CN110210645A - A kind of development trend data capture method, device and readable storage medium storing program for executing - Google Patents
A kind of development trend data capture method, device and readable storage medium storing program for executing Download PDFInfo
- Publication number
- CN110210645A CN110210645A CN201910319456.7A CN201910319456A CN110210645A CN 110210645 A CN110210645 A CN 110210645A CN 201910319456 A CN201910319456 A CN 201910319456A CN 110210645 A CN110210645 A CN 110210645A
- Authority
- CN
- China
- Prior art keywords
- data
- prediction
- business revenue
- factor
- historical
- 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
- 238000011161 development Methods 0.000 title claims abstract description 124
- 238000000034 method Methods 0.000 title claims abstract description 81
- 238000013481 data capture Methods 0.000 title claims abstract description 12
- 238000010801 machine learning Methods 0.000 claims abstract description 71
- 238000012549 training Methods 0.000 claims abstract description 31
- 238000005259 measurement Methods 0.000 claims description 67
- 230000006870 function Effects 0.000 claims description 50
- 238000012512 characterization method Methods 0.000 claims description 20
- 238000013075 data extraction Methods 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 11
- 230000000694 effects Effects 0.000 claims description 10
- 230000008901 benefit Effects 0.000 claims description 4
- 241000406668 Loxodonta cyclotis Species 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 13
- 235000013399 edible fruits Nutrition 0.000 description 5
- 241001269238 Data Species 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 2
- 238000003360 curve fit method Methods 0.000 description 1
- 238000013479 data entry Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- 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
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0203—Market surveys; Market polls
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention proposes a kind of development trend data capture method, device and readable storage medium storing program for executing, this method comprises: determining object type belonging to prediction object；Determine at least one factor corresponding with object type；At least two objects of object type will be belonged to as sample object, and obtain the historical development data of each sample object respectively；The corresponding first historical factors data of each factor are extracted from the historical development data of each sample object respectively；Correspond to the machine learning model of object type by each first historical factors data training extracted；The corresponding second historical factors data of each factor are extracted from the historical development data of prediction object；Each second historical factors scanning machine device learning model is obtained into the first prediction data；The development trend data for characterizing prediction object development trend are determined according to the first prediction data.This programme can reduce analysis personnel and carry out the cost that business revenue prediction is paid to listed company.
Description
Technical field
The present invention relates to technical field of data processing more particularly to a kind of development trend data capture methods, device and can
Read storage medium.
Background technique
According to the relevant regulations of company law, listed company is obligated to be predicted simultaneously externally to announce to its business revenue, therefore on
The analysis personnel of company, city need periodically to carry out business revenue prediction to the listed company where it.Under normal conditions, analysis personnel with
Season is to predict the business revenue of listed company in the period.
Currently, analysis personnel mainly by the following method predict the business revenue of listed company: according to institute, listed company
The industry of category, which is determined, has the multiple factors directly affected to listed company's business revenue, utilizes the real-time of each factor determined
Data carry out linear fit or fitting of a polynomial, and then according to the linear function or polynomial function fitted to listed company
Business revenue predicted.
For the method predicted at present listed company's business revenue, Linear Quasi is carried out using the real time data of each factor
Conjunction or fitting of a polynomial, therefore influence of the timeliness of data corresponding to each factor to prediction result is very big, in order to guarantee
The accuracy of business revenue prediction, analysis personnel need to collect the corresponding real time data of each factor in time, cause the analysis personnel to be
Listed company's business revenue predicts paid higher cost.
Summary of the invention
The present invention provides a kind of development trend data capture method, device and readable storage medium storing program for executing, main purpose and is
It is defeated by the historical development data that will predict to exclusively enjoy using the historical development data training machine learning model of multiple sample objects
Enter trained machine learning model, obtains the first prediction data that machine learning model is exported, and then according to the first prediction
Data predict the development trend data of object to determine.The development trend data capture method is applied to the business revenue of listed company
When prediction, real time data corresponding to listed company is collected in time without analyzing personnel, so as to reduce analysis personnel to upper
Company, city carries out the cost that business revenue prediction is paid.
In a first aspect, the embodiment of the invention provides a kind of development trend data capture methods, comprising:
Determine object type belonging to prediction object；
Determine at least one factor corresponding with the object type, wherein the different factors are corresponding with different
Data statistics rule；
At least two objects of the object type will be belonged to as sample object, and obtain each described sample respectively
The historical development data of object；
Each described factor corresponding first is extracted from the historical development data of sample object described in each respectively
Historical factors data；
Correspond to the machine learning of the object type by each first historical factors data training extracted
Model；
The corresponding second historical factors number of each described factor is extracted from the historical development data of the prediction object
According to；
Each second historical factors data are inputted into the machine learning model, it is defeated to obtain the machine learning model
The first prediction data out；
The development trend data for characterizing the prediction object development trend are determined according to first prediction data.
Optionally,
It is determined according to first prediction data for characterizing the development trend for predicting object development trend described
Before data, further comprise:
Utilize the historical development data fit polynomial function of at least two sample objects, wherein described in each
The historical development data of sample object are all satisfied the polynomial function；
The historical development data of the prediction object are inputted into the polynomial function, obtain the polynomial function output
The second prediction data；
The development trend number determined according to first prediction data for characterizing the prediction object development trend
According to, comprising:
The development trend of the prediction object is determined according to first prediction data and second prediction data
Data.
Optionally,
In the hair for determining the prediction object according to first prediction data and second prediction data
Before opening up trend data, further comprise:
Utilize the historical development data fit time series model of at least two sample objects, wherein each institute
State sample object historical development data change with time rule meet the time series models；
The historical development data of the prediction object are inputted into the time series models, obtain the time series models
The third prediction data of output；
The development that the prediction object is determined according to first prediction data and second prediction data
Trend data, comprising:
The prediction pair is determined according to first prediction data, second prediction data and the third prediction data
The development trend data of elephant.
Optionally,
The historical development data fit polynomial function using at least two sample objects, comprising:
Determine predetermined period that prediction of the development trend is carried out to the prediction object；
Each system is extracted from the historical development data of sample object described in each respectively according to described predetermined period
Period corresponding first historical development data are counted, wherein the measurement period and described predetermined period are opposite on time span
It answers；
According to the corresponding each first historical development data of each measurement period, following multinomial letter is fitted
Number, wherein each corresponding described first historical development data of each described sample object are all satisfied the multinomial letter
Number；
Wherein, the M characterizes second prediction data relative to current time；The kiCharacterization passes through machine learning
The weight coefficient fitted；X characterization is relative to upper one of the current time measurement period corresponding described the
One historical development data；The xiCharacterize the described measurement period of upper i+1 relative to the current time corresponding described the
One historical development data；The t+1 characterizes the number of the current time foregoing description measurement period.
Optionally,
The benefit utilizes the historical development data fit time series model of at least two sample objects, comprising:
Determine predetermined period that prediction of the development trend is carried out to the prediction object；
Each system is extracted from the historical development data of sample object described in each respectively according to described predetermined period
Period corresponding second historical development data are counted, wherein the measurement period and described predetermined period are opposite on time span
It answers；
According to the corresponding each second historical development data fit time series model of each measurement period,
In, the corresponding each second historical development data of each described sample object change with time rule when meeting described
Between series model；
The form of the time series models is as follows:
(ΔMt)2=K+k1(ΔMt-1)2-k2(ΔMt-2)2+εt-k3εt-1
Wherein, the Δ MtCharacterize the third prediction data and upper the one of the current time relative to current time
The difference of the corresponding second historical development data of a measurement period；The Δ Mt-1Characterize upper the one of the current time
Second statistics week before a corresponding second historical development data of the measurement period and the current time
The difference of the phase corresponding second historical development data；The Δ Mt-2Second system before characterizing the current time
It is corresponding with the third measurement period before the current time to count the period corresponding second historical development data
The difference of the second historical development data；The εtCharacterize the third prediction data relative to the current time；It is described
εt-1Characterize the corresponding second historical development data of upper one of the current time measurement period；It is the K, described
k1, the k2With the k3It is the weight coefficient fitted by machine learning.
Optionally, described according to the corresponding each second historical development data fit time of each measurement period
Series model, comprising:
Each second historical development data corresponding to each measurement period carry out second order difference, obtain opposite
The difference sequence answered；
According to the difference sequence, target equation corresponding with model is defined using tabulating method；
The target equation is carried out to solve the estimated result for obtaining the model；
It is detected based on fitting effect of the goodness of fit to the model；
After determining that the fitting effect of the model reaches preset target, the residual error of the model is examined
It surveys；
When the residual error for determining the model fluctuates in preset fluctuation range, the model is determined as described
Time series models.
Optionally, described true according to first prediction data, second prediction data and the third prediction data
The development trend data of the fixed prediction object, comprising:
First prediction data, second prediction data and the third prediction data are weighted, obtained
Obtain the development trend data of the prediction object.
Optionally, each first historical factors data training by extracting corresponds to the object type
Machine learning model, comprising:
For the factor described in each, the past at least two is obtained from the corresponding first historical factors data of the factor
At least one corresponding factor data of the factor of each year in year；
It is answered respectively with factor pair described in each using the corresponding factor data of each factor as sample training
Factor coefficient；
It is constructed as follows using each factor coefficient got for calculating the formula of first prediction data；
Wherein, the M ' characterization first prediction data；The n characterizes the number of the factor；Described in the m characterization
The number in the first historical factors data covered history year；The x(i, 1)It characterizes the prediction object the 1st year before this and corresponds to
The factor data of i-th of factor；The x(i, 2)It characterizes the prediction object the 2nd year before this and corresponds to i-th of factor
Factor data；The kiCharacterize the factor coefficient that current time corresponds to i-th of factor；The x(i, j)It characterizes described pre-
Survey the factor data that object corresponds to i-th of factor jth year before this；
Building includes the machine learning model of the formula.
Second aspect, the embodiment of the invention also provides a kind of development trend data acquisition facilities, comprising:
Classification identification module, factor identification module, data acquisition module, the first data extraction module, model training module,
Second data extraction module, model processing modules and data processing module；
The classification identification module, for determining object type belonging to prediction object；
The factor identification module, it is corresponding with the object type that the classification identification module is determined for determination
At least one factor, wherein the different factors are corresponding with different data statistics rules；
The data acquisition module, for the object type that the classification identification module is determined will to be belonged at least
Two objects obtain the historical development data of each sample object as sample object respectively；
First data extraction module, for respectively from sample object described in each and by the data acquisition mould
It is corresponding that each described factor that the factor identification module is determined is extracted in the historical development data that block is got
First historical factors data；
The model training module, each first history for being extracted by first data extraction module
Factor data training corresponds to the machine learning model of the object type；
Second data extraction module is known for extracting the factor from the historical development data of the prediction object
Other module determines the corresponding second historical factors data of each described factor；
The model processing modules, each second history for extracting second data extraction module because
Subdata inputs the machine learning model that the model training module trains, and obtains the machine learning model output
First prediction data；
The data processing module, first prediction data for being got according to the model processing modules determine
For characterizing the development trend data of the prediction object development trend.
The third aspect, it is described to deposit the embodiment of the invention also provides a kind of computer equipment, including memory and processor
Reservoir is stored with computer program, and the processor is realized any described in above-mentioned first aspect when executing the computer program
Development trend data capture method.
Fourth aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, are stored thereon with computer
Program, the computer program realize any development trend data acquisition in above-mentioned first aspect when being executed by processor
Method.
Development trend data capture method, device and computer equipment provided in an embodiment of the present invention and computer-readable
Storage medium after determining object type belonging to prediction object, determines one or more factors corresponding with object type, it
The historical development data for belonging at least two sample objects of the object type are obtained afterwards, and from the history of each sample object
The corresponding first historical factors data of each factor are extracted in development dataset, and are mentioned from the historical development data of prediction object
The corresponding second historical factors data of each factor are taken, correspond to prediction using the training of each first historical factors data later
Each second historical factors data are being inputted trained machine learning mould by the machine learning model of the affiliated object type of object
The first prediction data is obtained after type, and then the development trend data of prediction object can be determined according to the first prediction data.Thus
As it can be seen that using listed company as prediction object, using development trend data as when business revenue prediction data, utilization and listed company
Belong to the history business revenue data of multiple sample companies of same industry classification to predict the business revenue of listed company, to it is each because
The timeliness requirement of first historical factors data corresponding to son is lower, and it is corresponding to collect each factor in time without analysis personnel
Real time data carries out the cost that business revenue prediction is paid so as to reduce analysis personnel to listed company.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is the present invention
Some embodiments for those of ordinary skill in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 is a kind of flow chart of listed company's business revenue prediction technique provided by one embodiment of the present invention；
Fig. 2 is a kind of flow chart of Polynomial curve-fit method provided by one embodiment of the present invention；
Fig. 3 is a kind of flow chart of time series models approximating method provided by one embodiment of the present invention；
Fig. 4 is the flow chart of another listed company business revenue prediction technique provided by one embodiment of the present invention；
Fig. 5 is a kind of schematic diagram of listed company's business revenue prediction meanss provided by one embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments, based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.It answers
Work as understanding, the specific embodiments described herein are merely illustrative of the present invention, is not intended to limit the present invention.
Below using predict object as listed company, development trend data be listed company's business revenue prediction data for, to this
Development trend data capture method and device provided by inventive embodiments are described in detail.Specifically, with development trend number
Corresponding according to acquisition methods is listed company's business revenue prediction technique, corresponding with development trend data acquisition facility for listing
Company's business revenue prediction meanss.
As shown in Figure 1, one embodiment of the invention provides a kind of listed company's business revenue prediction technique, comprising:
Step 101: determination needs to carry out category of employment belonging to the listed company of business revenue prediction；
Step 102: determining at least one factor corresponding with category of employment, wherein the different factors are corresponding with different
Data statistics rule；
Step 103: at least two companies of category of employment will be belonged to as sample company, and obtain each sample respectively
The history business revenue data of company；
Step 104: extracting each factor corresponding first from the history business revenue data of each sample company respectively
Historical factors data；
Step 105: corresponding to the machine learning of category of employment by each first historical factors data training extracted
Model；
Step 106: the corresponding second historical factors number of each factor is extracted from the history business revenue data of listed company
According to；
Step 107: by each second historical factors scanning machine device learning model, obtaining machine learning model output
First business revenue prediction result；
Step 108: the prediction business revenue data of listed company are determined according to the first business revenue prediction result.
Listed company's business revenue prediction technique provided in an embodiment of the present invention determines the listed company for needing to carry out business revenue prediction
After affiliated category of employment, determines one or more factors corresponding with category of employment, obtain belong to category of employment later
The history business revenue data of at least two sample companies, and extract from the history business revenue data of each sample company each because
The corresponding first historical factors data of son, and each factor corresponding second is extracted from the history business revenue data of listed company
Historical factors data correspond to the machine of the affiliated category of employment of listed company using the training of each first historical factors data later
Learning model obtains the first business revenue prediction knot after each second historical factors data are inputted trained machine learning model
Fruit, and then the prediction business revenue data of listed company can be determined according to the first business revenue prediction result.It can be seen that using with it is upper
Company, city belongs to the history business revenue data of multiple sample companies of same industry classification to predict the business revenue of listed company,
It is lower to the timeliness requirement of the first historical factors data corresponding to each factor, each factor is collected in time without analysis personnel
Corresponding real time data carries out the cost that business revenue prediction is paid so as to reduce analysis personnel to listed company.
In embodiments of the present invention, determine the affiliated category of employment of listed company it is corresponding at least one because of the period of the day from 11 p.m. to 1 a.m, need
It is carried out according to the period that the affiliated category of employment of listed company carries out business revenue prediction.For example, the affiliated category of employment of listed company is usual
Business revenue data are counted according to season, then in the business revenue prediction to listed company, are usually predicted one under listed company
The business revenue data in season.Correspondingly, preceding quarter business revenue, preceding quarter total assets, same quarter last year business revenue and same quarter last year is total
Assets are determined as corresponding 4 factors of the affiliated category of employment of listed company.
For example, being determined by preceding quarter business revenue, preceding quarter total assets, same quarter last year business revenue and same quarter last year total assets
After 4 factors, 3000 companies for belonging to the affiliated category of employment of listed company are determined as sample company, and obtain each
10 years history business revenue data are gone over by sample company, and then from the history business revenue data got from obtaining each sample respectively
Company go over each season in 10 years each years season business revenue, season total assets is as the first historical factors data.Lead to later
240,000 (3000*10*4*2) the first historical factors data training machine learning model extracted is crossed, institute, listed company is obtained
Belong to the corresponding machine learning model of category of employment.The preceding quarter business revenue for the listed company that pending business revenue is predicted later, upper season
It spends total assets, same quarter last year business revenue and same quarter last year total assets and inputs machine learning model, it is defeated to obtain machine learning model
Listed company's the coming season prediction business revenue data out are as the first business revenue prediction result.
Optionally, on the basis of listed company's business revenue prediction technique shown in Fig. 1, due to each first history business revenue data
Reflect the business revenue situation of each sample company before this, thus can be determined according to each first history business revenue data and correspond to difference
The factor coefficient of the factor utilizes each factor coefficient determined to react each sample company business revenue variation tendency over the years,
And then can use each factor coefficient building machine learning model determined, by constructed machine learning model to listing
Second historical factors data of company are handled to predict the business revenue of listed company, and the first business revenue prediction knot is obtained
Fruit.The method of specific building machine learning model may comprise steps of:
S1: being directed to each factor, and the past at least two is obtained from the corresponding each first historical factors data of the factor
At least one corresponding factor data of the factor of each year in year；
S2: each factor data that will acquire is as sample training factor coefficient corresponding with each factor respectively；
S3: it is constructed as follows using each factor coefficient got for calculating the company of the first business revenue prediction result；
Wherein, the first business revenue of M ' characterization prediction result；The number of the n characterization factor；M characterizes the first historical factors data and is covered
The number in lid history year；x(i, 1)Characterize the factor data that listed company corresponds to i-th of factor for the 1st year before this；x(i, 2)Characterization
Listed company corresponds to the factor data of i-th of factor for the 2nd year before this；kiCharacterize current time correspond to i-th factor because
Subsystem number；x(i, j)Characterize the factor data that listed company corresponds to i-th of factor jth year before this；
S4: building includes the machine learning model of above-mentioned formula.
For example, 4 factors determined for listed company are respectively preceding quarter business revenue, preceding quarter total assets, last year in same season
Business revenue and same quarter last year total assets are spent, the first historical factors data got are the business revenue number that 3000 companies go over 10 years
According to being then directed to this Graph One factor of preceding quarter business revenue can determine that 10*3000*2 amounts to 60,000 factor datas, accordingly for upper
Season total assets, same quarter last year business revenue and same quarter last year total assets can determine 6 factor datas.Later by this 24
Ten thousand factor datas carry out machine learning as sample data, fit 4 factor systems for corresponding respectively to above-mentioned 4 factors
Number.Fit 4 factor coefficients are substituted into above-mentioned formulas later, and Shang Shi company corresponded to over the years before this above-mentioned 4 because
The history business revenue data of son substitute into above-mentioned formula, can calculate the first business revenue prediction that business revenue prediction is carried out to listed company
As a result.
In embodiments of the present invention, using each sample company history business revenue data fitting correspond to each factor because
Subsystem number, and then correspond to the affiliated category of employment machine learning model of listed company using the factor coefficient building fitted, it should
Machine learning model reflects the business revenue variation tendency of the affiliated category of employment of listed company, and then can use constructed machine
Learning model predicts the business revenue of listed company, due to reference to identical other companies of category of employment business revenue variation tendency and
The history business revenue situation of listed company to be predicted, so as to more accurately predict the business revenue of listed company.
Optionally, by training machine learning model, and will extracted from the history business revenue data of listed company
Two historical factors scanning machine device learning models obtain the first business revenue prediction result, can directly predict the first business revenue later
As a result as the prediction business revenue data of listed company, can be combined with the business revenue prediction result obtained by other prediction techniques come
Determine the prediction business revenue data of listed company.It is pre- for being determined by the business revenue prediction result obtained in conjunction with other prediction techniques
The method for surveying business revenue data, can specifically determine the prediction business revenue data of listed company by the following two kinds mode:
Mode one: by the first business revenue prediction result obtained by machine learning model and pass through what polynomial function obtained
Second business revenue prediction result combines, to determine the prediction business revenue data of listed company；
Mode two: it is obtained by the first business revenue prediction result obtained by machine learning model, by polynomial function
Second business revenue prediction result and by time series models obtain third business revenue prediction result combine, to determine listed company
Prediction business revenue data.
Determine the side of prediction business revenue data by combining multiple business revenue prediction results to above two separately below
Method is illustrated.
For mode one:
On the basis of listed company's business revenue prediction technique shown in Fig. 1, the is obtained by machine learning model in step 107
After one business revenue prediction result, and the prediction business revenue data of listed company are determined in step 108 according to the first business revenue prediction result
Before, the history business revenue data fit polynomial function that can use each sample company, so that each sample company is gone through
History business revenue data are all satisfied polynomial function made of fitting, and then the input of the history business revenue data of listed company is fitted
Polynomial function obtains the second business revenue prediction result of polynomial function output.Correspondingly, step 108 is pre- according to the first business revenue
It, can be according to the first business revenue prediction result and the second business revenue prediction result when survey result determines the prediction business revenue data of listed company
It determines the prediction business revenue data of listed company, can specifically calculate the first business revenue prediction result and the second business revenue prediction result
Weighted average, using calculated weighted average as the prediction business revenue data of listed company.
According to the history business revenue data fit polynomial function of each sample company, and then by the history business revenue of listed company
Data input polynomial function and obtain the second business revenue prediction result, pre- according to the first business revenue prediction result and the second business revenue later
Result is surveyed to determine the prediction business revenue data of listed company, due to combining two kinds of business revenues of machine learning model and polynomial function
Prediction technique predicts the business revenue of listed company, so as to improve the business revenue to listed company predicted it is accurate
Property.
In embodiments of the present invention, using the history business revenue data fit polynomial function of each sample company, such as Fig. 2 institute
Show, the process of fit polynomial function can be achieved by the steps of:
Step 201: determining predetermined period that business revenue prediction is carried out to listed company；
Step 202: being extracted from the history business revenue data of each sample company respectively according to the predetermined period determined
The corresponding first history business revenue data of each measurement period, wherein measurement period and predetermined period are opposite on time span
It answers；
Step 203: according to the corresponding each first history business revenue data of each measurement period, fitting following multinomial letter
Number, so that each corresponding first history business revenue data of each sample company are all satisfied the polynomial function；
Wherein, M characterizes the business revenue prediction result relative to current time；kiCharacterize the weight fitted by machine learning
Coefficient；X characterizes the corresponding first history business revenue data of a upper measurement period relative to current time；xiCharacterization is relative to working as
The corresponding first history business revenue data of the upper i+1 measurement period of preceding time；T+1 characterizes of measurement period before current time
Number.
For example, now needing to predict the business revenue of listed company's lower first quarter, then business revenue prediction is carried out to listed company
Predetermined period be season, correspondingly, the measurement period counted to the history business revenue data of various kinds our company is also season.
For each sample company, each season is obtained on the sample corporate history from the history business revenue data of the sample company
Business revenue is as the first history business revenue data.By intending the corresponding each first history business revenue data of each sample company
It closes, obtains the polynomial function that the corresponding each first history business revenue data of each sample company can be made to be all satisfied.
After fitting polynomial function, listed company is obtained from the history business revenue data of listed company corresponding to each
The data of measurement period, the data corresponding to each measurement period that will acquire later input polynomial function, obtain multinomial
Second business revenue prediction result of formula function output.For example, determining that the predetermined period for carrying out business revenue prediction to listed company is season
After degree, the business revenue data in listed company's each season in history are obtained from the history business revenue data of listed company, and then will
The business revenue data in the listed company got each season in history input the polynomial function that fits, obtain correspond to it is upper
Second business revenue prediction result of company, city.
After the polynomial function by fitting obtains the second business revenue prediction result, according to pre- for the first business revenue in advance
The weighted value for surveying result and the setting of the second business revenue prediction result, carries out the first business revenue prediction result and the second business revenue prediction result
Ranking operation, using operation result as the prediction business revenue data of listed company.Specifically, it can will be tied for the prediction of the first business revenue
The weighted value of fruit and the second business revenue prediction result is set as 0.5, i.e., by the first business revenue prediction result and the second business revenue prediction knot
Prediction business revenue data of the weighted average of fruit as listed company.
For mode two:
On the basis of aforesaid way one provides prediction business revenue data determination method, obtained by machine learning model
First business revenue prediction result and by polynomial function obtain the second business revenue prediction result after, can use each sample company
History business revenue data fit time series model, the rule so that the history business revenue data of each sample company change with time
Meet the time series models fitted, and then the history business revenue data of listed company are inputted to the time series models fitted
Afterwards, the third business revenue prediction result of time series models output is obtained.It correspondingly, can be according to the first business revenue prediction result,
Two business revenue prediction results and third business revenue prediction result determine the prediction business revenue data of listed company.
Machine learning model, polynomial function and time series mould are obtained according to the history business revenue data of each sample company
Type, and then the history business revenue data of listed company are inputted into machine learning model, polynomial function and time series models respectively,
The first business revenue prediction result, the second business revenue prediction result and the third business revenue prediction result for corresponding to listed company are obtained respectively,
The pre- of listed company is determined according to the first business revenue prediction result, the second business revenue prediction result and third business revenue prediction result later
Survey business revenue data.Due to combining three kinds of machine learning model, polynomial function and time series models business revenue prediction techniques pair
The business revenue of listed company predicted, the accuracy that the business revenue so as to further increase to listed company is predicted.
In embodiments of the present invention, it can use the history business revenue data fit time series model of each sample company,
As shown in figure 3, the process of fit time series model can be achieved by the steps of:
Step 301: determining predetermined period that business revenue prediction is carried out to listed company；
Step 302: each system is extracted from the history business revenue data of each sample company respectively according to predetermined period
Period corresponding second history business revenue data are counted, wherein measurement period is corresponding on time span with predetermined period；
Step 303: according to the corresponding each second history business revenue data fit time series model of each measurement period, making
The corresponding each second history business revenue data of each the sample company rule that changes with time meets the time series models；
The form of time series models is as follows:
(ΔMt)2=K+k1(ΔMt-1)2-k2(ΔMt-2)2+εt-k3εt-1
Wherein, Δ MtCharacterization is relative to the business revenue prediction result of current time and a upper measurement period pair for current time
The difference for the second history business revenue data answered；ΔMt-1Characterize the corresponding second history business revenue of a upper measurement period of current time
The difference of the corresponding second history business revenue data of second measurement period before data and current time；ΔMt-2When characterizing current
Between before the corresponding second history business revenue data of second measurement period and current time before third measurement period pair
The difference for the second history business revenue data answered；εtCharacterize the business revenue prediction result relative to current time；εt-1Characterize current time
The corresponding second history business revenue data of a upper measurement period；K,k1、k2And k3It is the weight system fitted by machine learning
Number.
For example, now needing to predict the business revenue of listed company's lower first quarter, then business revenue prediction is carried out to listed company
Predetermined period be season, correspondingly, the measurement period counted to the history business revenue data of each sample company is also season
Degree.For each sample company, each season is obtained on the sample corporate history from the history business revenue data of the sample company
The business revenue of degree is as the second history business revenue data.By intending the corresponding second history business revenue data of each sample company
It closes, obtaining can make each second history business revenue data of each sample company change with time the regular time being all satisfied
Series model.
After fitting time series models, listed company is obtained from the history business revenue data of listed company and is corresponded to respectively
The data of a measurement period, the data entry time series model corresponding to each measurement period that will acquire later obtain
The third business revenue prediction result of time series models output.For example, determining the prediction week for carrying out business revenue prediction to listed company
After phase is season, the business revenue data in listed company's each season in history are obtained from the history business revenue data of listed company,
And then the business revenue data in the listed company that will acquire each season in history input the time series models fitted, obtain
Third business revenue prediction result corresponding to listed company.
It should be noted that during practical business is realized, the first history business revenue number of user's fit polynomial function
It can be identical data according to the second history business revenue data for fit time series model.
Optionally, on the basis of the method for listed company's prediction business revenue data is determined provided by aforesaid way two,
It, can be pre- to the first business revenue after obtaining the first business revenue prediction result, the second business revenue prediction result and third business revenue prediction result
It surveys result, the second business revenue prediction result and third business revenue prediction result to be weighted, using the result of ranking operation as upper
The prediction business revenue data of company, city.
It specifically, can be according to actual needs to the first business revenue prediction result, the second business revenue prediction result and third business revenue
Corresponding weighting coefficient is arranged in prediction result, to control each business revenue prediction result to the weight of final prediction business revenue data.
For example, 0.4 can be set by the corresponding weighting coefficient of the first business revenue prediction result, by the second business revenue prediction result it is corresponding plus
Weight coefficient is set as 0.3, sets 0.3 for the corresponding weighting coefficient of third business revenue prediction result.
Respectively the first business revenue prediction result, the second business revenue prediction result and the setting of third business revenue prediction result are corresponding
Weighting coefficient seeks the first business revenue prediction result, the second business revenue prediction result and third according to weighting coefficient set in advance
It receives prediction result to be weighted, using the result of ranking operation as the prediction business revenue data of listed company.Added by setting
On the one hand power system can balance three business revenue prediction results to the influence degree of operation result, weigh three kinds of business revenue prediction sides
The pros and cons of method so that final prediction business revenue data obtained are more accurate, on the other hand allow user according to demand oneself
Row adjusts weighting coefficient, to meet the individual demand of different user, promotes being applicable in for listed company's business revenue prediction technique
Property.
Optionally, on the basis of the method for fit time series model shown in Fig. 3, step 303 is according to each statistics week
Phase corresponding each second history business revenue data fit time series model, can specifically be fitted ARIMA time series models, and
The fit procedure of ARIMA time series models can be accomplished in that
Each second history business revenue data corresponding to each measurement period carry out second order difference, obtain corresponding difference
Sequence；
According to difference sequence, target equation corresponding with model is defined using tabulating method；
Target equation is carried out to solve the estimated result for obtaining model；
It is detected based on fitting effect of the goodness of fit to model；
After the fitting effect for determining model reaches preset target, the residual error of model is detected；
When the residual error for determining model fluctuates in preset fluctuation range, model is determined as time series mould
Type.
In embodiments of the present invention, it during fit time series model, is primarily based on the goodness of fit and model is intended
It closes effect to be detected, the linear residual error of model be detected after fitting effect reaches preset target, in model
Linear residual error be located in preset fluctuation range after, model is determined as time series models.By intending model
It closes effect and linear residual error is detected, it is after fitting effect and linear residual error are all satisfied preset condition, model is true
It is set to time series models, guarantees the accuracy of generated time series models, and then can guarantees to pass through time series models
The accuracy of obtained third business revenue prediction result.
Below in the above described manner for the two prediction business revenue data capture methods provided, on provided in an embodiment of the present invention
Company, city business revenue prediction technique is described in further detail, as shown in figure 4, this method may comprise steps of:
Step 401: determination needs to carry out category of employment belonging to the listed company of business revenue prediction.
In embodiments of the present invention, when needing the business revenue to a listed company to predict, it is necessary first to which determining should
Category of employment belonging to listed company.
For example, the business revenue to listed company A next season is now needed to predict, then listed company A institute is determined first
The category of employment A of category.
Step 402: obtaining the history business revenue data for belonging to multiple sample companies of determined category of employment.
In embodiments of the present invention, the category of employment belonging to the listed company for determining to need to carry out business revenue prediction it
Afterwards, it is subordinated in each company of determined category of employment and randomly chooses at least two companies as sample company, later
Obtain the history business revenue data of each sample company.
It selects 3000 companies as sample company for example, being subordinated in the company of category of employment A, obtains respectively later
10 years history business revenue data are gone over by each sample company in 3000 sample companies.Wherein, if the establishment of sample company
Time less than 10 years, then obtains all history of the sample company and stubbornly insists data.
Step 403: passing through the history business revenue data training machine learning model of each sample company.
In embodiments of the present invention, the category of employment belonging to the listed company for determining to need to carry out business revenue prediction it
Afterwards, at least one factor for corresponding to the sector classification is determined.It is directed to each sample company later, from going through for the sample company
The corresponding first historical factors data of each factor are extracted in history business revenue data.Each first history extracted is utilized later
Machine learning model of the factor data training relative to determined category of employment.Wherein, the machine learning model trained can
To be predicted according to history business revenue data following business revenue data.Specifically, machine learning model may include as follows
Formula:
Wherein, the first business revenue of M ' characterization prediction result；The number of the n characterization factor；M characterizes the first historical factors data and is covered
The number in lid history year；x(i, 1)Characterize the factor data that listed company corresponds to i-th of factor for the 1st year before this；x(i, 2)Characterization
Listed company corresponds to the factor data of i-th of factor for the 2nd year before this；kiCharacterize current time correspond to i-th factor because
Subsystem number；x(i, j)Characterize the factor data that listed company corresponds to i-th of factor jth year before this.
For example, preceding quarter business revenue, preceding quarter total assets, same quarter last year business revenue and same quarter last year total assets are determined as
Corresponding 4 factors of category of employment A.Later for sample company, each of 3000 sample companies, from the sample company
The season business revenue and season that each each season in year in 10 years is gone over by the sample company is extracted in past 10 years history business revenue data
Total assets is spent as the first historical factors data, can be extracted 3000*10*4*2 in this way and be amounted to 240,000 the first historical factors
Data utilize this 240,000 the first historical factors data training machine learning model A later.
Step 404: being handled using history business revenue data of the machine learning model to listed company, obtain the first business revenue
Prediction result.
In embodiments of the present invention, each factor corresponding second is extracted from the history business revenue data of listed company to go through
The each second historical factors data extracted are inputted the machine learning model trained, by machine by history factor data later
Learning model predicts the business revenue of listed company according to each second historical factors data, obtains machine learning model output
The first business revenue prediction result.
For example, the history business revenue data that listed company A goes over 10 years are obtained, later from the history business revenue data got
Extract listed company A go over 10 years in each each season in year season business revenue and season total assets as the second historical factors number
According to, can extract in this way 10*4*2 amount to 80 the second historical factors data.Later by this 80 the second historical factors data
Machine learning model A is inputted, the first business revenue prediction result of machine learning model A output is obtained.It specifically can be by the 80 of acquisition
A second historical factors data substitute into above-mentioned formula, calculate the first business revenue prediction result.
Step 405: passing through the history business revenue data fit polynomial function of each sample company.
In embodiments of the present invention, according to the predetermined period for carrying out business revenue prediction to listed company, from each sample public affairs
The first history business revenue data of each measurement period are extracted in the history business revenue data of department, utilize each extracted later
One history business revenue data fit polynomial function, so that the corresponding each first history business revenue data of each sample company are full
The polynomial function fitted enough.Wherein, measurement period is corresponding on time span with predetermined period.
The form of the polynomial function fitted is as follows:
Wherein, M characterizes the business revenue prediction result relative to current time；kiCharacterize the weight fitted by machine learning
Coefficient；X characterizes the corresponding first history business revenue data of a upper measurement period relative to current time；xiCharacterization is relative to working as
The corresponding first history business revenue data of the upper i+1 measurement period of preceding time；T+1 characterizes of measurement period before current time
Number.
For example, since the predetermined period for carrying out business revenue prediction to listed company A is season, in 3000 sample companies
Each sample company, from the history business revenue data of the sample company extract in the past the 10 years sample company each years it is each
The business revenue data in season can extract 3000*10*4 and amount to 120,000 the first history business revenues as the first history business revenue data
Data.This 120,000 the first history business revenue data fit polynomial functions are utilized later, so that each sample company is corresponding
Each first history business revenue data meet the polynomial function.
Step 406: being handled using history business revenue data of the polynomial function to listed company, it is pre- to obtain the second business revenue
Survey result.
In embodiments of the present invention, it is extracted according to history business revenue data of the measurement period to listed company, acquisition pair
Should be in the history business revenue data of each measurement period, the history business revenue number corresponding to each measurement period that will extract later
According to input polynomial function, the business revenue of listed company is carried out in advance according to each history business revenue data of input by polynomial function
It surveys, obtains the second business revenue prediction result of polynomial function output.
For example, the history business revenue data that listed company A goes over 10 years are obtained, later from the history business revenue data got
The season business revenue that listed company A goes over each each season in year in 10 years is extracted, it is 40 total that 10*4 can be extracted in this way
Season business revenue.Later this business revenue is inputted into following polynomial function A in 40 season, obtains the second of polynomial function A output
Business revenue prediction result；
Wherein, in polynomial function A, M is characterized relative to the second business revenue prediction result；kiCharacterization passes through machine learning
The weight coefficient fitted；The season business revenue in an x characterization listed company A upper season；xiCharacterize i+1 season on listed company A
The season business revenue of degree.
Step 407: passing through the history business revenue data fit time series model of each sample company.
In embodiments of the present invention, according to the predetermined period for carrying out business revenue prediction to listed company, from each sample public affairs
The second history business revenue data of each measurement period are extracted in the history business revenue data of department, utilize each extracted later
Two history business revenue data fit time series models so that the corresponding each second history business revenue data of each sample company with
The changing rule of time meets the time series models.Wherein, measurement period is corresponding on time span with predetermined period.
The form of fitted time series models is as follows:
(ΔMt)2=K+k1(ΔMt-1)2-k2(ΔMt-2)2+εt-k3εt-1
Wherein, Δ MtCharacterization is relative to the business revenue prediction result of current time and a upper measurement period pair for current time
The difference for the second history business revenue data answered；ΔMt-1Characterize the corresponding second history business revenue of a upper measurement period of current time
The difference of the corresponding second history business revenue data of second measurement period before data and current time；ΔMt-2When characterizing current
Between before the corresponding second history business revenue data of second measurement period and current time before third measurement period pair
The difference for the second history business revenue data answered；εtCharacterize the business revenue prediction result relative to current time；εt-1Characterize current time
The corresponding second history business revenue data of a upper measurement period；K,k1、k2And k3It is the weight system fitted by machine learning
Number.
For example, since the predetermined period for carrying out business revenue prediction to listed company A is season, in 3000 sample companies
Each sample company, from the history business revenue data of the sample company extract in the past the 10 years sample company each years it is each
The business revenue data in season can extract 3000*10*4 and amount to 120,000 the second history business revenues as the second history business revenue data
Data.This 120,000 the second history business revenue data fit time series models are utilized later, so that each sample company is corresponding
Each second history business revenue data meet more time series models.
Step 408: being handled using history business revenue data of the time series models to listed company, obtain third business revenue
Prediction result.
In embodiments of the present invention, it is extracted according to history business revenue data of the measurement period to listed company, acquisition pair
Should be in the history business revenue data of each measurement period, the history business revenue number corresponding to each measurement period that will extract later
According to input time series model, by time series models according to each history business revenue data of input to the business revenue of listed company into
Row prediction, obtains the third business revenue prediction result of polynomial function output.
For example, the history business revenue data that listed company A goes over 10 years are obtained, later from the history business revenue data got
The season business revenue that listed company A goes over each each season in year in 10 years is extracted, it is 40 total that 10*4 can be extracted in this way
Season business revenue.Later this business revenue is inputted into following time series models A in 40 season, obtains time series models A output
Third business revenue prediction result；
(ΔM40)2=K+k1(ΔM39)2-k2(ΔM38)2+ε40-k3ε39
Wherein, Δ M40Characterize third business revenue prediction result and a listed company A upper season season business revenue difference；Δ
M39Characterize a listed company A upper season season business revenue and listed company A before this second season season business revenue difference；Δ
M38Characterize listed company A before this second season season business revenue and listed company A before this third season season business revenue it
Difference；ε40Characterize the third business revenue prediction result of listed company A；ε39Characterized the season business revenue in a listed company A upper season；K,
k1、k2And k3It is the weight coefficient fitted by machine learning.
Step 409: being determined according to the first business revenue prediction result, the second business revenue prediction result and third business revenue prediction result
The prediction business revenue data of company, city.
In embodiments of the present invention, the first business revenue prediction result, the second business revenue prediction result and third business revenue are being got
After prediction result, fortune is weighted to the first business revenue prediction result, the second business revenue prediction result and third business revenue prediction result
It calculates, using the result of ranking operation as the prediction business revenue data of listed company.
Specifically, add to the first business revenue prediction result, the second business revenue prediction result and third business revenue prediction result
It can in advance be respectively that the first business revenue prediction result, the second business revenue prediction result and third business revenue prediction result are set when weighing operation
Fixed corresponding weighting coefficient.First business revenue prediction result, the second business revenue prediction result and third business revenue prediction result corresponding 3
A weighting coefficient can be equal, and 3 weighting coefficients are 1/3 at this time.Further, it is also possible to be the first business revenue prediction result, second
Business revenue prediction result and third business revenue prediction result set different weighting coefficients, specifically can use and obtain the prediction of the first business revenue
As a result, the method for the second business revenue prediction result and third business revenue prediction result predicts this previous history year of listed company
Business revenue, and then weighting coefficient is determined according to the true business revenue situation in the prediction result of three estimated history year.For example, utilizing
The business revenue that the method for obtaining the first business revenue prediction result predicts last quarter is X1, utilizes the side for obtaining the second business revenue prediction result
The business revenue that method predicts last quarter is X2, is using the business revenue that the method for obtaining third business revenue prediction result predicts last quarter
X3, and it is X4 that the business revenue of last quarter is practical, and then can be determined according to the absolute value of difference between X1, X2 and X3 and X4 pair
Should in the weighting coefficient of the first business revenue prediction result, the second business revenue prediction result and third business revenue prediction result, the difference
The more big then corresponding weighting coefficient of absolute value is smaller.
For example, calculating the prediction business revenue data of listed company by following formula:(first
The+the second business revenue of business revenue prediction result prediction result+third business revenue prediction result).
As shown in figure 5, one embodiment of the invention provides a kind of listed company's business revenue prediction meanss, comprising: classification is known
Other module 501, factor identification module 502, data acquisition module 503, the first data extraction module 504, model training module
505, the second data extraction module 506, model processing modules 507 and data processing module 508；
Classification identification module 501 needs to carry out category of employment belonging to the listed company of business revenue prediction for determination；
Factor identification module 502, it is corresponding with the category of employment that classification identification module 501 is determined at least for determination
One factor, wherein the different factors are corresponding with different data statistics rules；
Data acquisition module 503, at least two for that will belong to the category of employment that classification identification module 501 is determined are public
Department is used as sample company, and obtains the history business revenue data of each sample company respectively；
First data extraction module 504, for company and being obtained respectively from each sample by data acquisition module 503
To history business revenue data in the corresponding first historical factors number of each factor determined of extraction factor identification module 502
According to；
Model training module 505, each first historical factors number for being extracted by the first data extraction module 504
Correspond to the machine learning model of category of employment according to training；
Second data extraction module 506, for the extraction factor identification module 502 from the history business revenue data of listed company
Determine the corresponding second historical factors data of each factor；
Model processing modules 507, each second historical factors data for extracting the second data extraction module 506
The machine learning model that input model training module 505 trains, the first business revenue for obtaining machine learning model output predict knot
Fruit；
Data processing module 508, the first business revenue prediction result for being got according to model processing modules 507 determine
The prediction business revenue data of company, city.
In embodiments of the present invention, classification identification module 501 can be used for executing the step 101 in above method embodiment,
Factor identification module 502 can be used for executing the step 102 in above method embodiment, and data acquisition module 503 can be used for executing
Step 103 in above method embodiment, the first data, which extract model 504, can be used for executing the step in above method embodiment
104, model training module 505 can be used for executing the step 105 in above method embodiment, and the second data extraction module 506 can
For executing the step 106 in above method embodiment, model processing modules 507 can be used for executing in above method embodiment
Step 107, data processing module 508 can be used for executing the step 108 in above method embodiment.
It should be noted that in information exchange, implementation procedure between modules included by present apparatus embodiment etc.
Hold, since based on the same inventive concept, particular content may refer to chatting in above method embodiment with above method embodiment
It states, details are not described herein again.In addition, present apparatus embodiment can also include other modules, for executing above method embodiment
In each step.
The embodiment of the invention also provides a kind of computer equipments, include memory and processor, store on memory
There is computer program, above-mentioned each embodiment institute may be implemented when processor executes the computer program stored on memory
Listed company's business revenue prediction technique of offer.
The embodiment of the invention also provides a kind of computer readable storage medium, stored on the computer readable storage medium
There is computer program, the listing public affairs that above-mentioned each embodiment may be implemented and provide are provided in the computer storage that it is stored
Take charge of business revenue prediction technique.
In conclusion listed company's business revenue prediction technique, device and computer that each embodiment of the present invention provides are set
Standby and computer readable storage medium, determine need to carry out category of employment belonging to the listed company of business revenue prediction after, determine with
The corresponding one or more factors of category of employment, obtain the history battalion for belonging at least two sample companies of category of employment later
Data are received, and extract the corresponding first historical factors number of each factor from the history business revenue data of each sample company
According to, and the corresponding second historical factors data of each factor are extracted from the history business revenue data of listed company, it utilizes later
The training of each first historical factors data corresponds to the machine learning model of the affiliated category of employment of listed company, by each second
Historical factors data obtain the first business revenue prediction result after inputting trained machine learning model, and then can be according to the first battalion
Prediction result is received to determine the prediction business revenue data of listed company.It can be seen that belonging to same industry class using with listed company
The history business revenue data of other multiple sample companies predict the business revenue of listed company, go through to corresponding to each factor first
The timeliness requirement of history factor data is lower, collects the corresponding real time data of each factor in time without analysis personnel, thus
Analysis personnel can be reduced, the cost that business revenue prediction is paid is carried out to listed company.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, device, article or the method that include a series of elements not only include those elements, and
And further include other elements that are not explicitly listed, or further include for this process, device, article or method institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including being somebody's turn to do
There is also other identical elements in the process, device of element, article or method.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.Pass through above embodiment party
The description of formula, it is required general that those skilled in the art can be understood that above-described embodiment method can add by software
The mode of hardware platform is realized, naturally it is also possible to which by hardware, but in many cases, the former is more preferably embodiment.It is based on
Such understanding, substantially the part that contributes to existing technology can be with software product in other words for technical solution of the present invention
Form embody, which is stored in a storage medium (such as ROM/RAM, magnetic disk, light as described above
Disk) in, including some instructions use is so that a terminal device (can be mobile phone, computer, server or the network equipment
Deng) execute method described in each embodiment of the present invention.
The above is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of development trend data capture method characterized by comprising
Determine object type belonging to prediction object；
Determine at least one factor corresponding with the object type, wherein the different factors are corresponding with different data
Statistical rules；
At least two objects of the object type will be belonged to as sample object, and obtain each described sample object respectively
Historical development data；
Corresponding first history of each described factor is extracted from the historical development data of sample object described in each respectively
Factor data；
Correspond to the machine learning model of the object type by each first historical factors data training extracted；
The corresponding second historical factors data of each described factor are extracted from the historical development data of the prediction object；
Each second historical factors data are inputted into the machine learning model, obtain the machine learning model output
First prediction data；
The development trend data for characterizing the prediction object development trend are determined according to first prediction data.
2. the method according to claim 1, wherein
It is determined according to first prediction data for characterizing the development trend data for predicting object development trend described
Before, further comprise:
Utilize the historical development data fit polynomial function of at least two sample objects, wherein each described sample
The historical development data of object are all satisfied the polynomial function；
The historical development data of the prediction object are inputted into the polynomial function, obtain the of the polynomial function output
Two prediction data；
The development trend data determined according to first prediction data for characterizing the prediction object development trend, packet
It includes:
The development trend data of the prediction object are determined according to first prediction data and second prediction data.
3. according to the method described in claim 2, it is characterized in that,
Determine that the development of the prediction object becomes according to first prediction data and second prediction data described
Before gesture data, further comprise:
Utilize the historical development data fit time series model of at least two sample objects, wherein each described sample
The historical development data of this object change with time rule meet the time series models；
The historical development data of the prediction object are inputted into the time series models, obtain the time series models output
Third prediction data；
The development trend that the prediction object is determined according to first prediction data and second prediction data
Data, comprising:
The prediction object is determined according to first prediction data, second prediction data and the third prediction data
The development trend data.
4. according to the method in claim 2 or 3, which is characterized in that the going through using at least two sample objects
History development dataset fit polynomial function, comprising:
Determine predetermined period that prediction of the development trend is carried out to the prediction object；
Extract each statistics week from the historical development data of sample object described in each respectively according to described predetermined period
Phase corresponding first historical development data, wherein the measurement period and described predetermined period are corresponding on time span；
According to the corresponding each first historical development data of each measurement period, following polynomial function is fitted,
Wherein, each corresponding described first historical development data of each described sample object are all satisfied the polynomial function；
Wherein, the M characterizes second prediction data relative to current time；The kiCharacterization is fitted by machine learning
Weight coefficient out；The x characterization is gone through relative to upper one measurement period corresponding described first of the current time
History development dataset；The xiIt characterizes and is gone through relative to the upper i+1 measurement periods corresponding described first of the current time
History development dataset；The t+1 characterizes the number of the current time foregoing description measurement period.
5. according to the method described in claim 3, it is characterized in that, the benefit utilizes the history of at least two sample objects
Development dataset fit time series model, comprising:
Determine predetermined period that prediction of the development trend is carried out to the prediction object；
Extract each statistics week from the historical development data of sample object described in each respectively according to described predetermined period
Phase corresponding second historical development data, wherein the measurement period and described predetermined period are corresponding on time span；
According to the corresponding each second historical development data fit time series model of each measurement period, wherein
The corresponding each second historical development data of each the described sample object rule that changes with time meets the time
Series model；
The form of the time series models is as follows:
(ΔMt)2=K+k1(ΔMt-1)2-k2(ΔMt-2)2+εt-k3εt-1
Wherein, the Δ MtCharacterize a upper institute for the third prediction data and the current time relative to current time
State the difference of the corresponding second historical development data of measurement period；The Δ Mt-1Characterize a upper institute for the current time
State second measurement period pair before the corresponding second historical development data of measurement period and the current time
The difference for the second historical development data answered；The Δ Mt-2Second statistics week before characterizing the current time
The phase corresponding second historical development data are corresponding with the third measurement period before the current time described
The difference of second historical development data；The εtCharacterize the third prediction data relative to the current time；The εt-1Table
Levy the corresponding second historical development data of upper one of the current time measurement period；The K, the k1, institute
State k2With the k3It is the weight coefficient fitted by machine learning.
6. according to the method described in claim 5, it is characterized in that, described according to the corresponding each institute of each measurement period
State the second historical development data fit time series model, comprising:
Each second historical development data corresponding to each measurement period carry out second order difference, obtain corresponding
Difference sequence；
According to the difference sequence, target equation corresponding with model is defined using tabulating method；
The target equation is carried out to solve the estimated result for obtaining the model；
It is detected based on fitting effect of the goodness of fit to the model；
After determining that the fitting effect of the model reaches preset target, the residual error of the model is detected；
When the residual error for determining the model fluctuates in preset fluctuation range, the model is determined as the time
Series model.
7. according to method described in claim 3,5 or 6, which is characterized in that
It is described that the prediction pair is determined according to first prediction data, second prediction data and the third prediction data
The development trend data of elephant, comprising:
First prediction data, second prediction data and the third prediction data are weighted, institute is obtained
State the development trend data of prediction object；
And/or
The machine learning for corresponding to the object type by each first historical factors data training extracted
Model, comprising:
For the factor described in each, obtained in past at least 2 years from the corresponding first historical factors data of the factor
At least one corresponding factor data of the factor of each year；
Using the corresponding factor data of each factor as sample training respectively it is corresponding with the factor described in each because
Subsystem number；
It is constructed as follows using each factor coefficient got for calculating the formula of first prediction data；
Wherein, the M ' characterization first prediction data；The n characterizes the number of the factor；The m characterization described first
The number in historical factors data covered history year；The x(i, 1)It characterizes the prediction object the 1st year before this and corresponds to i-th
The factor data of the factor；The x(i, 2)Characterize the factor that the prediction corresponds to i-th of factor for object the 2nd year before this
Data；The kiCharacterize the factor coefficient that current time corresponds to i-th of factor；The x(i, i)Characterize the prediction object
Jth year corresponds to the factor data of i-th of factor before this；
Building includes the machine learning model of the formula.
8. a kind of development trend data acquisition facility characterized by comprising classification identification module, factor identification module, data
It obtains at module, the first data extraction module, model training module, the second data extraction module, model processing modules and data
Manage module；
The classification identification module, for determining object type belonging to prediction object；
The factor identification module, it is corresponding extremely with the classification identification module object type determined for determination
Few factor, wherein the different factors are corresponding with different data statistics rules；
The data acquisition module, for at least two of the object type that the classification identification module is determined will to be belonged to
Object obtains the historical development data of each sample object as sample object respectively；
First data extraction module, for respectively from sample object described in each and obtained by the data acquisition module
Each described factor corresponding first that the factor identification module is determined is extracted in the historical development data got
Historical factors data；
The model training module, each first historical factors for being extracted by first data extraction module
Data training corresponds to the machine learning model of the object type；
Second data extraction module, for extracting the factor identification mould from the historical development data of the prediction object
Block determines the corresponding second historical factors data of each described factor；
The model processing modules, each second historical factors number for extracting second data extraction module
The machine learning model trained according to the model training module is inputted obtains the first of the machine learning model output
Prediction data；
The data processing module, first prediction data determination for being got according to the model processing modules are used for
Characterize the development trend data of the prediction object development trend.
9. a kind of computer equipment, including memory and processor, the memory are stored with computer program, feature exists
In the step of processor realizes any one of claims 1 to 7 the method when executing the computer program.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of any one of claims 1 to 7 the method is realized when being executed by processor.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910319456.7A CN110210645A (en) | 2019-04-19 | 2019-04-19 | A kind of development trend data capture method, device and readable storage medium storing program for executing |
| PCT/CN2019/103060 WO2020211245A1 (en) | 2019-04-19 | 2019-08-28 | Development trend data acquisition method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910319456.7A CN110210645A (en) | 2019-04-19 | 2019-04-19 | A kind of development trend data capture method, device and readable storage medium storing program for executing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110210645A true CN110210645A (en) | 2019-09-06 |
Family
ID=67786096
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910319456.7A Pending CN110210645A (en) | 2019-04-19 | 2019-04-19 | A kind of development trend data capture method, device and readable storage medium storing program for executing |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110210645A (en) |
| WO (1) | WO2020211245A1 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112668772A (en) * | 2020-12-24 | 2021-04-16 | 润电能源科学技术有限公司 | State development trend prediction method, device, equipment and storage medium |
| WO2022141883A1 (en) * | 2020-12-31 | 2022-07-07 | 平安科技（深圳）有限公司 | Enterprise revenue trend prediction method and apparatus, and computer device and storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102968670B (en) * | 2012-10-23 | 2016-08-17 | 北京京东世纪贸易有限公司 | The method and apparatus of prediction data |
| CN107194489A (en) * | 2016-03-14 | 2017-09-22 | 阿里巴巴集团控股有限公司 | Data predication method and device |
| US10078337B1 (en) * | 2017-07-14 | 2018-09-18 | Uber Technologies, Inc. | Generation of trip estimates using real-time data and historical data |
| CN108550047A (en) * | 2018-03-20 | 2018-09-18 | 阿里巴巴集团控股有限公司 | The prediction technique and device of trading volume |
-
2019
- 2019-04-19 CN CN201910319456.7A patent/CN110210645A/en active Pending
- 2019-08-28 WO PCT/CN2019/103060 patent/WO2020211245A1/en active Application Filing
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112668772A (en) * | 2020-12-24 | 2021-04-16 | 润电能源科学技术有限公司 | State development trend prediction method, device, equipment and storage medium |
| CN112668772B (en) * | 2020-12-24 | 2024-03-12 | 润电能源科学技术有限公司 | State development trend prediction method, device, equipment and storage medium |
| WO2022141883A1 (en) * | 2020-12-31 | 2022-07-07 | 平安科技（深圳）有限公司 | Enterprise revenue trend prediction method and apparatus, and computer device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020211245A1 (en) | 2020-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Curry et al. | Manufacturing systems modeling and analysis | |
| CN103593479B (en) | Application program menu recommendation method and device | |
| CN108491951A (en) | A kind of prediction technique, device and electronic equipment for taking out distribution time | |
| Xiang et al. | An expanded robust optimisation approach for the berth allocation problem considering uncertain operation time | |
| CN110309434B (en) | Track data processing method and device and related equipment | |
| CN107153874A (en) | Water quality prediction method and system | |
| CN115643285A (en) | Smart city parking lot recommendation method, internet of things system, device and storage medium | |
| CN110288350A (en) | User's Value Prediction Methods, device, equipment and storage medium | |
| CN111476441A (en) | Load prediction method for electric vehicle charging equipment and related device | |
| CN108446944A (en) | A kind of the determination method, apparatus and electronic equipment in resident city | |
| CN110210645A (en) | A kind of development trend data capture method, device and readable storage medium storing program for executing | |
| CN110458325A (en) | A kind of traffic zone passenger flow forecasting and device in short-term | |
| CN109615414A (en) | House property predictor method, device and storage medium | |
| Morton et al. | A Bayesian stochastic programming approach to an employee scheduling problem | |
| CN114037140A (en) | Prediction model training method, prediction model training device, prediction model data prediction method, prediction model data prediction device, prediction model data prediction equipment and storage medium | |
| CN113837383B (en) | Model training method and device, electronic equipment and storage medium | |
| Du et al. | Analysis of university fitness center data uncovers interesting patterns, enables prediction | |
| US20170236087A1 (en) | Method and System for Recommendation Engine Optimization | |
| Li et al. | Discrete event simulation analysis of a reservation-based, one-way car-sharing system | |
| CN112488752A (en) | Sales prediction method, sales prediction device, electronic equipment and storage medium | |
| CN115311001A (en) | Method and system for predicting user change tendency based on multiple voting algorithm | |
| CN115203545A (en) | Health maintenance service intelligent matching method and system based on deep learning and knowledge graph | |
| CN114417884A (en) | Intelligent display device and method for information technology consultation service | |
| CN114418450A (en) | Data processing method and device | |
| Gupta et al. | Risks in supply chain management and its mitigation |
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: 20190906 |
|
| RJ01 | Rejection of invention patent application after publication |