CN105869017A - Method and system for predicting ticket prices - Google Patents
Method and system for predicting ticket prices Download PDFInfo
- Publication number
- CN105869017A CN105869017A CN201610187700.5A CN201610187700A CN105869017A CN 105869017 A CN105869017 A CN 105869017A CN 201610187700 A CN201610187700 A CN 201610187700A CN 105869017 A CN105869017 A CN 105869017A
- Authority
- CN
- China
- Prior art keywords
- fare
- date
- model
- forecast
- pricing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0283—Price estimation or determination
- G06Q30/0284—Time or distance, e.g. usage of parking meters or taximeters
Abstract
The invention provides a method and a system for predicting ticket prices. The method comprises the following steps: acquiring all historical data of an object flight; extracting n time sequence ticket price data from the historical data so as to constitute n training samples; dividing the training samples into first training samples and second training samples; establishing basic predicting models which respectively correspond with the training samples, and training respective basic predicting model so as to determine respective parameter of the model; establishing a ticket price predicting model, introducing each second training sample to a corresponding basic predicting model for prediction so as to obtain n basic predicting values, and training the ticket price predicting model so as to determine respective parameter of the model; extracting n time sequence ticket price data so as to constitute n new training samples; introducing each new training sample to a corresponding basic predicting model for prediction so as to obtain n basic predicting values; introducing the acquired n basic predicting values to the ticket price predicting model for prediction so as to obtain a ticket price prediction value. The method can accurately predict ticket prices.
Description
Technical field
The present invention relates to a kind of fare-pricing estimate method and system, particularly to a kind of ticket based on machine learning
Valency Forecasting Methodology and system.
Background technology
Along with the gradually development of aircraft industry, airline's great majority use complicated pricing strategy to improve it
Income, thus cause the admission fee of aircraft to present high uncertainty, and due to price institute foundation
Factors is not the most public domain, and ordinary consumer is often difficult to hold the alteration trend of price.Air ticket
Price expectation, is the air ticket predicting some the given flight tool at following one or more time points
Body price.Predict machine profile exactly, no matter budget is saved for consumer, or be that air ticket is sold
Business holds market trends great forward facilitation.
Machine profile prediction is a typical recurrence learning problem, and the admission fee as output valve is a company
The variable of ideotype.In current research with put into practice field, it is directed to theory and the side of machine profile prediction
Method is less, and relevant research great majority concentrate in the optimisation strategy to air ticket purchase opportunity, i.e. a certain
Individual given time point consumer should select to buy two classification problems being also to wait for.Existing method is also
Great majority use the sorting algorithm in the middle of machine learning to build solution, and fresh few directly prediction is concrete
Machine profile.Additionally, due to price expectation relates to multiple different flight and different dates,
Modeling to price data is the most difficult.Therefore, the most feasible price expectation method still phase
To scarcity.
The machine profile observed the most in the same time can be considered a time series.In time series analysis
In, general employing moving average MA, autoregression model AR and difference slip autoregression model
The methods such as ARIMA are predicted, but owing to these traditional method great majority assume history and following number
According to linear, therefore it is not suitable for the prediction of the machine profile of Nonlinear Time Series.
Summary of the invention
The technical problem to be solved in the present invention is not possess feasible air ticket valency in prior art to overcome
The defect of lattice Predicting Technique, it is provided that a kind of fare-pricing estimate method and system.
The present invention solves above-mentioned technical problem by following technical proposals:
The present invention provides a kind of fare-pricing estimate method, and its feature is, it comprises the following steps:
S1, gather all historical datas of a target order of classes or grades at school, each historical data includes this target order of classes or grades at school
Start the date, apart from the natural law on this starting date and admission fee at that time;
S2, start on the basis of date starts by current, according to starting the date and/or distance starts the date
The time sequencing of natural law from those historical datas, extract n time series fare data to constitute n
Individual training sample;
S3, be the first training sample and the second training sample by each training sample random division;
S4, set up n and n training sample basic forecast model one to one, and to each base
Plinth forecast model is trained determining the parameters in this basic forecast model, wherein, each base
The input value of plinth forecast model be the first training sample that this basic forecast model is corresponding, output valve be current
Date admission fee;
S5, set up a Model of Fare-Pricing Forecast, the basis that each second training sample substitutes into its correspondence is pre-
Survey in model and predict to obtain n basic forecast value, and be trained determining to this Model of Fare-Pricing Forecast
Parameters in this Model of Fare-Pricing Forecast, wherein, the input value of this Model of Fare-Pricing Forecast is this n base
Plinth predictive value, output valve are current date admission fee on the same day;
S6, start the date by target and on the basis of this target starts the natural law on date, extract n
Time series fare data is to constitute n new training sample;
S7, each new training sample is substituted into and the basic forecast model of its correspondence is predicted obtain n
Basic forecast value;
S8, by step S7N basic forecast value of middle acquisition substitutes in this Model of Fare-Pricing Forecast to be predicted to obtain
Obtain admission fee predictive value.
It is preferred that in step S2In, extract 5 time series fare data successively, be respectively as follows:
Laterally sequence fare data:D represents and currently starts the date, if i=0,
Represent that the distance current starting date is admission fee when one;
Longitudinal sequence fare data: Represent current the previous day starting the date
The admission fee on the same day;
Longitudinal sequence fare data: Before representing the current starting date
The admission fee on the 7th day same day;
Longitudinal sequence fare data: Before representing the current starting date
The admission fee on the 30th day same day;
Longitudinal sequence fare data: Represent current to start residing for the date
The admission fee on the same day on correspondence date the previous year in time.
It is preferred that this basic forecast model uses K nearest neighbor algorithm to build.
It is preferred that this Model of Fare-Pricing Forecast uses linear regression model (LRM).
It is preferred that the data volume in this first training sample is not less than the data in this second training sample
Amount.
The present invention also provides for a kind of fare-pricing estimate system, and its feature is, comprising:
One acquisition module, for gathering all historical datas of a target order of classes or grades at school, each historical data includes
This target order of classes or grades at school start the date, apart from the natural law on this starting date and admission fee at that time;
One first abstraction module, on the basis of starting by current starting date, according to starting the date
And/or the time sequencing of the natural law on distance starting date extracts n time sequence from those historical datas
Row fare data is to constitute n training sample;
Stroke sub-module, being used for each training sample random division is the first training sample and the second instruction
Practice sample;
One first training module, is used for setting up n and n training sample basic forecast mould one to one
Type, and each basic forecast model is trained each ginseng determining in this basic forecast model
Number, wherein, the input value of each basic forecast model is the first training that this basic forecast model is corresponding
Sample, output valve are current date admission fee on the same day;
One second training module, is used for setting up a Model of Fare-Pricing Forecast, by each second training sample generation
Enter and the basic forecast model of its correspondence is predicted to obtain n basic forecast value, and to this fare-pricing estimate mould
Type is trained determining the parameters in this Model of Fare-Pricing Forecast, wherein, this Model of Fare-Pricing Forecast
Input value be this n basic forecast value, output valve be current date admission fee on the same day;
One second abstraction module, for the natural law with the target starting date and apart from this target starting date be
Benchmark, extracts n time series fare data to constitute n new training sample;
One first prediction module, for substituting into the basic forecast model of its correspondence by each new training sample
Middle prediction is to obtain n basic forecast value;
One second prediction module, for calling n the basic forecast value generation that this first prediction module obtains
Enter and this Model of Fare-Pricing Forecast is predicted obtain fare-pricing estimate value.
It is preferred that this first abstraction module is for extracting 5 time series fare data successively, respectively
For:
Laterally sequence fare data:D represents and currently starts the date, if i=0,
Represent that the distance current starting date is admission fee when one;
Longitudinal sequence fare data: Represent current the previous day starting the date
The admission fee on the same day;
Longitudinal sequence fare data: Before representing the current starting date
The admission fee on the 7th day same day;
Longitudinal sequence fare data: Before representing the current starting date
The admission fee on the 30th day same day;
Longitudinal sequence fare data: Represent current to start residing for the date
The admission fee on the same day on correspondence date the previous year in time.
It is preferred that this basic forecast model uses K nearest neighbor algorithm to build.
It is preferred that this Model of Fare-Pricing Forecast uses linear regression model (LRM).
It is preferred that the data volume in this first training sample is not less than the data in this second training sample
Amount.
On the basis of meeting common sense in the field, above-mentioned each optimum condition, can combination in any, i.e. get Ben Fa
Bright each preferred embodiments.
The most progressive effect of the present invention is:
The present invention proposes a kind of Data Modeling Method based on historic fare, is different from conventional time series
Only considering the way of one-dimensional sequence in prediction, the present invention is by multiple different cycles such as year, the moon, week, days
The time series extracted carries out integrated forecasting, thus avoids risk and the limitation of loss of learning, for machine
Admission fee lattice provide a kind of practicable Forecasting Methodology, thus the present invention can go out target and rise by Accurate Prediction
Dynamic date and the machine profile when this target starts the natural law on date.
Accompanying drawing explanation
Fig. 1 is that the sample of the flight prices table of present pre-ferred embodiments is shown.
Fig. 2 is the arrangement flow chart of the fare-pricing estimate method of present pre-ferred embodiments.
Detailed description of the invention
Further illustrate the present invention below by the mode of embodiment, but the most therefore limit the present invention to
Among described scope of embodiments.
Embodiment of the present invention is classified into model training and two parts of fare-pricing estimate, wherein, the mesh of training
Be to build available Model of Fare-Pricing Forecast, it follows the steps below:
Step 1: determine the target flight of prediction, and obtain this target flight institute to current date
There is historical data.Such as, target flight is set to Beijing MU5137 the flight to Shanghai, current date
Be JIUYUE in 2015 15, then historical data set should comprise all JIUYUE in 2015 15 days and
The historgraphic data recording observed in the past, each historgraphic data recording comprises concrete taking off day
Phase, apart from the natural law of this date and price at that time.
Step 2: work out the price list of this target flight on the basis of historical data set.Fig. 1 is
The price list woven, in table, every string represents a fixing date, and every a line represents a spy
Fixed distance take off before natural law, the machine profile that the value in cell then observes for this moment.By Fig. 1
Visible, owing to current date is JIUYUE in 2015 15, the flight historical price taken off in this day is
Known quantity, and the flight taken off for 16th in JIUYUE 1 day and price earlier before take-off are also for known quantity,
By that analogy.
Step 3: form training dataset.Make d represent currently and start the date, if i=0,Represent and work as
The admission fee of the admission fee of front starting date, i.e. this flight JIUYUE in 2015 same day on the 15th is (in Fig. 1
653), for each known units lattice in price list, according to the different time cycles, extract successively
Five time serieses:
1) horizontal sequence fare dataI.e. table is positioned atAll cells with on the right side of a line:
2) longitudinal sequence fare dataI.e. table is positioned atAll cells with above string:
3) longitudinal sequence fare dataI.e. table is positioned atWith owning with 7 as cycle above string
Cell:
4) longitudinal sequence fare dataI.e. table is positioned atWith above string with one month as cycle
All cells, wherein, the length that each cycle crosses over should according to of that month total natural law, from 28 to
Value in 31:
5) longitudinal sequence fare dataI.e. table is positioned atWith institute with 1 year as cycle above string
Having cell, wherein, the length that each cycle crosses over should take 365 or 366 according to total natural law then:
In the price list of Fig. 1, as a example by the cell of 2015-09-15 row the 0th row, by different wire
Five time series fare data are indicated.The fare data of each cell and extract five thereof
Time series component constitutes five training samples.Significantly, since data are limited, time series
Fare data can not infinitely extend, it is therefore necessary to limits each sequence queue size.In the application, each
The sequence length that visual angle generates according to the size of data volume or should be determined by experiment.In this example, take
Window size is 4, and the most each sequence comprises the fare data in 4 historical datas.
Step 4: as shown in Figure 2, is randomly divided into two parts by each training sample, the i.e. first instruction
Practicing sample and the second training sample, those first training samples composing training collection a, training set a is used for instructing
The basic forecast model that five time series components of white silk are corresponding, those second training samples composing training collection b,
Training set b is used for training Model of Fare-Pricing Forecast, i.e. extra arbitration learner in Fig. 2.The ration of division
Can determine according to concrete data volume, such as, can distribute training set by the proportion of 75% and 25%
A and training set b.
Step 5: generate five basic forecast model f1,f2,f3,f4,f5, these five basic forecast models are respectively
With five the first training sample one_to_one corresponding.Use training set a generated in step 4 that it is carried out respectively
Training is to determine the parameters in this basic forecast model, as shown in the solid arrow in Fig. 2, each
The input value of individual basic forecast model is that the first training sample that this basic forecast model is corresponding, output valve are
Current date admission fee on the same day.In this example, the K nearest neighbor algorithm (KNN) in machine learning is used
Being used as basic forecast model, each KNN model is respectively acting onWhen five
Between in sequence one.
It is pointed out that KNN algorithm the most merely illustrative, in practical operation, any can
Machine learning algorithm or other models of predicted time sequence can be replaced in the framework of this method,
Therefore, its equivalent form of value also should be protected according to the claim detailed rules and regulations in the present invention.
Step 6: generate an extra arbitration learner, i.e. Model of Fare-Pricing Forecast, and with in step 4
It is trained by training set b generated.Dotted portion in Fig. 2 illustrates the process of this step, first
First, for the second training sample of each in training set b, five basic forecast moulds all it are inputted to
Type is to obtain five different basic forecast values.Secondly, using five basic forecast values as arbitration learner
Input, it, as learning target, is further trained by current date admission fee on the same day.In this example
In, this arbitration learner has selected linear regression model (LRM), and it is substantially the prediction for five basic models
Value imparts different weights, draws final predicting the outcome by weighting.
After above step, it is possible to obtain five the basic forecast models trained and one secondary
Cut out learner, at this point it is possible to the blank cell of lower section in price list is predicted.
Concretely comprising the following steps of prediction:
Step 7: determine prediction target.Such as, target flight is that JIUYUE in 2015 is taken off on the 25th
MU5137 flight, current date is JIUYUE 15 days (distance is taken off first 10 days) in 2015, it was predicted that
Target is this flight price JIUYUE (distance is taken off first 9 days) on the 16th.
Step 8: for the cell of d=2015-09-25 in price list, i=9, on the basis of it, press
Five time series characteristic components of correspondence are extracted according to the method in step 3With
Constitute five new training samples.
Step 9: five new training samples are flowed to respectively the basic forecast model of correspondence, obtains five
Basic forecast value.
Step 10: flow to five basic forecast values arbitrate learner, it is thus achieved that final predicts the outcome i.e.
Fare-pricing estimate value.
The present embodiment proposes a kind of data modeling based on price list and fare-pricing estimate method, greatly simplifies
The most complicated machine profile data represent, also the different price observation of order in time interrelated
Clearly presented.It is different from the way only considering one-dimensional sequence in conventional time series prediction, this enforcement
The time series that example is extracted by multiple different cycles such as year, the moon, week, days carries out integrated forecasting, from
And avoid risk and the limitation of loss of learning, provide a kind of practicable prediction side for machine profile
Method.
Although the foregoing describing the detailed description of the invention of the present invention, but those skilled in the art should managing
Solving, these are merely illustrative of, and protection scope of the present invention is defined by the appended claims.This
The technical staff in field, can be to these embodiment party on the premise of without departing substantially from the principle of the present invention and essence
Formula makes various changes or modifications, but these changes and amendment each fall within protection scope of the present invention.
Claims (10)
1. a fare-pricing estimate method, it is characterised in that it comprises the following steps:
S1, gather all historical datas of a target order of classes or grades at school, each historical data includes this target order of classes or grades at school
Start the date, apart from the natural law on this starting date and admission fee at that time;
S2, start on the basis of date starts by current, according to starting the date and/or distance starts the date
The time sequencing of natural law from those historical datas, extract n time series fare data to constitute n
Individual training sample;
S3, be the first training sample and the second training sample by each training sample random division;
S4, set up n and n training sample basic forecast model one to one, and to each base
Plinth forecast model is trained determining the parameters in this basic forecast model, wherein, each base
The input value of plinth forecast model be the first training sample that this basic forecast model is corresponding, output valve be current
Date admission fee;
S5, set up a Model of Fare-Pricing Forecast, the basis that each second training sample substitutes into its correspondence is pre-
Survey in model and predict to obtain n basic forecast value, and be trained determining to this Model of Fare-Pricing Forecast
Parameters in this Model of Fare-Pricing Forecast, wherein, the input value of this Model of Fare-Pricing Forecast is this n base
Plinth predictive value, output valve are current date admission fee on the same day;
S6, start the date by target and on the basis of this target starts the natural law on date, extract n
Time series fare data is to constitute n new training sample;
S7, each new training sample is substituted into and the basic forecast model of its correspondence is predicted obtain n
Basic forecast value;
S8, by step S7N basic forecast value of middle acquisition substitutes in this Model of Fare-Pricing Forecast to be predicted to obtain
Obtain admission fee predictive value.
2. fare-pricing estimate method as claimed in claim 1, it is characterised in that in step S2In, depend on
Secondary extract 5 time series fare data, be respectively as follows:
Laterally sequence fare data:D represents and currently starts the date, if i=0,
Represent that the distance current starting date is admission fee when one;
Longitudinal sequence fare data: Represent current the previous day starting the date
The admission fee on the same day;
Longitudinal sequence fare data: Before representing the current starting date
The admission fee on the 7th day same day;
Longitudinal sequence fare data: Before representing the current starting date
The admission fee on the 30th day same day;
Longitudinal sequence fare data: Represent current to start residing for the date
The admission fee on the same day on correspondence date the previous year in time.
3. fare-pricing estimate method as claimed in claim 1, it is characterised in that this basic forecast model
K nearest neighbor algorithm is used to build.
4. fare-pricing estimate method as claimed in claim 1, it is characterised in that this Model of Fare-Pricing Forecast
Use linear regression model (LRM).
5. fare-pricing estimate method as claimed in claim 1, it is characterised in that this first training sample
In data volume not less than the data volume in this second training sample.
6. a fare-pricing estimate system, it is characterised in that comprising:
One acquisition module, for gathering all historical datas of a target order of classes or grades at school, each historical data includes
This target order of classes or grades at school start the date, apart from the natural law on this starting date and admission fee at that time;
One first abstraction module, on the basis of starting by current starting date, according to starting the date
And/or the time sequencing of the natural law on distance starting date extracts n time sequence from those historical datas
Row fare data is to constitute n training sample;
Stroke sub-module, being used for each training sample random division is the first training sample and the second instruction
Practice sample;
One first training module, is used for setting up n and n training sample basic forecast mould one to one
Type, and each basic forecast model is trained each ginseng determining in this basic forecast model
Number, wherein, the input value of each basic forecast model is the first training that this basic forecast model is corresponding
Sample, output valve are current date admission fee on the same day;
One second training module, is used for setting up a Model of Fare-Pricing Forecast, by each second training sample generation
Enter and the basic forecast model of its correspondence is predicted to obtain n basic forecast value, and to this fare-pricing estimate mould
Type is trained determining the parameters in this Model of Fare-Pricing Forecast, wherein, this Model of Fare-Pricing Forecast
Input value be this n basic forecast value, output valve be current date admission fee on the same day;
One second abstraction module, for the natural law with the target starting date and apart from this target starting date be
Benchmark, extracts n time series fare data to constitute n new training sample;
One first prediction module, for substituting into the basic forecast model of its correspondence by each new training sample
Middle prediction is to obtain n basic forecast value;
One second prediction module, for calling n the basic forecast value generation that this first prediction module obtains
Enter and this Model of Fare-Pricing Forecast is predicted obtain fare-pricing estimate value.
7. fare-pricing estimate system as claimed in claim 6, it is characterised in that this first abstraction module
For extracting 5 time series fare data successively, it is respectively as follows:
Laterally sequence fare data:D represents and currently starts the date, if i=0,
Represent that the distance current starting date is admission fee when one;
Longitudinal sequence fare data: Represent current the previous day starting the date
The admission fee on the same day;
Longitudinal sequence fare data: Before representing the current starting date
The admission fee on the 7th day same day;
Longitudinal sequence fare data: Before representing the current starting date
The admission fee on the 30th day same day;
Longitudinal sequence fare data: Represent current to start residing for the date
The admission fee on the same day on correspondence date the previous year in time.
8. fare-pricing estimate system as claimed in claim 6, it is characterised in that this basic forecast model
K nearest neighbor algorithm is used to build.
9. fare-pricing estimate system as claimed in claim 6, it is characterised in that this Model of Fare-Pricing Forecast
Use linear regression model (LRM).
10. fare-pricing estimate system as claimed in claim 6, it is characterised in that this first training sample
In data volume not less than the data volume in this second training sample.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610187700.5A CN105869017A (en) | 2016-03-29 | 2016-03-29 | Method and system for predicting ticket prices |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610187700.5A CN105869017A (en) | 2016-03-29 | 2016-03-29 | Method and system for predicting ticket prices |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105869017A true CN105869017A (en) | 2016-08-17 |
Family
ID=56625325
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610187700.5A Pending CN105869017A (en) | 2016-03-29 | 2016-03-29 | Method and system for predicting ticket prices |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105869017A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107767094A (en) * | 2017-09-22 | 2018-03-06 | 北京京东尚科信息技术有限公司 | A kind of logistics ladder freight charges optimization method and device |
CN109472399A (en) * | 2018-10-23 | 2019-03-15 | 上海交通大学 | Consider the air ticket purchase decision method and system of uncertainty in traffic |
CN110111036A (en) * | 2019-03-28 | 2019-08-09 | 跨越速运集团有限公司 | Logistics goods amount prediction technique and system based on LSTM Model Fusion |
CN111091407A (en) * | 2019-10-28 | 2020-05-01 | 海南太美航空股份有限公司 | Airline passenger seat rate prediction method and system |
CN112132323A (en) * | 2020-08-25 | 2020-12-25 | 汉海信息技术(上海)有限公司 | Method and device for predicting value amount of commodity object and electronic equipment |
CN112767035A (en) * | 2021-01-25 | 2021-05-07 | 海南太美航空股份有限公司 | Flight fare estimation method, system and electronic equipment |
-
2016
- 2016-03-29 CN CN201610187700.5A patent/CN105869017A/en active Pending
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107767094A (en) * | 2017-09-22 | 2018-03-06 | 北京京东尚科信息技术有限公司 | A kind of logistics ladder freight charges optimization method and device |
CN109472399A (en) * | 2018-10-23 | 2019-03-15 | 上海交通大学 | Consider the air ticket purchase decision method and system of uncertainty in traffic |
CN110111036A (en) * | 2019-03-28 | 2019-08-09 | 跨越速运集团有限公司 | Logistics goods amount prediction technique and system based on LSTM Model Fusion |
CN111091407A (en) * | 2019-10-28 | 2020-05-01 | 海南太美航空股份有限公司 | Airline passenger seat rate prediction method and system |
CN112132323A (en) * | 2020-08-25 | 2020-12-25 | 汉海信息技术(上海)有限公司 | Method and device for predicting value amount of commodity object and electronic equipment |
CN112767035A (en) * | 2021-01-25 | 2021-05-07 | 海南太美航空股份有限公司 | Flight fare estimation method, system and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105869017A (en) | Method and system for predicting ticket prices | |
CN107895283B (en) | Merchant passenger flow volume big data prediction method based on time series decomposition | |
CN106909933B (en) | A kind of stealing classification Forecasting Methodology of three stages various visual angles Fusion Features | |
Xie et al. | Variable selection methods for probabilistic load forecasting: Empirical evidence from seven states of the united states | |
CN103729689B (en) | Power grid electric quantity prediction method based on industry classifications and leading industry data | |
CN107844915A (en) | A kind of automatic scheduling method of the call center based on traffic forecast | |
CN105023066B (en) | A kind of Business Process System analysing and predicting system and method based on seasonal adjustment | |
CN109214719A (en) | A kind of system and method for the marketing inspection analysis based on artificial intelligence | |
CN111062757A (en) | Information recommendation method and system based on multi-path optimization matching | |
CN103310286A (en) | Product order prediction method and device with time series characteristics | |
CN102541920A (en) | Method and device for improving accuracy degree by collaborative filtering jointly based on user and item | |
CN112700027A (en) | Multi-precision index comprehensive evaluation method and system for electric power prediction model | |
CN108388974A (en) | Top-tier customer Optimum Identification Method and device based on random forest and decision tree | |
CN107909392A (en) | Stimulation behavior susceptibility Forecasting Methodology, device, equipment and readable storage medium storing program for executing | |
CN105912686A (en) | Search engine marketing bid method and system based on machine learning | |
CN106779478A (en) | A kind of load scheduling Valuation Method | |
CN104346698A (en) | Catering member big data analysis and checking system based on cloud computing and data mining | |
CN108846695A (en) | The prediction technique and device of terminal replacement cycle | |
CN110458668A (en) | Determine the method and device of Products Show algorithm | |
CN108154311A (en) | Top-tier customer recognition methods and device based on random forest and decision tree | |
CN107679103A (en) | For entity attributes analysis method and system | |
Ripoll-Zarraga et al. | Exploring the reasons for efficiency in Spanish airports | |
CN114782065A (en) | Commodity sales volume prediction method and device based on model combination and storage medium | |
CN114239292B (en) | Comprehensive evaluation method and system for low-carbon economic operation-oriented multifunctional demand park | |
CN104794322B (en) | Solar cell module limited buffer multiple target lot size scheduling method based on DNSGA II |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160817 |