CN109754086A - One kind being based on markovian ridge regression Numerical Predicting Method - Google Patents
One kind being based on markovian ridge regression Numerical Predicting Method Download PDFInfo
- Publication number
- CN109754086A CN109754086A CN201910015180.3A CN201910015180A CN109754086A CN 109754086 A CN109754086 A CN 109754086A CN 201910015180 A CN201910015180 A CN 201910015180A CN 109754086 A CN109754086 A CN 109754086A
- Authority
- CN
- China
- Prior art keywords
- prediction
- model
- determined
- numerical
- markovian
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to belong to machine learning and the field of data mining based on markovian ridge regression Numerical Predicting Method, it is characterised in that uses following steps: (1) determining the output of model(2) loss function L (ω) is determined；(3) value of ω is determined；(4) residual sequence E={ e is determined1, e2..., en}；(5) determine that m walks state-transition matrix p(m)；(6) known state transfer matrix p(m)With original state ei, establish Markov Chain and prediction result be modified；(7) using test set sample as input, model training is carried out, prediction result is obtained.The present invention is based on markovian ridge regression Numerical Predicting Methods, initially set up ridge regression model and are predicted, are then corrected using the residual error that Markov Chain generates prediction, realize that numerical value is precisely predicted.By multi-group data experimental result it is found that comparing compared to other models, the present invention provides the prediction technique for enhancing model generalization ability on the basis of ensure that precision of prediction for numerical prediction.
Description
Technical field
The present invention relates to machine learning and the field of data mining, mainly a kind of Numerical Predicting Method.
Background technique
It is directed to numerical prediction problem at present, most models can achieve very high precision of prediction, but model is easy
Existing over-fitting, with the raising of Training Capability, predictive ability declines instead.These models are often shown on training set very well
Estimated performance, however on test set or in some unknown data, predictive ability be will be greatly reduced.Especially nerve net
Network, although neural network can achieve very high precision of prediction in terms of numerical prediction, it is easy to appear over-fittings, and
Model is excessively complicated, and the training time is too long.By taking BP neural network as an example, that there are approximation capabilities is weak for the BP neural network of early stage, receives
A series of problems, such as it is slow to hold back speed, easily falls into local optimum.The generalization ability of BP neural network is easy to be influenced by topological structure,
A kind of unification there is no and complete theoretical direction to the selection of its network structure at present, can only generally be selected by experience.This is one
Determine to increase the uncertainty during model training in degree.In addition, the selection of neural network initial parameter is to network performance
With very big influence.Network is easily trapped into local minimum point during training, although proposing a series of parameter
Optimization method, such as genetic algorithm, cuckoo search, mesh parameter optimizing etc., but these method calculation process are complicated, calculate
Amount is big, for example genetic algorithm is needed by crossing over many times, variation mode is by the breeding and the survival of the fittest in N generation, needs to expend big
The amount time carries out parameter adjustment.Greatly reduce the applicability of model.
China also achieves some abundant achievements in terms of numerical prediction.For example, using deep learning in annual precipitation, friendship
Through-current capacity, stock trend etc. realize preferable numerical prediction.With computer and artificial intelligence fast development and
The exponential growth of mass data, more stringent requirements are proposed for precision of prediction and predetermined speed to model.Therefore, Yao Shixian
Accurately and efficiently the numerical value in certain fields is predicted, reaches the requirement of numerical prediction, it is necessary to establish it is a kind of accurate, be not easy
The Numerical Predicting Method of over-fitting is further simplified the training process of model, improves predetermined speed and precision of model, is China
The fields such as economic and business provide a kind of accurate, efficient numerical prediction mode, provide constructive suggestions for following management.
Summary of the invention
For above-mentioned problems of the prior art, the technical problem to be solved in the present invention is to provide a kind of accurate, high
The numerical value trend forecasting method of effect, detailed process are as shown in Figure 1.
Technical solution implementation steps are as follows:
(1) output of model is determined
Wherein, Yk-1={ y (0), y (1) ..., y (k-1) }, U (k)={ u (0), u (1) ..., u (k) }, u (k) is defeated
Enter variable, θ is that continually changing parameter, k are discrete times to m dimension at any time.
(2) loss function L (w) is determined:
L (w)=wTu(k)Tu(k)w-y(k)Tu(k)w+wTu(k)Ty(k)-y(k)Ty(k)+λwTw
In formula,For model predication value, y (k) is actual value,For the regular terms of addition.
(3) value of w is determined:
SeparatelyI.e.
Obtain w=(u (k)Tu(k)+λE)-1u(k)Ty
(4) residual sequence E={ e is determined1, e2..., en}:
In formula,For model predication value, y (k) is actual value.
(5) determine that m walks state-transition matrix p(m):
(6) known state transfer matrix p(m)With original state ei, establish Markov Chain and prediction result be modified.
(7) using test set sample as input, model training is carried out, obtains prediction result, completes to be based on Markov Chain
Ridge regression Numerical Predicting Method.
The present invention has the advantage that than the prior art:
(1) present invention employs by the mode of ridge regression prediction and Markov chain combination, the advantage of the two has been merged, more
Respective shortcoming is mended.It initially sets up ridge regression model to be predicted, then prediction is generated using Markov Chain
Residual error be corrected.The accurate prediction for realizing numerical value, improves the precision of prediction of model.
(2) present invention joined regular terms in ridge regression model, intend so that model was less prone in the training process
It closes, on test set or in some unknown data, model prediction ability is strong.
(3) present invention tests multi-group data, and experimental result all achieves obvious compared with prior art
Advantage, and data result is relatively stable.This illustrates that the present invention improves the general of model on the basis of ensure that precision of prediction
Change ability can preferably complete numerical prediction task.
For a better understanding of the present invention, it is further described with reference to the accompanying drawing.
Fig. 1 is to establish the step flow chart based on markovian ridge regression Numerical model；
Fig. 2 is to establish the algorithm flow chart based on markovian ridge regression Numerical model；
Fig. 3 is the experimental result comparison of a variety of models；
Specific embodiment
Below by case study on implementation, invention is further described in detail.
By taking photovoltaic power generation power prediction as an example, the data set of selection is certain -2018 years 2017 power generation of photo-voltaic power generation station
Data meteorological data corresponding with its includes the generated output that daily 8:00-19:00 has 12 integral point moment altogether in data set,
Share 4380 datas record.Wherein, 3504 datas are used as training set, and 876 datas are used as test set.
Numerical prediction overall flow provided by the present invention is as shown in Figure 1, the specific steps are as follows:
(1) loss function L (w) is determined:
L (w)=wTu(k)Tu(k)w-y(k)Tu(k)w+wTu(k)Ty(k)-y(k)Ty(k)+λwTw
In formula,For model predication value, y (k) is actual value,For the regular terms of addition.
(2) value of w is determined:
It enablesI.e.
?
W=(u (k)Tu(k)+λE)-1u(k)Ty
(3) output of model is determined
Wherein, Yk-1={ y (0), y (1) ..., y (k-1) }, U (k)={ u (0), u (1) ..., u (k) }, u (k) is defeated
Enter variable, θ is that continually changing parameter, k are discrete times to m dimension at any time.In this example, the value of k is set as 5.
(4) residual sequence E={ e is determined1, e2..., e3504}:
In formula,For model predication value, y (k) is actual value.
(5) determine that m walks state-transition matrix p(m):
In this example, the value of m is set as 4, that is, seeks 4 step state-transition matrix p(4)。
(6) known state transfer matrix p(m)With original state ei, establish Markov Chain and prediction result be modified.
(7) using test set sample as input, model training is carried out, obtains prediction result, completes to be based on Markov Chain
Ridge regression Numerical Predicting Method.
In order to verify the precision of logarithm prediction of the present invention and the generalization ability of model, the present invention is carried out with test set
Multiple groups numerical prediction emulation experiments, and result and some other prediction models are compared, simulation result such as 1 institute of table
Show.
More than a kind of model experiment results comparison of table
| Experimental method | MAE (%) |
| Based on the Numerical Predicting Method for improving CS algorithm optimization Elman-IOC neural network | 1.65 |
| The Numerical Predicting Method of improved BP algorithm | 1.78 |
| Numerical Predicting Method based on machine learning Xgboost model | 2.54 |
| The present invention | 1.26 |
By simulation result table 1 it is found that the present invention with Markov Chain by being carried out using same data set
After residual prediction amendment, the prediction result of model has reached very high precision, mean absolute error 1.26%.With other three
Kind method is compared, and has higher precision and stronger applicability.This shows that the Numerical Predicting Method that the present invention establishes is accurate
, effective method is provided to establish accurate Numerical model.
Claims (1)
1. being based on markovian ridge regression Numerical Predicting Method, specific classifying step is as follows:
(1) output of model is determined
Wherein, Yk-1={ y (0), y (1) ..., y (k-1) }, U (k)={ u (0), u (1) ..., u (k) }, u (k) are that input becomes
Amount, θ are that continually changing parameter, k are discrete times to m dimension at any time；
(2) loss function L (ω) is determined:
L (ω)=ωTu(k)Tu(k)ω-y(k)Tu(k)ω+ωTu(k)Ty(k)-y(k)Ty(k)+λωTω
In formula,For model predication value, y (k) is actual value,For the regular terms of addition；
(3) value of ω is determined:
SeparatelyI.e.
Obtain ω=(u (k)Tu(k)+λE)-1u(k)Ty
(4) residual sequence E={ e is determined1, e2..., en}:
In formula,For model predication value, y (k) is actual value；
(5) determine that m walks state-transition matrix p(m):
(6) known state transfer matrix p(m)With original state ei, establish Markov Chain and prediction result be modified；
(7) using test set sample as input, model training is carried out, obtains prediction result, completes to be based on markovian ridge
Return Numerical Predicting Method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910015180.3A CN109754086A (en) | 2019-01-13 | 2019-01-13 | One kind being based on markovian ridge regression Numerical Predicting Method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910015180.3A CN109754086A (en) | 2019-01-13 | 2019-01-13 | One kind being based on markovian ridge regression Numerical Predicting Method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109754086A true CN109754086A (en) | 2019-05-14 |
Family
ID=66405299
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910015180.3A Pending CN109754086A (en) | 2019-01-13 | 2019-01-13 | One kind being based on markovian ridge regression Numerical Predicting Method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109754086A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111369070A (en) * | 2020-03-13 | 2020-07-03 | 西安理工大学 | Envelope clustering-based multimode fusion photovoltaic power prediction method |
| CN111967652A (en) * | 2020-07-22 | 2020-11-20 | 国网浙江省电力有限公司电力科学研究院 | Double-layer cooperative real-time correction photovoltaic prediction method |
| CN117150876A (en) * | 2022-05-23 | 2023-12-01 | 北京理工大学 | Method for controlling charge density of pressed mixed explosive based on ridge regression algorithm |
-
2019
- 2019-01-13 CN CN201910015180.3A patent/CN109754086A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111369070A (en) * | 2020-03-13 | 2020-07-03 | 西安理工大学 | Envelope clustering-based multimode fusion photovoltaic power prediction method |
| CN111369070B (en) * | 2020-03-13 | 2023-06-27 | 西安理工大学 | Multimode fusion photovoltaic power prediction method based on envelope clustering |
| CN111967652A (en) * | 2020-07-22 | 2020-11-20 | 国网浙江省电力有限公司电力科学研究院 | Double-layer cooperative real-time correction photovoltaic prediction method |
| CN111967652B (en) * | 2020-07-22 | 2023-10-24 | 国网浙江省电力有限公司电力科学研究院 | Double-layer collaborative real-time correction photovoltaic prediction method |
| CN117150876A (en) * | 2022-05-23 | 2023-12-01 | 北京理工大学 | Method for controlling charge density of pressed mixed explosive based on ridge regression algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108900346B (en) | Wireless network flow prediction method based on LSTM network | |
| CN108170529A (en) | A kind of cloud data center load predicting method based on shot and long term memory network | |
| CN110969290B (en) | Runoff probability prediction method and system based on deep learning | |
| CN109754086A (en) | One kind being based on markovian ridge regression Numerical Predicting Method | |
| Gao et al. | Forecasting annual natural gas consumption via the application of a novel hybrid model | |
| CN103150609A (en) | Modeling method for short time traffic flow predicting model | |
| CN105447510B (en) | Fluctuating wind speed prediction technique based on artificial bee colony optimization LSSVM | |
| CN104408518A (en) | Method of learning and optimizing neural network based on particle swarm optimization algorithm | |
| CN112001556B (en) | Reservoir downstream water level prediction method based on deep learning model | |
| CN108876001A (en) | A kind of Short-Term Load Forecasting Method based on twin support vector machines | |
| CN111461463A (en) | Short-term load prediction method, system and equipment based on TCN-BP | |
| CN109754122A (en) | A kind of Numerical Predicting Method of the BP neural network based on random forest feature extraction | |
| CN111144663A (en) | Ultra-short-term wind power prediction method for offshore wind farm considering output fluctuation process | |
| CN109934422A (en) | Neural network wind speed prediction method based on time series data analysis | |
| CN114219181A (en) | Wind power probability prediction method based on transfer learning | |
| CN111667377A (en) | Full-automatic power consumption prediction method and system | |
| CN109214610B (en) | Saturated power load prediction method based on long-term and short-term memory neural network | |
| CN113595798A (en) | Network flow prediction method and system for improving lightning connection process optimization algorithm | |
| CN117521511A (en) | Granary temperature prediction method based on improved wolf algorithm for optimizing LSTM | |
| CN117439053A (en) | Method, device and storage medium for predicting electric quantity of Stacking integrated model | |
| CN113300884B (en) | GWO-SVR-based step-by-step network flow prediction method | |
| CN105913144A (en) | Product life prediction method based on target oriented best matching similarity | |
| CN104537224A (en) | Multi-state system reliability analysis method and system based on self-adaptive learning algorithm | |
| CN114943277A (en) | Method for identifying operation condition and predicting operation performance of comprehensive energy system | |
| CN114372615A (en) | Short-term photovoltaic power prediction method and system considering spatial correlation |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190514 |
|
| WD01 | Invention patent application deemed withdrawn after publication |