CN106780021A - A kind of stock based on local and global gram statistics recommends method - Google Patents
A kind of stock based on local and global gram statistics recommends method Download PDFInfo
- Publication number
- CN106780021A CN106780021A CN201611045969.6A CN201611045969A CN106780021A CN 106780021 A CN106780021 A CN 106780021A CN 201611045969 A CN201611045969 A CN 201611045969A CN 106780021 A CN106780021 A CN 106780021A
- Authority
- CN
- China
- Prior art keywords
- stock
- amount
- global
- local
- statistics
- 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
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
- 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)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Complex Calculations (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Recommend method the invention discloses a kind of stock based on local and global gram statistics.Method regards the amount of increase and amount of decrease quantized value of stock short time sequence as the symbol string in natural language, to predicting then similar to the prediction to the following symbol in symbol string for the next day of trade amount of increase and amount of decrease of stock；Then with reference to the reality of stock, using the thought of n-gram, the stock certificate data of historical data and the overall situation to stock itself carries out the rule-statistical of stock grammer, is then utilized respectively local grammer and global grammer calculates the amount of increase and amount of decrease score of stock next day of trade.Comprehensive score finally is calculated to two kinds of results, the recommendation of stock is carried out after sequence.Method is in addition to the daily recommendation that can be used for stock, it may also be used for the daily real-time recommendation of stock.
Description
Technical field
The present invention relates to stock certificate data digging technology field, more particularly, to a kind of based on local and overall situation gram statistics
Stock recommends method.
Background technology
Stock market is the market of a risk and benefit paragenesis, the warp of the modeling and forecasting research of stock market to China
Ji development and finance construction have great importance.
The purpose of stock technical Analysis is following trend of prediction markets price, and the means used to reach this purpose are
The analysis past and present market behavior in stock market.The market behavior includes：(1) change of the height and price of price；(2) send out
The adjoint exchange hand of raw these changes；(3) these change elapsed times are completed.The theoretical foundation of technical Analysis is mainly
Three big hypothesis:The market behavior includes all information；Price fluctuates along trend and keeps trend;History repeat itself.
In the Prediction of Stock Index field that data are extremely abundant, the method for data mining has obtained more and more extensive research and has answered
With.The research that stock certificate data is excavated is concentrated mainly on 4 aspects：Similar sequences matching, Prediction of Stock Price, Mining Exchange Rules
And temporal mode finds.Conventional method has neutral net, evolution algorithm, fuzzy logic, rough set, SVMs etc..
The content of the invention
Recommend method the invention discloses a kind of stock based on local and global gram statistics.When method is stock segment
Between the amount of increase and amount of decrease quantized value of sequence regard symbol string in natural language as, to the prediction of the next day of trade amount of increase and amount of decrease of stock then
Similar to the prediction to the following symbol in symbol string.
The reality of the inventive method combination stock, using the thought of n-gram, to the historical data and the overall situation of stock itself
Stock certificate data carry out the rule-statistical of stock grammer, be then utilized respectively local grammer and global grammer calculate the next friendship of stock
The amount of increase and amount of decrease score of Yi.Comprehensive score finally is calculated to two kinds of results, the recommendation of stock is carried out after sequence.Method is except available
In outside the daily recommendation of stock, it may also be used for the daily real-time recommendation of stock.
Assuming that stock list is S, S=[S1, S2,…,Si,…,Sm], m is the quantity of stock in stock pond, such as in China
The quantity of city's stock or the quantity of listed stock of the U.S..The step of the inventive method, is as follows：
（1）Ups and downs amplitude to stock quantifies；
（2）Local polynary gram statistics are carried out based on stock itself historical data；
（3）Global polynary gram statistics are carried out based on all stock historical datas；
（4）Obtain the recent amount of increase and amount of decrease data of stock to be predicted and quantify；
（5）Calculate the local syntax prediction score of stock to be predicted；
（6）Calculate the global syntax prediction score of stock to be predicted；
（7）Calculate every comprehensive score of stock and carry out sort recommendations.
Wherein, the ups and downs amplitude to stock of step (1) quantifies, specially：For every stock, when obtaining certain
Between put since（Such as on January 1st, 2005）Amount of increase and amount of decrease data, then ups and downs amplitude is quantified, i.e., ups and downs amplitude is carried out
The operation for rounding up, is transformed to [- 10,10] interval integer value；Every stock in last stock pond is all transformed to include
The array of advance versus decline width integer value.
Wherein, step (2) based on stock itself historical data carries out local polynary gram statistics, specially：For every
Stock, it is assumed that current stock is Si, i=1 ..., m, to the stock, the historical data of itself carries out k metagrammar statistics, and k takes 1
To 5.One k dimension group is set first, and the length of each dimension is 21, i.e., often dimension there are 21 grooves.Then S is traveled throughiAdvance versus decline
Width integer value array, takes two adjacent values every time, and the number of times to the common appearance of the two values carries out cumulative statistics, and is put into
In the corresponding groove of k dimension groups.After k metagrammars have been counted, with reference to the result of k-1 metagrammars, design conditions probability, for k dimensions
Every a line in group, calculates the conditional probability that the numerical value of last dimension occurs, wherein a metagrammar disregards calculation conditional probability.Bar
The molecule of part probability is the statistical number of the integer value combination that k dimension groups are often gone, and denominator is that k dimensions group often remove finally by row integer value
The corresponding statistical number in k-1 metagrammars of the combination of.If k-1 metagrammars do not have corresponding integer value when calculating denominator
The calculating of the conditional probability is then skipped in combination.Last every stock has the k metagrammar statisticses of oneself.
Wherein, step (3) based on all stock historical datas carries out global polynary gram statistics, specially：For institute
The historical data for having stock carries out k metagrammar statistics.Statistic processes similar step（2）, the difference is that being directed to all stock history
Data are counted, and last all stocks only have one group of statistics of k metagrammars.
Wherein, the recent amount of increase and amount of decrease data of the acquisition stock to be predicted of step (4) and quantify, specially obtain stock to be predicted
Then this 4 days amount of increase and amount of decrease data is quantified by the amount of increase and amount of decrease data of ticket nearly 4 day of trade, i.e., carry out four to ups and downs amplitude
House five enters operation, forms integer value, is designated as D4, D3, D2, D1, and the approximate integral ups and downs of stock nearly four day of trade are represented respectively
Width.And the approximate amount of increase and amount of decrease of next day of trade that will be predicted is designated as D0.
Wherein, the local syntax prediction score of the calculating stock to be predicted of step (5),；Specially：With D4, D3, D2, D1
The preceding four dimensions of the local five metagrammar arrays of search matching, obtain 3 maximum numerical value APercent of conditional probabilityk, and its
The numerical value AD0 of corresponding 5th dimensionk；With D3, first three dimension of the local four metagrammar arrays of D2, D1 search matching obtains bar
3 numerical value BPercent of part maximum probabilityk, and its corresponding fourth dimension numerical value BD0k；With D2, D1 search matching local three
The first two dimension of metagrammar array, obtains 3 maximum numerical value CPercent of conditional probabilityk, and its corresponding third dimension
Numerical value CD0k；With the previous dimension of the D1 local two-dimensional grammar arrays of search matching, 3 maximum numerical value of conditional probability are obtained
DPercentk, and its corresponding second numerical value DD0 for tieing upk；Summation finally is weighted to all results, as local grammer
Score is predicted, specific publicity is：
（A）
；
（B）。
Wherein, the global syntax prediction score of the calculating stock to be predicted of step (6), specially：With D4, D3, D2, D1 is searched
Rope matches the preceding four dimensions of global five metagrammars array, obtains 3 maximum numerical value APercent of conditional probabilityk, and its it is right
The numerical value AD0 of the 5th dimension answeredk；With D3, first three dimension of D2, D1 the search global four metagrammars array of matching obtains condition
3 numerical value BPercent of maximum probabilityk, and its corresponding fourth dimension numerical value BD0k；With D2, the global ternary of D1 search matchings
The first two dimension of grammer array, obtains 3 maximum numerical value CPercent of conditional probabilityk, and its corresponding third dimension number
Value CD0k；With the previous dimension of the D1 global two-dimensional grammar arrays of search matching, 3 maximum numerical value of conditional probability are obtained
DPercentk, and its corresponding second numerical value DD0 for tieing upk；Summation finally is weighted to all results, as global grammer
Score is predicted, specific publicity is：
（A）
；
（B）。
Wherein, every comprehensive score of stock of calculating of step (7) and sort recommendations are carried out, specially：For every stock
Ticket, calculates the comprehensive score of stock, and specific publicity is：
WhereinIt is weight coefficient, it is considered to the characteristics of stock itself, value here.Comprehensive score to all stocks is entered
Row sequence from big to small, h before being taken out after sequence, these stocks are the stock of recommendation.The cycle of recommendation is to push away daily
Recommend, therefore the term of validity recommended is only one day, is only recommended to do stock next day of trade.The gained comprehensive score of recommendation stock can
As one kind reference of the approximate amount of increase in the next day of the stock.
Brief description of the drawings
Fig. 1 is the flow chart that stock of the present invention based on local and global gram statistics recommends method.
Fig. 2 is the recommendation stock list on a certain date based on the inventive method output.On November 10th, 1
The stock of recommendation, tertial score may be considered the prediction to next approximate amount of increase of transaction, while being provided in the 4th row
The turnover rate of same day stock is used as reference.
Specific embodiment
Below in conjunction with the accompanying drawings and example, the present invention is described in detail.
((n-gram grammar) is built upon a kind of probabilistic grammar on Markov model to n-gram.It is by right
The statistics of n symbol probability of occurrence simultaneously infers the structural relation of sentence in the symbol string of natural language.As n=2,
Referred to as two-dimensional grammar, bi-gram（During n=2）Referred to as single order Markov Chain；When n=3, referred to as three metagrammars, ternary
Grammatical model is referred to as second order Markov Chain.
The inventive method regards the amount of increase and amount of decrease quantized value of stock short time sequence as the symbol string in natural language, and
To the prediction of the next day of trade amount of increase and amount of decrease of stock similar to the prediction to the following symbol in symbol string.
The reality of the inventive method combination stock, using the thought of n-gram, to the historical data and the overall situation of stock itself
Stock certificate data carry out stock syntax rule statistics, be then utilized respectively local grammer and global grammer calculate the next transaction of stock
The amount of increase and amount of decrease score of day.Comprehensive score finally is calculated to two kinds of results, the recommendation of stock is carried out after sequence.The inventive method is pushed away
It is daily recommendation to recommend mode.
Assuming that stock list is S, S=[S1, S2,…,Si,…,Sm], m is the quantity of stock in stock pond, such as in China
The quantity of city's stock or the quantity of listed stock of the U.S..Stock recommendation process based on local and global gram statistics is specific such as
Under.
First, the ups and downs amplitude to stock quantifies.
For every stock, since obtaining certain time point（Such as on January 1st, 2005）Amount of increase and amount of decrease data, then to rising
Drop range value is quantified, i.e., the operation for being rounded up to ups and downs amplitude, is transformed to [- 10,10] interval integer value；Finally
Every stock in stock pond is all transformed to the array comprising advance versus decline width integer value.
2nd, local polynary gram statistics are carried out based on stock itself historical data.
For every stock, it is assumed that current stock is Si, i=1 ..., m, then the polynary gram statistics process of the stock is such as
Under.
2.1 local unitary gram statistics.
Because stock has amount of increase and amount of decrease to limit, the integer value after quantization has the interval integer in 21 kinds of situations, i.e. [- 10,10], because
This sets an array for having 21 grooves, travels through SiThe amount of increase and amount of decrease integer value array of stock, occurrence is gone out according to amount of increase and amount of decrease integer value
Number carries out cumulative statistics, and is put into corresponding groove.Finally, every stock has the unitary gram statistics array of oneself.
2.2 local two-dimensional grammar statistics.
One two-dimensional array of 21X21 is set, S is traveled throughiAdvance versus decline width integer value array, takes adjacent two every time
Value, the number of times to the common appearance of the two values carries out cumulative statistics, and is put into the corresponding groove of two-dimensional array.Every stock has
The two-dimensional grammar statistics array of oneself.
After two-dimensional grammar has been counted, with reference to the result of a metagrammar, design conditions probability.For every stock two-dimensional grammar
Every a line in array, it is assumed that common two integer values for occurring are AB, then conditional probability P (B)=P (AB)/P that consequent B occurs
(A) number of times that, i.e. AB occurs jointly accounts for the ratio of the independent occurrence numbers of A.The number of times that A individually occurs is united in a metagrammar
Obtained in meter.And the number of times that AB occurs jointly is obtained in the statistics of two-dimensional grammar.If former piece A did not occurred, skip
The calculating of conditional probability.
2.3 local ternary gram statistics.
One three-dimensional array of 21X21 X21 is set, S is traveled throughiAdvance versus decline width integer value array, takes adjacent three every time
Individual value, the number of times to the common appearance of these three values carries out cumulative statistics, and is put into the corresponding groove of three-dimensional array.Every stock
There is the ternary gram statistics array of oneself.
After ternary gram statistics are complete, with reference to the result of two-dimensional grammar, design conditions probability.For every metagrammar of stock three
Every a line in array, it is assumed that common three integer values for occurring are ABC, then conditional probability P (C)=P that consequent C occurs
(ABC) number of times that/P (AB), i.e. ABC occur jointly accounts for the ratio of AB occurrence numbers.The number of times that AB occurs is in two-dimensional grammar
Obtained in statistics.And the number of times that ABC occurs jointly is obtained in the statistics of three metagrammars.If former piece AB did not occurred,
The calculating of skip condition probability.
2.4 local quaternary gram statistics.
One four-dimensional array of 21X21 X21 X21 is set, S is traveled throughiAdvance versus decline width integer value array, takes adjacent every time
Four values, the statistics that adds up is carried out to the number of times of the common appearance of this four values, and be put into the corresponding groove of four-dimensional array.Every
Stock has the quaternary gram statistics array of oneself.
After quaternary gram statistics are complete, with reference to the result of three metagrammars, design conditions probability.For every metagrammar of stock four
Every a line in array, it is assumed that common four integer values for occurring are ABCD, then conditional probability P (D)=P that consequent D occurs
(ABCD) number of times that/P (ABC), i.e. ABCD occur jointly accounts for the ratio of ABC occurrence numbers.The number of times that ABC occurs is three
Obtained in metagrammar statistics.And the number of times that ABCD occurs jointly is obtained in the statistics of four metagrammars.If former piece ABC does not go out
Now cross, then the calculating of skip condition probability.
2.5 local five metagrammars statistics.
One five dimension group of 21X21 X21 X21 X21 is set, S is traveled throughiAdvance versus decline width integer value array, takes every time
Five adjacent values, the number of times to the common appearance of this five values carries out cumulative statistics, and is put into the corresponding groove of five dimension groups.
Every stock has the five metagrammars statistics array of oneself.
After five metagrammars have been counted, with reference to the result of four metagrammars, design conditions probability.For every metagrammar of stock five
Every a line in array, it is assumed that common five integer values for occurring are ABCDE, then conditional probability P (E)=P that consequent E occurs
(ABCDE) number of times that/P (ABCD), i.e. ABCDE occur jointly accounts for the ratio of ABCD occurrence numbers.The number of times that ABCD occurs is
Obtained in quaternary gram statistics.And the number of times that ABCDE occurs jointly is obtained in the statistics of five metagrammars.If former piece
ABCD did not occurred, then the calculating of skip condition probability.
3rd, global polynary gram statistics are carried out based on all stock historical datas.
For all stocks, the statistic processes of global many metagrammars is as follows.
3.1 global unitary gram statistics.
One array for there are 21 grooves is set, the amount of increase and amount of decrease integer value array of all stocks is traveled through, according to amount of increase and amount of decrease integer value
Occurrence number carry out cumulative statistics, and be put into corresponding groove.Finally, a unitary gram statistics array for the overall situation is formed.
3.2 global two-dimensional grammar statistics.
One two-dimensional array of 21X21 is set, all advance versus decline width integer value arrays are traveled through, adjacent two are taken every time
Value, the number of times to the common appearance of the two values carries out cumulative statistics, and is put into the corresponding groove of two-dimensional array.Finally, formed
One two-dimensional grammar statistics array of the overall situation.
After two-dimensional grammar has been counted, with reference to the result of a metagrammar, design conditions probability.To in global two-dimensional grammar array
Every a line, it is assumed that common two integer values occurring are AB, then conditional probability P (B)=P (AB)/P (A) that consequent B occurs,
That is the number of times that AB occurs jointly accounts for the ratio of the independent occurrence numbers of A.The number of times that A individually occurs is in global unitary gram statistics
In obtain.And the number of times that AB occurs jointly is obtained in the statistics of global two-dimensional grammar.If former piece A did not occurred, jump
Cross the calculating of conditional probability.
3.3 global ternary gram statistics.
One three-dimensional array of 21X21 X21 is set, all advance versus decline width integer value arrays are traveled through, is taken every time adjacent
Three values, the number of times to the common appearance of these three values carries out cumulative statistics, and is put into the corresponding groove of three-dimensional array.Finally,
Form a ternary gram statistics array for the overall situation.
After ternary gram statistics are complete, with reference to the result of two-dimensional grammar, design conditions probability.To in global three metagrammars array
Every a line, it is assumed that common three integer values for occurring are ABC, then conditional probability P (C)=P (ABC)/P that consequent C occurs
(AB) number of times that, i.e. ABC occurs jointly accounts for the ratio of AB occurrence numbers.The number of times that AB occurs is united in global two-dimensional grammar
Obtained in meter.And the number of times that ABC occurs jointly is obtained in the statistics of global three metagrammar.If former piece AB did not occurred,
The then calculating of skip condition probability.
3.4 global quaternary gram statistics.
One four-dimensional array of 21X21 X21 X21 is set, all advance versus decline width integer value arrays are traveled through, phase is taken every time
Four adjacent values, the number of times to the common appearance of this four values carries out cumulative statistics, and is put into the corresponding groove of four-dimensional array.Most
Afterwards, a quaternary gram statistics array for the overall situation is formed.
After quaternary gram statistics are complete, with reference to the result of three metagrammars, design conditions probability.To in global four metagrammars array
Every a line, it is assumed that common four integer values for occurring are ABCD, then conditional probability P (D)=P (ABCD)/P that consequent D occurs
(ABC) number of times that, i.e. ABCD occurs jointly accounts for the ratio of ABC occurrence numbers.The number of times that ABC occurs is in global ternary language
Obtained in method statistics.And the number of times that ABCD occurs jointly is obtained in the statistics of global four metagrammar.If former piece ABC does not go out
Now cross, then the calculating of skip condition probability.
3.5 global five metagrammars statistics.
One five dimension group of 21X21 X21 X21 X21 is set, all advance versus decline width integer value arrays is traveled through, every time
Five adjacent values are taken, the number of times to the common appearance of this five values carries out cumulative statistics, and is put into the corresponding groove of five dimension groups
In.Finally, a five metagrammars statistics array for the overall situation is formed.
After five metagrammars have been counted, with reference to the result of four metagrammars, design conditions probability.To in global five metagrammars array
Every a line, it is assumed that common five integer values for occurring are ABCDE, then conditional probability P (E)=P (ABCDE)/P that consequent E occurs
(ABCD) number of times that, i.e. ABCDE occurs jointly accounts for the ratio of ABCD occurrence numbers.The number of times that ABCD occurs is global four
Obtained in metagrammar statistics.And the number of times that ABCDE occurs jointly is obtained in the statistics of global five metagrammar.If former piece
ABCD did not occurred, then the calculating of skip condition probability.
4th, the recent amount of increase and amount of decrease data of stock are obtained and is quantified.
Obtain the amount of increase and amount of decrease data of stock nearly 4 day of trade, then this 4 days amount of increase and amount of decrease data is quantified, i.e., it is right
Ups and downs amplitude carries out the operation that rounds up, and forms integer value, is designated as D4, D3, D2, D1, and stock nearly four day of trade is represented respectively
Approximate integral amount of increase and amount of decrease.And the approximate amount of increase and amount of decrease of next day of trade that will be predicted is designated as D0.
5th, the prediction score of local grammer is calculated.
To every recent quantization amount of increase and amount of decrease of stock, the local gram statistics array of stock itself is searched for, and calculate score.
Detailed process is as follows.
5.1 obtain conditional probability most with D4, the preceding four dimensions of the local five metagrammar arrays of D3, D2, D1 search matching
3 big numerical value APercentk, and its corresponding 5th numerical value AD0 for tieing upk, k=1,2,3.Simultaneously weight 4 is distributed to the part.
5.2 obtain conditional probability maximum with D3, first three dimension of the local four metagrammar arrays of D2, D1 search matching
3 numerical value BPercentk, and its corresponding fourth dimension numerical value BD0k, k=1,2,3.Simultaneously weight 3 is distributed to the part.
5.3 obtain 3 of conditional probability maximum with D2, the first two dimension of the local three metagrammar arrays of D1 search matchings
Numerical value CPercentk, and its corresponding third dimension numerical value CD0k, k=1,2,3.Simultaneously weight 2 is distributed to the part.
5.4, with the previous dimension of the D1 local two-dimensional grammar arrays of search matching, obtain 3 maximum numbers of conditional probability
Value DPercentk, and its corresponding second numerical value DD0 for tieing upk, k=1,2,3.Simultaneously weight 1 is distributed to the part.
5.5 calculate local syntax prediction score
（1）
；
（2）。
In calculating process, if the conditional probability of denominator part and be 0, the contribution score value of the partial fraction is skipped.
6th, the prediction score of global grammer is calculated.
To every recent quantization amount of increase and amount of decrease of stock, the global gram statistics array of search stock, and calculate score.Specifically
Process is as follows.
6.1 obtain conditional probability most with D4, the preceding four dimensions of D3, D2, D1 the search global five metagrammars array of matching
3 big numerical value APercentk, and its corresponding 5th numerical value AD0 for tieing upk, k=1,2,3.Simultaneously weight 4 is distributed to the part.
6.2 obtain the 3 of conditional probability maximum with D3, first three dimension of D2, D1 the search global four metagrammars array of matching
Individual numerical value BPercentk, and its corresponding fourth dimension numerical value BD0k, k=1,2,3.Simultaneously weight 3 is distributed to the part.
6.3 obtain 3 of conditional probability maximum with D2, the first two dimension of the D1 search global three metagrammars arrays of matching
Numerical value CPercentk, and its corresponding third dimension numerical value CD0k, k=1,2,3.Simultaneously weight 2 is distributed to the part.
6.4, with the previous dimension of the D1 global two-dimensional grammar arrays of search matching, obtain 3 maximum numbers of conditional probability
Value DPercentk, and its corresponding second numerical value DD0 for tieing upk, k=1,2,3.Simultaneously weight 1 is distributed to the part.
6.5 calculate global syntax prediction score
（1）
；
（2）。
In calculating process, if the conditional probability of denominator part and be 0, the contribution score value of the partial fraction is skipped.
7th, calculate comprehensive score and carry out sort recommendations.
7.1, for every stock, calculate the comprehensive score of stock.Specially：
WhereinIt is weight coefficient, it is considered to the characteristics of stock itself, value here。
The 7.2 pairs of comprehensive scores of all stocks carry out sequence from big to small, h, these stocks before being taken out after sequence
The stock as recommended.The term of validity of recommendation is only one day, is only recommended to do stock next day of trade.Recommend the gained of stock comprehensive
Closing score can be as one kind reference of the approximate amount of increase in the next day of the stock.
In sum, method is recommended the invention discloses a kind of stock based on local and global gram statistics.Method pair
The stock certificate data of the historical data of stock itself and the overall situation carries out stock syntax rule statistics, be then utilized respectively local grammer and
Global grammer calculates the possibility amount of increase and amount of decrease score of stock next day of trade.Comprehensive score finally is calculated to two kinds of results.Method is removed
Can be used for outside the daily recommendation of stock, it may also be used for stock daily real-time recommendation.When carrying out real-time recommendation, it is only necessary to will be used
Day line number according to change into minute line number according to.
The inventive method is similarly applied to security class has the data of time series feature, such as fund, futures.Cause
This, although disclosing specific embodiments and the drawings of the invention for the purpose of illustration, its object is to help understand in of the invention
Hold and implement according to this, but it will be appreciated by those skilled in the art that：The essence of claim of the invention and appended is not being departed from
In god and scope, various replacements, to change and modifications all be impossible.Therefore, the present invention should not be limited to most preferred embodiment and
Accompanying drawing disclosure of that.Presently disclosed embodiment should be understood illustrative rather than it be claimed in all respects
Scope limitation.
Claims (5)
1. a kind of stock based on local and global gram statistics recommends method, it is characterised in that methods described includes following step
Suddenly：
（1）Ups and downs amplitude to stock quantifies；
（2）Local polynary gram statistics are carried out based on stock itself historical data；
（3）Global polynary gram statistics are carried out based on all stock historical datas；
（4）Obtain the recent amount of increase and amount of decrease data of stock to be predicted and quantify；
（5）Calculate the local syntax prediction score of stock to be predicted；
（6）Calculate the global syntax prediction score of stock to be predicted；
（7）Calculate every comprehensive score of stock and carry out sort recommendations.
2. a kind of stock based on local and global gram statistics according to claim 1 recommends method, it is characterised in that
When the statistics of many metagrammars is carried out to stock, during the amount of increase and amount of decrease quantized value of stock short time sequence is regarded as natural language
Symbol string, to the next day of trade amount of increase and amount of decrease of stock prediction then similar to the prediction to the following symbol in symbol string.
3. a kind of stock based on local and global gram statistics according to claim 1 recommends method, it is characterised in that
The stock certificate data of historical data and the overall situation to stock itself has carried out the rule-statistical of stock grammer, and this mode is taken into account simultaneously
The similitude of local and global stock fragment combination.
4. a kind of stock based on local and global gram statistics according to claim 1 recommends method, it is characterised in that
Prediction based on local and global grammer to stock next day of trade possible amount of increase and amount of decrease, includes under different k metagrammars statistics
The prediction of corresponding various possible situations, finally carries out synthesis calculating timesharing to the classic predictive of various situations again.
5. a kind of stock based on local and global gram statistics according to claim 1 recommends method, it is characterised in that
The prediction score of the prediction score of local grammer, global grammer, and last comprehensive score calculation so that can be near
Seemingly using the score as the stock approximate amount of increase of next day.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611045969.6A CN106780021A (en) | 2016-11-25 | 2016-11-25 | A kind of stock based on local and global gram statistics recommends method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611045969.6A CN106780021A (en) | 2016-11-25 | 2016-11-25 | A kind of stock based on local and global gram statistics recommends method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106780021A true CN106780021A (en) | 2017-05-31 |
Family
ID=58974138
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611045969.6A Pending CN106780021A (en) | 2016-11-25 | 2016-11-25 | A kind of stock based on local and global gram statistics recommends method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106780021A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108022133A (en) * | 2017-12-22 | 2018-05-11 | 洪志令 | A kind of Stock Evaluation method based on the similar income of itself history |
-
2016
- 2016-11-25 CN CN201611045969.6A patent/CN106780021A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108022133A (en) * | 2017-12-22 | 2018-05-11 | 洪志令 | A kind of Stock Evaluation method based on the similar income of itself history |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mehtab et al. | Stock price prediction using CNN and LSTM-based deep learning models | |
| Wang et al. | DeepTrader: a deep reinforcement learning approach for risk-return balanced portfolio management with market conditions Embedding | |
| Das et al. | A decoder-only foundation model for time-series forecasting | |
| Sadaei et al. | Combining ARFIMA models and fuzzy time series for the forecast of long memory time series | |
| CN105512727A (en) | Markov-based personal path prediction method | |
| CN106503853A (en) | A kind of foreign exchange transaction forecast model based on multiple scale convolutional neural networks | |
| CN110619540A (en) | Click stream estimation method of neural network | |
| CN115145901A (en) | Multi-scale-based time series prediction method and system | |
| CN113360848A (en) | Time sequence data prediction method and device | |
| CN115408505A (en) | Dialogue recommendation algorithm based on two-channel hypergraph interest modeling | |
| Tao et al. | Modeling and forecasting multifractal volatility established upon the heterogeneous market hypothesis | |
| CN115982567A (en) | Refrigerating system load prediction method based on sequence-to-sequence model | |
| CN115062832A (en) | Waste household appliance recovery amount prediction method based on multi-time scale attention network | |
| CN106780021A (en) | A kind of stock based on local and global gram statistics recommends method | |
| Sun et al. | GraphSAGE with deep reinforcement learning for financial portfolio optimization | |
| CN106776757A (en) | User completes the indicating means and device of Net silver operation | |
| CN113672871A (en) | High-proportion missing data filling method and related device | |
| Ronaghi et al. | ND-SMPF: A noisy deep neural network fusion framework for stock price movement prediction | |
| CN109919688B (en) | Electronic cigarette product line planning method considering market factors | |
| CN105678430A (en) | Improved user recommendation method based on neighbor project slope one algorithm | |
| US20230334360A1 (en) | Model-independent feature selection | |
| CN106530085A (en) | Stock prediction method based on similar n-gram statistics | |
| Vancsura et al. | Analysis of the performance of predictive models during Covid-19 and the Russian-Ukrainian war | |
| Fayyad et al. | Application of Recurrent Neural Network Bi-Long Short-Term Memory, Gated Recurrent Unit and Bi-Gated Recurrent Unit for Forecasting Rupiah Against Dollar (USD) Exchange Rate | |
| Pavithya et al. | Quantitative analysis and sentiment analysis for stock price forecast: the case of Colombo Stock Exchange |
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: 20170531 |
|
| RJ01 | Rejection of invention patent application after publication |