CN103226554A - Automatic stock matching and classifying method and system based on news data - Google Patents

Abstract

The invention relates to a matching and classifying method and system for stock information, in particular to an automatic stock matching and classifying method and system based on news data. The automatic matching and classifying method is characterized by comprising the following steps: establishing a local data base; performing word segmentation and screening on historical news data, extracting key word sequences, constructing sequence models for individual share key word sequence correlation, calculating the correlation between the individual shares, and classifying stocks by combining a cluster analysis algorithm; and performing word segmentation and screening on real-time news data, extracting a real-time key word sequence, calculating sequences for real-time key word sequence correlation, and performing automatic matching with the sequence models for individual share key word sequence correlation. The automatic stock matching and classifying method and system adopt the stock key word sequence excavation technology based on the news data to achieve automatic classification of the stocks; the method is comprehensive, accurate, simple, convenient and feasible, and provides better investment reference for investors; and stocks with higher matching degree are given automatically aiming at breaking news events.

Description

Based on the automatic match classifying method and system of the stock of news data

Technical field

The present invention relates to the match classifying method and system of stock information, relate in particular to the automatic match classifying method and system of a kind of stock based on news data.

Background technology

Traditional internet search engine technology uses particular keywords that Website page is marked, and provides Search Results by user search keyword and webpage keyword are mated.Along with the development of Web technology, all have a large amount of pictures, blog, video distribution on the net every day, and the extreme blast of information makes and is difficult to help the information that people find them to need by simple keyword.And the appearance of recommended engine, the mode that makes the user obtain information is transformed into the more senior abundanter INFORMATION DISCOVERY of contextual information that more meets people's use habit from simple keyword search.By excavating the correlativity of webpage and user data, structure webpage and user's keyword sequence, calculate between webpage, the user on this basis and the correlativity between webpage and the user, thereby obtain Search Results more accurately, and continue to optimize Search Results according to user behavior.

Traditional stock classification is static often, normally based on a certain feature stock is divided, for example based on industry, region, notion, style index etc.Yet the business of listed company is developing always, and the main management direction also can change to some extent, and in addition, the stock of different plates is not all inequality to the feedback of same media event, and therefore, static stock classification method can not reflect the attribute of stock sometimes well.And traditional selecting stocks normally judged according to individual's experience, and general stock invester can grind newspaper and makes oneself judgement by reading relevant news; Yet in the face of the news data of ever-increasing stock quantity and magnanimity, individual's energy and experience are limited, are difficult to the stock market is observed all-sidedly and accurately and analyzed.

Summary of the invention

The invention provides the automatic match classifying method and system of a kind of stock based on news data, from the magnanimity news data, excavation can embody the keyword sequence of stock attribute more comprehensively, by the correlativity of analyzing between the stock keyword sequence stock is classified, by the keyword sequence of analysis news data and the correlativity between the stock keyword sequence, provide stock simultaneously with the news data optimum matching.

In order to achieve the above object, the present invention adopts following technical scheme:

The automatic match classifying method of a kind of stock based on news data,, it is characterized in that comprising the steps:

(A) set up a local data base;

(B) grasp the historical news data of relevant personal share at least first from the external world, be stored in the local data base, set up corresponding personal share sequence number, and each then historical news data is carried out participle, screening, extract wherein relevant keyword sequence with personal share;

(C) frequency that each keyword occurs in the keyword sequence that statistics is relevant with above-mentioned personal share is calculated corresponding keyword relative coefficient on this basis, structure personal share keyword sequence correlativity series model, the line parameter optimization of going forward side by side; Calculate correlativity between the personal share according to personal share keyword sequence correlativity series model then, and,, stock is classified in conjunction with cluster algorithm according to the correlativity between the personal share; And

(D) grasp real-time news data from the external world, be stored in the local data base, this news data is carried out participle, screening, extract real-time keyword sequence, calculate real-time keyword sequence correlativity sequence, by the real-time keyword sequence correlativity sequence of real-time news data and the personal share keyword sequence correlativity series model among the above-mentioned steps C are mated automatically, choose the highest personal share of correlativity as final best matching result.

In described step B, the personal share sequence number is stock A, stock B ..., stock N; Keyword sequence is a keyword 1, keyword 2 ... keyword M; In described step C, personal share keyword sequence correlativity series model is write as matrix form and is: C _NMRelative coefficient between expression stock N and the keyword M.

Correlativity between the described personal share, promptly the computing formula of Euclidean distance is:

Wherein, || vector|| represents the Euclid norm of vector; The correlativity sequence of described stock A is vector _A=[C _A1C _A2C _AM], then || vector _A|| account form be:

In described step D, the computing method of keyword sequence correlativity sequence are as follows in real time: vector _Event=[e ₁e ₂E _M], wherein, e _MRepresent the relative coefficient between real-time news data and the keyword M.

In described step D, the real-time keyword sequence correlativity sequence of real-time news data and the computing method that personal share keyword sequence correlativity series model mates automatically are as follows:

Wherein, || vector|| represents the Euclid norm of vector; The correlativity sequence of described stock A is vector _A=[C _A1C _A2C _AM], then || vector _A|| account form be: $\left|\right|{\mathrm{vector}}_A\left|\right|=\sqrt{C_{A1}^2+C_{A2}^2+\·\·\·+C_{\mathrm{AM}}^2}.$

In described step C, described parameter optimization is keyword sequence to be optimized with existing ripe stock classification result by the comparison classification results.

The automatic match classifying of a kind of stock based on news data system is characterized in that it comprises:

One local data base is used to store historical news data and real-time news data, and sets up corresponding personal share sequence number and keyword sequence within it;

One participle screening module is carried out participle, screening to each then historical news data, extracts wherein relevant with personal share keyword sequence, and real-time news data is carried out participle, screening, extracts real-time keyword sequence;

One statistical generic module, be used for adding up the frequency that each keyword of keyword sequence occurs, and calculate corresponding keyword relative coefficient, structure personal share keyword sequence correlativity series model, the line parameter optimization of going forward side by side, and the correlativity between the calculating personal share, and stock is classified in conjunction with cluster algorithm;

One automatic matching module, be used to calculate real-time keyword sequence correlativity sequence, automatically mate by real-time keyword sequence correlativity sequence and personal share keyword sequence correlativity series model, choose the highest personal share of correlativity as final best matching result with real-time news data.

The present invention adopts the stock keyword sequence digging technology based on news data, the attribute that reflects stock with keyword sequence objectively, realize automatic classification on this basis to stock, this method comprehensively, accurately, and at the burst media event provide the higher stock of matching degree automatically, method is simple, for the stock invester provides better investment foundation.

Description of drawings

Fig. 1 is a process flow diagram of the present invention.

Embodiment

Now the present invention will be further described in conjunction with the accompanying drawings and embodiments.

As one embodiment of the present invention, referring to Fig. 1, the automatic match classifying method of its stock based on news data comprises the steps:

(A) set up a local data base.

(B) grasp the historical news data of relevant personal share at least first from the external world, be stored in the local data base, the source of historical news data comprises mainstream news portal website, national correlation department bulletin, transaction on exchange data, social networks etc.; Set up corresponding personal share sequence number then, and each then historical news data is carried out participle, screening, rejecting function word and some do not have the participle of quantity of information, extract wherein relevant with personal share keyword sequence; Suppose that the personal share sequence number is stock A, stock B ..., stock N; Keyword sequence is a keyword 1, keyword 2 ... keyword M.

(C) frequency that each keyword occurs in the keyword sequence that statistics is relevant with above-mentioned personal share is calculated corresponding keyword relative coefficient on this basis, structure personal share keyword sequence correlativity series model, the line parameter optimization of going forward side by side.The building method of personal share keyword sequence correlativity series model is as follows:

Suppose structure keyword sequence [keyword 1 keyword 2 keywords 3 ... keyword M], the number of times that these keywords occur in the difference statistical history news data, as shown in the table:

Relative coefficient	Keyword 1	Keyword 2	Keyword 3	Keyword 4	......	Keyword M
							Stock A	20	30	12	5	......	8
Stock B	30	9	25	2	......	1
							......	......	......	......	......	......	......
Stock N	8	60	9	25	......	2

Make normalized by the keyword frequency to each personal share, promptly divided by the root mean square of the quadratic sum of all keyword frequencys of personal share, with stock A be example promptly divided by

Obtain personal share keyword sequence correlation models thus, as follows:

Relative coefficient	Keyword 1	Keyword 2	Keyword 3	Keyword 4	......	Keyword M
							Stock A	0.02	0.3	0.12	0.05	......	0.08
Stock B	0.3	0.09	0.25	0.02	......	0.1?
							......	......	......	......	......	......	......
Stock N	0.08	0.6	0.09	0.25	......	0.02

Being write as matrix form is:

C _NMRelative coefficient between expression stock N and the keyword M.

Calculate correlativity between the personal share according to personal share keyword sequence correlativity series model then, this patent reflects correlativity between sequence by calculating Euclidean distance between keyword correlativity sequence, and the near more expression correlativity of distance is high more:

The keyword correlativity sequence of supposing stock A, stock B is respectively:

$\begin{array}{c}{\mathrm{vector}}_A=\left[\begin{array}{cccc}{C}_{A1}& C_{A2}& \·\·\·& C_{\mathrm{AM}}\end{array}\right]\\ {\mathrm{vector}}_B=\left[\begin{array}{cccc}{C}_{B1}& C_{B2}& \·\·\·& C_{\mathrm{BM}}\end{array}\right]\end{array}$ So, their Euclidean distance can be calculated by following:

Wherein, || vector|| represents the Euclid norm of vector, with || vector _A|| be example, its account form is:

$\left|\right|{\mathrm{vector}}_A\left|\right|=\sqrt{C_{A1}^2+C_{A2}^2+\·\·\·+C_{\mathrm{AM}}^2}.$

Suppose that strand keyword sequence correlativity series model is as follows one by one:

?	Energy-conserving and environment-protective	Water utilities	Instrument	Finance	Interest rate	Lithium battery	School bus	Universal wheel
									Stock A	0.6614	0.2835	0.6614	0.0945	0.1890	0	0	0
Stock B	0.1021	0	0.1021	0.2041	0.1021	0.7144	0.4082	0
									Stock C	0.3094	0	0	0.1031	0.1031	0.5157	0.3094	0.7220

Then can obtain personal share keyword correlativity sequence is respectively:

vector _A＝[0.6614?0.2835?0.6614?0.0945?0.1890?0?0?0]

vector _B＝[0.1021?0?0.1021?0.2041?0.1021?0.7144?0.4082?0]

Vector _C=[0.3094 00 0.1031 0.1031 0.5157 0.3094 0.7220] are calculated the Euclidean distance between them, i.e. correlativity respectively:

D _AB＝||vector _A-vector _B||＝1.0577

D _AC＝||vector _A-vector _C||＝1.2409

D _BC＝||vector _B-vector _C||＝0.7947

This shows that the correlativity between stock B and the C is bigger, and the accuracy that they are classified as a class stock is also higher.

At last, according to the correlativity between the personal share,, stock is classified in conjunction with cluster algorithm.By selecting different keyword sequences, sorting technique ripe in each classification results and the industry is artificially compared, according to results of comparison keyword sequence is optimized.

(D) grasp real-time news data from the external world, be stored in the local data base, the source of news data comprises mainstream news portal website, national correlation department bulletin, transaction on exchange data, social networks etc. in real time, this news data is carried out participle, screening, extract real-time keyword sequence, the computing method of keyword sequence are made as vector with step C in real time _Event=[e ₁e ₂E _M], e _MRepresent the relative coefficient between real-time news data and the keyword M.At last, by the real-time keyword sequence correlativity sequence of real-time news data and the personal share keyword sequence correlativity series model among the above-mentioned steps C are mated automatically, choose correlativity the highest, be that the personal share of Euclidean distance minimum is as final best matching result.

The real-time keyword sequence of supposing certain real-time news data is as shown in the table:

?	Energy-conserving and environment-protective	Water utilities	Instrument	Finance	Interest rate	Lithium battery	School bus	Universal wheel
									Vector	0.7127	0.4454	0.5345	0	0.0891	0	0	0

Be vector _Event=[0.7127 0.4454 0.5345 0 0.0891 00 0], calculate respectively with step C in the Euclidean distance of personal share correlativity sequence of stock A, B, C, then can obtain respectively:

D _A，event＝0.2595

D _B，event＝1.0788

D _C，event＝1.2498，

Can think that thus the optimum matching stock of this real-time news data is stock A.

The automatic match classifying of stock based on news data of the present invention system, it comprises:

Local data base is used to store historical news data and real-time news data, and sets up corresponding personal share sequence number and keyword sequence within it;

Participle screening module is carried out participle, screening to each then historical news data, extracts wherein relevant with personal share keyword sequence, and real-time news data is carried out participle, screening, extracts real-time keyword sequence;

The statistical classification module, be used for adding up the frequency that each keyword of keyword sequence occurs, and calculate corresponding keyword relative coefficient, structure personal share keyword sequence correlativity series model, the line parameter optimization of going forward side by side, and the correlativity between the calculating personal share, and stock is classified in conjunction with cluster algorithm;

Automatic matching module, be used to calculate real-time keyword sequence correlativity sequence, automatically mate by real-time keyword sequence correlativity sequence and personal share keyword sequence correlativity series model, choose the highest personal share of correlativity as final best matching result with real-time news data.

The present invention adopts the stock keyword sequence digging technology based on news data, the attribute that reflects stock with keyword sequence objectively, realize automatic classification on this basis to stock, this method comprehensively, accurately, and at the burst media event provide the higher stock of matching degree automatically, method is simple, for the stock invester provides better investment foundation.

Claims (7)

1. the automatic match classifying method of the stock based on news data is characterized in that comprising the steps:

(A) set up a local data base;

(B) grasp the historical news data of relevant personal share at least first from the external world, be stored in the local data base, set up corresponding personal share sequence number, and each then historical news data is carried out participle, screening, extract wherein relevant keyword sequence with personal share;

(C) frequency that each keyword occurs in the keyword sequence that statistics is relevant with above-mentioned personal share is calculated corresponding keyword relative coefficient on this basis, structure personal share keyword sequence correlativity series model, the line parameter optimization of going forward side by side; Calculate correlativity between the personal share according to personal share keyword sequence correlativity series model then, and,, stock is classified in conjunction with cluster algorithm according to the correlativity between the personal share; And

(D) grasp real-time news data from the external world, be stored in the local data base, this news data is carried out participle, screening, extract real-time keyword sequence, calculate real-time keyword sequence correlativity sequence, by the real-time keyword sequence correlativity sequence of real-time news data and the personal share keyword sequence correlativity series model among the above-mentioned steps C are mated automatically, choose the highest personal share of correlativity as final best matching result.

2. the automatic match classifying method of the stock based on news data according to claim 1 is characterized in that:

In described step B, the personal share sequence number is stock A, stock B ..., stock N; Keyword sequence is a keyword 1, keyword 2 ... keyword M; In described step C, personal share keyword sequence correlativity series model is write as matrix form and is: C _NMRelative coefficient between expression stock N and the keyword M.

3. the automatic match classifying method of the stock based on news data according to claim 2 is characterized in that: the correlativity between the described personal share, and promptly the computing formula of Euclidean distance is:

Wherein, || vector|| represents the Euclid norm of vector;

The correlativity sequence of described stock A is vector _A=[C _A1C _A2C _AM], then || vector _A|| account form be: $\left|\right|{\mathrm{vector}}_A\left|\right|=\sqrt{C_{A1}^2+C_{A2}^2+\·\·\·+C_{\mathrm{AM}}^2}.$

4. the automatic match classifying method of the stock based on news data according to claim 1 is characterized in that:

In described step D, the computing method of keyword sequence correlativity sequence are as follows in real time:

Vector _Event=[e ₁e ₂E _M], wherein, e _MRepresent the relative coefficient between real-time news data and the keyword M.

5. according to the automatic match classifying method of each described stock of claim 1-4 based on news data, it is characterized in that: in described step D, the real-time keyword sequence correlativity sequence of real-time news data and the computing method that personal share keyword sequence correlativity series model mates automatically are as follows:

Wherein, || vector|| represents the Euclid norm of vector;

The correlativity sequence of described stock A is vector _A=[C _A1C _A2C _AM], then || vector _A|| account form be: $\left|\right|{\mathrm{vector}}_A\left|\right|=\sqrt{C_{A1}^2+C_{A2}^2+\·\·\·+C_{\mathrm{AM}}^2}.$

6. the automatic match classifying method of the stock based on news data according to claim 1 is characterized in that:

In described step C, described parameter optimization is keyword sequence to be optimized with existing ripe stock classification result by the comparison classification results.

7. the automatic match classifying of the stock based on news data system is characterized in that it comprises:

One local data base is used to store historical news data and real-time news data, and sets up corresponding personal share sequence number and keyword sequence within it;

One participle screening module is carried out participle, screening to each then historical news data, extracts wherein relevant with personal share keyword sequence, and real-time news data is carried out participle, screening, extracts real-time keyword sequence;

One statistical generic module, be used for adding up the frequency that each keyword of keyword sequence occurs, and calculate corresponding keyword relative coefficient, structure personal share keyword sequence correlativity series model, the line parameter optimization of going forward side by side, and the correlativity between the calculating personal share, and stock is classified in conjunction with cluster algorithm;

One automatic matching module, be used to calculate real-time keyword sequence correlativity sequence, automatically mate by real-time keyword sequence correlativity sequence and personal share keyword sequence correlativity series model, choose the highest personal share of correlativity as final best matching result with real-time news data.