system according to claim 6 based on media information tensor supervised learning, it is special Sign is：The structure tensor includes：Build marketing information dimension, Media News information dimension, stock invester's emotion information dimension Degree, three rank tensors are built based on marketing information dimension, Media News information dimension, stock invester's emotion information dimension.
  9. 9. the Stock Price Fluctuation forecasting system according to claim 6 based on media information tensor supervised learning, it is special Sign is：The tensor resolution decomposes for Tucker, and the tensor is reconstructed into core tensor and is multiplied with factor matrix.
  10. 10. the Stock Price Fluctuation forecasting system according to claim 8 based on media information tensor supervised learning, it is special Sign is：The marketing information dimension is to carry out vectorization to marketing information combination；The Media News information dimension Degree is to contrast news and news dictionary, plus the weight of equivalent, is converted to vector；Stock invester's emotion information dimension passes through Calculate emotional factor of the public to the emotional factor of single branch stock and the public to whole market；
    Tensor XtRepresent the investment bad border residing for investor, tensor X during time tt∈ℝI 1 ×I 2 ×I 3Presentation medium information space, its I1、I2、I3Dimension of the tensor per single order is represented, i.e. there is I stock market information1Individual property value, Media News information have I2Individual attribute Value, stock invester's emotion information have I3Individual property value；T represents t-th of tensor sample, can represent that time t is thrown corresponding to stock market Money person after the media information that goes, i.e., the subscript different samples set；ai1,i2,i3Some element value in tensor X is represented, its Correspond to be designated as i under the first rank dimension1, i is designated as under second-order dimension2, i is designated as under the 3rd rank dimension3Element, on t-th Tensor sample XtConstruction it is as follows：
    ai1,1,1,1≤i1≤I1Marketing information dimension is represented, the property value number of the dimension hasI 1It is individual；
    a2, i2,2,1≤i2≤I2Presentation medium news information dimension, the property value number of the dimension haveI 2It is individual；
    a3,3, i3,1≤i3≤I3Stock invester's emotion information dimension is represented, the property value number of the dimension hasI 3It is individual.
CN201710596662.3A 2017-07-20 2017-07-20 Stock Price Fluctuation forecasting system and method based on media information tensor supervised learning Pending CN107392664A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710596662.3A CN107392664A (en) 2017-07-20 2017-07-20 Stock Price Fluctuation forecasting system and method based on media information tensor supervised learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710596662.3A CN107392664A (en) 2017-07-20 2017-07-20 Stock Price Fluctuation forecasting system and method based on media information tensor supervised learning

Publications (1)

Publication Number Publication Date
CN107392664A true CN107392664A (en) 2017-11-24

Family

ID=60336207

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710596662.3A Pending CN107392664A (en) 2017-07-20 2017-07-20 Stock Price Fluctuation forecasting system and method based on media information tensor supervised learning

Country Status (1)

Country Link
CN (1) CN107392664A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108804392A (en) * 2018-05-30 2018-11-13 福州大学 A kind of traffic data tensor fill method based on space-time restriction
CN109598380A (en) * 2018-12-03 2019-04-09 郑州云海信息技术有限公司 A kind of method and system of polynary real-time time series data prediction
CN110400225A (en) * 2019-07-29 2019-11-01 北京北信源软件股份有限公司 A kind of market value of stock management method
CN111159200A (en) * 2019-12-31 2020-05-15 华中科技大学鄂州工业技术研究院 Data storage method and device based on deep learning
CN111539770A (en) * 2020-04-27 2020-08-14 启迪数华科技有限公司 Intelligent data asset assessment method and system

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108804392A (en) * 2018-05-30 2018-11-13 福州大学 A kind of traffic data tensor fill method based on space-time restriction
CN109598380A (en) * 2018-12-03 2019-04-09 郑州云海信息技术有限公司 A kind of method and system of polynary real-time time series data prediction
CN110400225A (en) * 2019-07-29 2019-11-01 北京北信源软件股份有限公司 A kind of market value of stock management method
CN111159200A (en) * 2019-12-31 2020-05-15 华中科技大学鄂州工业技术研究院 Data storage method and device based on deep learning
CN111159200B (en) * 2019-12-31 2023-10-17 华中科技大学鄂州工业技术研究院 Data storage method and device based on deep learning
CN111539770A (en) * 2020-04-27 2020-08-14 启迪数华科技有限公司 Intelligent data asset assessment method and system
CN111539770B (en) * 2020-04-27 2023-06-16 国云数字科技(重庆)有限公司 Intelligent evaluation method and system for data assets

Similar Documents

Publication Publication Date Title
CN107392664A (en) Stock Price Fluctuation forecasting system and method based on media information tensor supervised learning
Hammou et al. Towards a real-time processing framework based on improved distributed recurrent neural network variants with fastText for social big data analytics
Zhang et al. Improving stock market prediction via heterogeneous information fusion
CN109033068A (en) It is used to read the method, apparatus understood and electronic equipment based on attention mechanism
CN111339415A (en) Click rate prediction method and device based on multi-interactive attention network
CN106447066A (en) Big data feature extraction method and device
CN104487936B (en) Method and system for carrying out area of computer aided consumption to the information from application data file
CN110968695A (en) Intelligent labeling method, device and platform based on active learning of weak supervision technology
CN106445988A (en) Intelligent big data processing method and system
CN115860006B (en) Aspect-level emotion prediction method and device based on semantic syntax
Pham et al. Deep learning, graph-based text representation and classification: a survey, perspectives and challenges
CN114610989B (en) Personalized thesis recommendation method and system based on heterogeneous graph dynamic information compensation
CN113298661B (en) Artificial intelligence based product recommendation method and device and electronic equipment
Zheng et al. Learn from model beyond fine-tuning: A survey
Malakouti et al. Predicting patient’s diagnoses and diagnostic categories from clinical-events in EHR data
Demirbaga HTwitt: a hadoop-based platform for analysis and visualization of streaming Twitter data
CN111221881B (en) User characteristic data synthesis method and device and electronic equipment
Rai et al. Analysis of customer churn prediction in telecom sector using cart algorithm
Dang et al. Adoption of social media search systems: An IS success model perspective
CN117216535A (en) Training method, device, equipment and medium for recommended text generation model
Zhang et al. Towards employing native information in citation function classification
Zarzour et al. Sentiment analysis based on deep learning methods for explainable recommendations with reviews
Mahmud et al. Deep learning based sentiment analysis from Bangla text using glove word embedding along with convolutional neural network
Liu et al. Aspect-Based Pair-Wise Opinion Generation in Chinese automotive reviews: design of the task, dataset and model
Ganesan et al. Deep learning approaches for accurate sentiment analysis of online consumer feedback

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: 20171124

RJ01 Rejection of invention patent application after publication