CN110297712A - A kind of ARIMA combination forecasting method towards block chain node load estimation - Google Patents
A kind of ARIMA combination forecasting method towards block chain node load estimation Download PDFInfo
- Publication number
- CN110297712A CN110297712A CN201910420703.2A CN201910420703A CN110297712A CN 110297712 A CN110297712 A CN 110297712A CN 201910420703 A CN201910420703 A CN 201910420703A CN 110297712 A CN110297712 A CN 110297712A
- Authority
- CN
- China
- Prior art keywords
- arima
- block chain
- chain node
- model
- combination forecasting
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5083—Techniques for rebalancing the load in a distributed system
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
Abstract
The invention discloses a kind of ARIMA combination forecasting methods towards block chain node load estimation.Carry out in the steps below: a. is according to the remaining load rate of the resource service condition calculation block chain node of block chain node, using remaining load rate as original time series;B. tranquilization processing is carried out to original time series;C. rank is determined to ARIMA according to the recognition rule of time series models, constructs ARIMA model, and predicted using the ARIMA model that building is completed, obtains initial predicted value, and calculate the residual error of initial predicted value, construct the residual sequence of initial predicted value;D. forecast analysis is carried out to residual error using BP neural network and obtains residual analysis value, to correct the residual sequence, be finally added the initial predicted value of ARIMA model with residual analysis value to obtain the combined prediction value of ARIMA-BP combination forecasting.The present invention has the characteristics that improve remaining load rate precision of prediction.
Description
Technical field
It is especially a kind of towards block chain node load estimation the present invention relates to the resource allocation techniques field of block chain
ARIMA combination forecasting method.
Background technique
Medical data shared system based on block chain mainly stores medical data and shares access for user.In the system
In, medical data storage is in block chain, and all nodes in system all have a complete block chain, and block chain sheet
It is the distributed data base of a decentralization in matter.Therefore, medical data is shared really looks into distributed data base
Ask operation.With the increase of user query request task quantity, the pressure of server is gradually increased.If server cannot be reasonable
Distribution task, it will it is largely idle part of server resource occur, the case where another part server resource wretched insufficiency, has
It may cause service performance and reduce and even stop service, resource utilization decline seriously affects service quality.The residue of server
Load factor can be very good to embody the current performance of server.Therefore the load variation of each node is obtained ahead of time, to server
Remaining load rate is predicted, can more reasonably be distributed resource for server and be played an important role.But at present to remaining load
There are a difficulties for the prediction of rate, because the remaining load rate of current most of server is by concurrent request number, service etc.
The influence of the many-sided enchancement factor of grade agreement etc., causes data to have very big random fluctuation, so cause both included in data
Linear data, and doped with nonlinear data.This just has larger impact to the accuracy of remaining load rate prediction.
Summary of the invention
The object of the present invention is to provide a kind of ARIMA combination forecasting methods towards block chain node load estimation.This
Invention has the characteristics that improve remaining load rate precision of prediction.
Technical solution of the present invention: a kind of ARIMA combination forecasting method towards block chain node load estimation, by following
Step carries out:
A. according to the remaining load rate of the resource service condition calculation block chain node of block chain node, by remaining load rate
As original time series;
B. tranquilization processing is carried out to original time series;
C. rank is determined to ARIMA according to the recognition rule of time series models, constructs ARIMA model, and complete using building
ARIMA model predicted, obtain initial predicted value, and calculate the residual error of initial predicted value, construct initial predicted value
Residual sequence;
D. forecast analysis is carried out to residual error using BP neural network and obtains residual analysis value, to correct the residual sequence,
Finally it is added the initial predicted value of ARIMA model to obtain the combined prediction of ARIMA-BP combination forecasting with residual analysis value
Value.
In step d described in ARIMA combination forecasting method above-mentioned towards block chain node load estimation, ARIMA- is obtained
After the combined prediction value of BP combination forecasting, the abnormal coefficient of the initial predicted value of ARIMA model is calculated using LOF algorithm,
When abnormal coefficient is in default LOF threshold value, use initial predicted value as the final predicted value of model;
When abnormal coefficient is in outside default LOF threshold value, use combined prediction value as the final predicted value of model.
In step a described in ARIMA combination forecasting method above-mentioned towards block chain node load estimation, block chain node
Resource service condition be the CPU of block chain node and the resource service condition of memory.
In step a described in ARIMA combination forecasting method above-mentioned towards block chain node load estimation, remaining load rate
It calculates as follows:
If block chain node set is P={ P1,P2,…,Pn};
When there is multiple queries to request while reaching, the resources left state L of current i-th node of system predictioni, and pass
To load balancer;Then have:
Li=α Lc+βLm
Wherein, LcFor CPU residue, LmFor memory residue, α is CPU weight, and β is memory weight, and alpha+beta=1,1≤i≤n;
The remaining load summation of block chain node in block chain node set are as follows:
The then remaining load rate E of i-th nodeiAre as follows:
In step a described in ARIMA combination forecasting method above-mentioned towards block chain node load estimation, block chain node
Resource service condition acquisition when, according in system in the past period query task average completion time determine acquisition frequency
Rate.
The average completion time described in ARIMA combination forecasting method above-mentioned towards block chain node load estimation
When being AT seconds, the frequency acquisition is the resource service condition at interval of the block chain node of acquisition in AT/3 seconds.
It is described to original in step b described in ARIMA combination forecasting method above-mentioned towards block chain node load estimation
Beginning time series carries out tranquilization processing, specifically carries out tranquilization processing to original time series using the method for difference.
In step c described in ARIMA combination forecasting method above-mentioned towards block chain node load estimation, the time
The recognition rule of series model, specifically BIC criterion.
It is described in step c described in ARIMA combination forecasting method above-mentioned towards block chain node load estimation
ARIMA model carries out parameter Estimation in building, using least square method or maximum-likelihood method, while verifying in building
ARIMA model diagnoses the residual sequence of the ARIMA model in building;When the residual sequence of the ARIMA model in building is not white
When noise sequence, rank is determined to ARIMA again, until completing when the residual sequence of the ARIMA model in building is white noise sequence
The building of ARIMA model.
In step c described in ARIMA combination forecasting method above-mentioned towards block chain node load estimation, completion is constructed
ARIMA model is specially (1,1,1) ARIMA.
Beneficial effect
Compared with prior art, ARIMA model and BP neural network are bonded ARIMA-BP combined prediction by the present invention
Model predicts remaining load rate by ARIMA-BP combination forecasting, overcomes and uses single ARIMA model prediction
When, bad to the treatment effect of nonlinear data, precision of prediction is not high, and when use single BP neural network prediction, to linear
The inadequate problem of data mining has given full play to advantage and BP neural network of the ARIMA model in terms of linear data prediction and has existed
The advantage of nonlinear data prediction aspect, finally improves ARIMA-BP combination forecasting to the remaining load of block chain node
The precision of prediction of rate.
The present invention calculates the abnormal coefficient of the initial predicted value of ARIMA model using LOF algorithm, rejects in ARIMA model
Abnormal point initial predicted value, using combined prediction value of the ARIMA-BP combination forecasting to the abnormal point substitute this initially
Predicted value is as final predicted value;In this way, the precision of prediction of ARIMA-BP combination forecasting is further improved, from
And remaining load rate precision of prediction is also improved, resource, which is more reasonably distributed, for server plays important function.
In order to prove beneficial effects of the present invention, applicant carried out following experiments:
1. apparatus and method for
1.1 equipment
Using the medical data shared system based on block chain as experimental subjects, the shared system is negative by 1 for this experiment
Carry balanced node and 2 service node compositions.Load balancing node is configured to 8 thread 2.6GHzCPU of 8GB memory/4 core;
Ubuntu-16.04.The configuration of two service nodes, one identical as load balancing node, and one is 4GB memory/4 core, 4 thread
2.6GHzCPU;Ubuntu-16.04.
1.2 method
By the relatively simple ARIMA model of this experiment, ARIMA-BP combination forecasting and using LOF algorithm to ARIMA-
BP combination forecasting further corrects the prediction of the remaining load rate of gained model (hereinafter referred to as LOF-ARIMA-BP model)
Ability.
Firstly, node collects the CPU of this node and the service condition of memory by monitor;Then, according to formulaCalculate remaining load rate and the historical data as prediction;Finally, predicting node using above three prediction model
Subsequent remaining load rate.
2. experimental evaluation index
The evaluation index of experiment is mean absolute percentage error (MAPE).Evaluation index is calculated by following formula:
Y in formulaiActual value is represented,Predicted value is represented, n represents the number predicted in the data of data set;MAPE value
It is lower, it is meant that the fitting degree of prediction model is higher, has better accuracy.
3. experimentation and parameter
3.1 experimental procedure
(1) data decimation.Because the average completion time of query task in system is 3s, Gu monitor 1s collects a minor node
CPU and memory resource service condition and calculate remaining load rate, then using the server remaining load rate of calculating as original
Beginning time series.It chooses 5 minutes, the data of totally 300 time intervals, are predicted using the data of this 300 time intervals not
Carry out the remaining load rate of the server of i.e. 10 time intervals 10s Nei.Time series such as Fig. 2 institute of server remaining load rate
Show.
(2) tranquilization processing is carried out to original time series.This experiment carries out tranquilization processing using difference method.Difference
Original time series afterwards are as shown in Figure 3.
(3) ARIMA (p, d, q) model is constructed.Rank is determined using BIC criterion progress model, model accuracy can be effectively prevent
Model complexity caused by excessively high is excessively high.The model of final choice is ARIMA (1,1,1).
(4) BP neural network corrects residual error.Calculating acquires the residual sequence of the predicted value of ARIMA (1,1,1) model, then
Forecast analysis is carried out to residual error by BP neural network, the residual sequence is corrected, finally obtains ARIMA (1,1,1) prediction model
To initial predicted value be added to obtain the combined prediction value of ARIMA-BP combination forecasting with residual analysis value.
(5) LOF detects exceptional value.The abnormal coefficient of the predicted value of ARIMA (1,1,1), normal point are calculated using LOF algorithm
Using the initial predicted value of ARIMA (1,1,1) prediction model as final predicted value, abnormal point uses ARIMA-BP combined prediction
The combined prediction value of model is as final predicted value.
Above-mentioned normal point and abnormal point refers specifically to, and in the present invention, the sample point that fixed factor is influenced is as original sample
This collection is normal point, and the sample point influenced by enchancement factor is as abnormal point.The practical sample point of so stage of stable development is normal
Point, the sample point of peak period are then abnormal point.In specifically used, the predicted value of ARIMA (1,1,1) is calculated by LOF algorithm
Abnormal coefficient, normal point or abnormal point are belonged to by abnormal coefficient decision sample point.
LOF algorithm is a kind of k arest neighbors method based on local density.For a test data, its LOF exception because
Son is the ratio of its density apart from k nearest neighbours region and its own region density.When test data
When local density is similar to the density of its neighbour, LOF Outlier factor can be relatively low, which belongs to normal data;Work as survey
The local density for trying data is lower than nearest-neighbors, then LOF Outlier factor can be relatively high, which belongs to abnormal data.
LOF-ARIMA-BP model detects the initial predicted value of ARIMA model by LOF algorithm, calculates the LOF of future position
Then abnormal coefficient is compared with default LOF threshold value.If abnormal coefficient is in outside threshold value, which is abnormal point,
Final predicted value is the combined prediction value of ARIMA-BP built-up pattern;If abnormal coefficient is in threshold value, which is
Normal point, final predicted value are directly the initial predicted value of ARIMA model.
3.2 analysis of experimental results
Fig. 4 is the predicted value of three models and the comparison diagram of true value.
The MAPE of three models is calculated separately, the results are shown in Table 1:
The error comparing result of 1 three kinds of prediction models of table
It can be seen that in this experiment from the experimental result of Fig. 4 and table 1, the prediction of ARIMA-BP combination forecasting is remaining
The more single ARIMA model of the precision of prediction of load factor improves a lot, and the precision of prediction of LOF-ARIMA-BP model is compared with ARIMA-
The precision of prediction of BP combination forecasting increases again.The mean absolute percentage error (MAPE) of each model respectively from
7.92% and 6.52% is reduced to 6.01%.
For LOF-ARIMA-BP model precision of prediction be higher than ARIMA-BP combination forecasting precision of prediction original
Cause, applicant analysis obtain: its reason primarily with respect to the remaining load rate of the stage of stable development (normal point) for, ARIMA model
Predictablity rate has had reached relatively high level, if the non-linear partial anticipation component along with BP neural network can make
Predicted value is integrally bigger than normal, to reduce whole precision of prediction.Therefore, in order to ensure whole precision of prediction, LOF is being used
After algorithm rejecting abnormalities point, the final predicted value of normal point still uses the initial predicted value of ARIMA model, and abnormal point is most
Whole predicted value then uses the combined prediction value of ARIMA-BP combination forecasting, further increases prediction model of the present invention with this
Precision of prediction.
To sum up, through the invention ARIMA-BP combination forecasting or LOF-ARIMA-BP model overcomes " single
Prediction model is generally difficult to obtain complete information content, and error is larger " defect, the advantage for realizing combination forecasting is mutual
It mends, achievees the purpose that improve predictablity rate.When client uprushes a large amount of inquiry requests, load balancer can be according to prediction
Accuracy requirement uses ARIMA-BP combination forecasting or the remaining load rate of each server of LOF-ARIMA-BP model prediction, root
It is predicted that remaining load rate select suitable server distributed tasks, achieve the purpose that load balancing, to improve system money
The utilization rate in source.
Detailed description of the invention
Fig. 1 is BP neural network algorithm flow chart;
Fig. 2 is the time series of server remaining load rate;
Fig. 3 is the time series of the remaining load rate after first-order difference;
Fig. 4 is the comparison of each model predication value and true value.
Specific embodiment
The present invention is further illustrated with reference to the accompanying drawings and examples, but be not intended as to the present invention limit according to
According to.
Embodiment 1.A kind of ARIMA combination forecasting method towards block chain node load estimation carries out in the steps below:
A. according to the remaining load rate of the resource service condition calculation block chain node of block chain node, by remaining load rate
As original time series;
B. tranquilization processing is carried out to original time series;
C. rank is determined to ARIMA according to the recognition rule of time series models, constructs ARIMA model, and complete using building
ARIMA model predicted, obtain initial predicted value, and calculate the residual error of initial predicted value, construct initial predicted value
Residual sequence;
D. forecast analysis is carried out to residual error using BP neural network and obtains residual analysis value, to correct the residual sequence,
Finally it is added the initial predicted value of ARIMA model to obtain the combined prediction of ARIMA-BP combination forecasting with residual analysis value
Value.
Specifically, it in aforementioned step d, after obtaining the combined prediction value of ARIMA-BP combination forecasting, is calculated using LOF
Method calculates the abnormal coefficient of the initial predicted value of ARIMA model,
When abnormal coefficient is in default LOF threshold value, use initial predicted value as the final predicted value of model;
When abnormal coefficient is in outside default LOF threshold value, use combined prediction value as the final predicted value of model.
The verification step of default LOF threshold value:
1. assuming in training data used in the training stage, the ratio of exceptional sample data is T1;
2. the model that the hypothesis training stage obtains finally is classified as shared by abnormal data when predicting test set
Ratio is T2;
3. LOF threshold value=(T1+T2)/2.
Specifically, in aforementioned step a, the resource service condition of block chain node is the CPU and memory of block chain node
Resource service condition.Specifically, in aforementioned step a, remaining load rate calculates as follows:
If block chain node set is P={ P1,P2,…,Pn};
When multiple queries are requested while being reached, the resources left state L of current i-th node of system predictioni, and be transmitted to
Load balancer;Then have:
Li=α Lc+βLm (1)
Wherein, LcFor CPU residue, LmFor memory residue, α is CPU weight, and β is memory weight, and alpha+beta=1,1≤i≤n;
The remaining load summation of block chain node in block chain node set are as follows:
The then remaining load rate E of i-th nodeiAre as follows:
Specifically, (silent in the past period when the resource service condition of block chain node acquires in aforementioned step a
Recognize and be set as 24 hours) average completion time of query task determines frequency acquisition in system (each block chain node).
Specifically, when average completion time above-mentioned is AT seconds, the frequency acquisition is at interval of acquisition one in AT/3 seconds
The resource service condition of secondary block chain node.
Specifically, described that tranquilization processing is carried out to original time series in aforementioned step b, specifically using poor
The method divided carries out tranquilization processing to original time series.
Specifically, in aforementioned step c, the recognition rule of the time series models, specifically BIC criterion.
Specifically, in aforementioned step c, the ARIMA model is in building, using least square method or maximum likelihood
Method carries out parameter Estimation, while verifying the ARIMA model in building, the residual sequence of the ARIMA model in diagnosis building;Work as structure
When the residual sequence of ARIMA model in building is not white noise sequence, rank is determined to ARIMA again, until the ARIMA mould in building
When the residual sequence of type is white noise sequence, the building of ARIMA model is completed.
Specifically, in aforementioned step c, the ARIMA model for constructing completion is specially (1,1,1) ARIMA.
Above-mentioned ARIMA model, is one of time series models, and time series models are that one in prediction field is important
Branch, it is one group of stochastic variable according to time-sequencing, it is usually to adopt within the period of equal intervals according to given
The result that sample rate is observed certain latent process.The basic thought of ARIMA model is approximate with certain regression model
The data sequence formed over time, the capability of fitting of testing model are described, and future value is predicted by model.ARIMA
The form of model is ARIMA (p, d, q), and expression formula is as follows:
Wherein, xtIndicate the value in t moment time series;xt-1, xt-2..., xt-pThe historical data of p phase before indicating;
εt-1, εt-2..., εt-qThe prediction error of q phase before indicating;It is current random disturbances for error term;εtFor zero-mean
White noise random error series;Parameter p is auto-regressive parameter, and it is model parameter to be estimated that q, which is rolling average parameter,.
The modeling process of ARIMA is divided into four steps:
(1) tranquilization processing is carried out to former sequence.ARIMA model can be only applied in stable time series data, such as
Fruit establishes prediction model using the time series of non-stationary, it is easy to the problem of False value, i.e. original input sequence occurs
Any relationship is not present between fitting sequence, but obtains between them that there are significant correlations.
(2) according to the recognition rule of time series models, corresponding model is established.The identification of model is also referred to as determining for model
Rank determines p, the value of q.
(3) parameter Estimation and model testing.It generallys use least square method or maximum-likelihood method carries out parameter Estimation, simultaneously
Model is verified, whether diagnosis residual sequence is white noise.If residual sequence is not white noise sequence, it is meant that residual sequence is also deposited
It is not extracted in useful information, then returns to (2) step and re-recognize model.
(4) using by examine model predicted.
Above-mentioned BP neural network, i.e. reverse transmittance nerve network algorithm are a kind of for feedforward multilayer neural network
Back propagation learning algorithm.BP neural network uses multilayered structure, by input layer, one or more hidden layers and output layer group
At.Neuron between adjacent layer realizes connection completely, and the neuron in same layer does not connect.Its process is divided into two ranks
Section, the i.e. propagated forward and error back propagation of sampled data signal.
First stage (propagated forward), i.e., when there is information input BP neural network, input information passes through one by input layer
Output layer is reached after layer or multilayer hidden layer calculation processing.The activation primitive requirement of each layer can be micro- therebetween, is generally used
Sigmoid function, input therein just can reflect the response characteristic of single neuron with the relationship of output, can effectively subtract
The number of nodes of few hidden layer accelerates convergence rate and improves convergence precision.Second stage (backpropagation), i.e., if output signal
It is different from expection, that is, there is error, then error signal along former connection path return, and to the connection weight of each layer neuron into
Row is suitably modified, and obtained output reaches desired error requirements after network calculates input information.BP
Neural network algorithm flow chart is as shown in Figure 1.
Claims (10)
1. a kind of ARIMA combination forecasting method towards block chain node load estimation, which is characterized in that in the steps below into
Row:
A. according to the remaining load rate of the resource service condition calculation block chain node of block chain node, using remaining load rate as
Original time series;
B. tranquilization processing is carried out to original time series;
C. rank is determined to ARIMA according to the recognition rule of time series models, constructs ARIMA model, and completed using building
ARIMA model is predicted, initial predicted value is obtained, and calculates the residual error of initial predicted value, constructs the residual of initial predicted value
Difference sequence;
D. forecast analysis is carried out to residual error using BP neural network and obtains residual analysis value, to correct the residual sequence, finally
The initial predicted value of ARIMA model is added to obtain the combined prediction value of ARIMA-BP combination forecasting with residual analysis value.
2. the ARIMA combination forecasting method according to claim 1 towards block chain node load estimation, feature exist
In after obtaining the combined prediction value of ARIMA-BP combination forecasting, LOF algorithm being used to calculate ARIMA model in step d
The abnormal coefficient of initial predicted value,
When abnormal coefficient is in default LOF threshold value, use initial predicted value as the final predicted value of model;
When abnormal coefficient is in outside default LOF threshold value, use combined prediction value as the final predicted value of model.
3. the ARIMA combination forecasting method according to claim 1 towards block chain node load estimation, feature exist
In in step a, the resource service condition of block chain node is the CPU of block chain node and the resource service condition of memory.
4. the ARIMA combination forecasting method according to claim 3 towards block chain node load estimation, feature exist
In in step a, remaining load rate calculates as follows:
If block chain node set is P={ P1,P2,…,Pn};
When there is multiple queries to request while reaching, the resources left state L of current i-th node of system predictioni, and be transmitted to negative
Carry balanced device;Then have:
Li=α Lc+βLm
Wherein, LcFor CPU residue, LmFor memory residue, α is CPU weight, and β is memory weight, and alpha+beta=1,1≤i≤n;
The remaining load summation of block chain node in block chain node set are as follows:
The then remaining load rate E of i-th nodeiAre as follows:
5. the ARIMA combination forecasting method according to claim 4 towards block chain node load estimation, feature exist
In in step a, when the resource service condition of block chain node acquires, according to query task in system in the past period
Average completion time determines frequency acquisition.
6. the ARIMA combination forecasting method according to claim 5 towards block chain node load estimation, feature exist
In when the average completion time is AT seconds, the frequency acquisition is at interval of the block chain node of acquisition in AT/3 seconds
Resource service condition.
7. the ARIMA combination forecasting method according to claim 1 towards block chain node load estimation, feature exist
In, it is described that tranquilization processing is carried out to original time series in step b, specifically using the method for difference to original time
Sequence carries out tranquilization processing.
8. the ARIMA combination forecasting method according to claim 1 towards block chain node load estimation, feature exist
In, in step c, the recognition rule of the time series models, specifically BIC criterion.
9. the ARIMA combination forecasting method according to claim 1 towards block chain node load estimation, feature exist
In in step c, the ARIMA model carries out parameter Estimation in building, using least square method or maximum-likelihood method, together
When verifying building in ARIMA model, diagnosis building in ARIMA model residual sequence;When the ARIMA model in building
When residual sequence is not white noise sequence, rank is determined to ARIMA again, until the residual sequence of the ARIMA model in building is white
When noise sequence, the building of ARIMA model is completed.
10. the ARIMA combination forecasting method according to claim 1 towards block chain node load estimation, feature exist
In in step c, the ARIMA model for constructing completion is specially (1,1,1) ARIMA.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910420703.2A CN110297712A (en) | 2019-05-20 | 2019-05-20 | A kind of ARIMA combination forecasting method towards block chain node load estimation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910420703.2A CN110297712A (en) | 2019-05-20 | 2019-05-20 | A kind of ARIMA combination forecasting method towards block chain node load estimation |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110297712A true CN110297712A (en) | 2019-10-01 |
Family
ID=68026996
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910420703.2A Pending CN110297712A (en) | 2019-05-20 | 2019-05-20 | A kind of ARIMA combination forecasting method towards block chain node load estimation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110297712A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111209111A (en) * | 2019-12-31 | 2020-05-29 | 卓尔智联(武汉)研究院有限公司 | Resource allocation method, device, equipment and storage medium based on block chain system |
CN111414999A (en) * | 2020-04-27 | 2020-07-14 | 新智数字科技有限公司 | Method and device for monitoring running state of equipment |
CN111600967A (en) * | 2020-07-24 | 2020-08-28 | 江苏荣泽信息科技股份有限公司 | Access pressure-based load balancing system among block chain nodes |
CN113890716A (en) * | 2021-10-26 | 2022-01-04 | 中国联合网络通信集团有限公司 | Method and device for identifying pRRU and storage medium |
CN114221877A (en) * | 2021-12-14 | 2022-03-22 | 中国联合网络通信集团有限公司 | Load prediction method, device, equipment and computer readable medium |
CN115131943A (en) * | 2022-07-07 | 2022-09-30 | 杭州申昊科技股份有限公司 | Acousto-optic linkage early warning method based on deep learning |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105703954A (en) * | 2016-03-17 | 2016-06-22 | 福州大学 | Network data flow prediction method based on ARIMA model |
CN105760952A (en) * | 2016-02-15 | 2016-07-13 | 国网山东省电力公司电力科学研究院 | Load prediction method based on Kalman filtering and self-adaptive fuzzy neural network |
CN108053054A (en) * | 2017-11-22 | 2018-05-18 | 中国农业大学 | A kind of water quality prediction of river method |
US20180225605A1 (en) * | 2017-02-06 | 2018-08-09 | American Express Travel Related Services Company, Inc. | Risk assessment and alert system |
US20190139159A1 (en) * | 2017-05-18 | 2019-05-09 | Drift Marketplace, Inc. | Energy opportunity optimization system |
-
2019
- 2019-05-20 CN CN201910420703.2A patent/CN110297712A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105760952A (en) * | 2016-02-15 | 2016-07-13 | 国网山东省电力公司电力科学研究院 | Load prediction method based on Kalman filtering and self-adaptive fuzzy neural network |
CN105703954A (en) * | 2016-03-17 | 2016-06-22 | 福州大学 | Network data flow prediction method based on ARIMA model |
US20180225605A1 (en) * | 2017-02-06 | 2018-08-09 | American Express Travel Related Services Company, Inc. | Risk assessment and alert system |
US20190139159A1 (en) * | 2017-05-18 | 2019-05-09 | Drift Marketplace, Inc. | Energy opportunity optimization system |
CN108053054A (en) * | 2017-11-22 | 2018-05-18 | 中国农业大学 | A kind of water quality prediction of river method |
Non-Patent Citations (2)
Title |
---|
孙凯俐等: "面向区块链节点负载预测的ARIMA组合预测方法", 《电子技术与软件工程》 * |
高兴旺等: "基于混合预测模型的负载均衡算法研究", 《计算机工程与设计》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111209111A (en) * | 2019-12-31 | 2020-05-29 | 卓尔智联(武汉)研究院有限公司 | Resource allocation method, device, equipment and storage medium based on block chain system |
CN111414999A (en) * | 2020-04-27 | 2020-07-14 | 新智数字科技有限公司 | Method and device for monitoring running state of equipment |
CN111414999B (en) * | 2020-04-27 | 2023-08-22 | 新奥新智科技有限公司 | Method and device for monitoring running state of equipment |
CN111600967A (en) * | 2020-07-24 | 2020-08-28 | 江苏荣泽信息科技股份有限公司 | Access pressure-based load balancing system among block chain nodes |
CN111600967B (en) * | 2020-07-24 | 2020-11-06 | 江苏荣泽信息科技股份有限公司 | Access pressure-based load balancing system among block chain nodes |
CN113890716A (en) * | 2021-10-26 | 2022-01-04 | 中国联合网络通信集团有限公司 | Method and device for identifying pRRU and storage medium |
CN114221877A (en) * | 2021-12-14 | 2022-03-22 | 中国联合网络通信集团有限公司 | Load prediction method, device, equipment and computer readable medium |
CN115131943A (en) * | 2022-07-07 | 2022-09-30 | 杭州申昊科技股份有限公司 | Acousto-optic linkage early warning method based on deep learning |
CN115131943B (en) * | 2022-07-07 | 2023-10-31 | 杭州申昊科技股份有限公司 | Acousto-optic linkage early warning method based on deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110297712A (en) | A kind of ARIMA combination forecasting method towards block chain node load estimation | |
Hu et al. | Web service recommendation based on time series forecasting and collaborative filtering | |
CN106952159B (en) | Real estate collateral risk control method, system and storage medium | |
US20120179638A1 (en) | Relative variable selection system and selection method thereof | |
CN114721833B (en) | Intelligent cloud coordination method and device based on platform service type | |
CN105989441A (en) | Model parameter adjustment method and device | |
CN101771758A (en) | Dynamic determine method for normal fluctuation range of performance index value and device thereof | |
CN110837915B (en) | Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning | |
CN114896899B (en) | Multi-agent distributed decision method and system based on information interaction | |
CN115345212A (en) | Control method and device based on machine learning model | |
CN113254719B (en) | Online social network information propagation method based on status theory | |
CN111666494A (en) | Clustering decision model generation method, clustering processing method, device, equipment and medium | |
CN112329997A (en) | Power demand load prediction method and system, electronic device, and storage medium | |
CN116562514B (en) | Method and system for immediately analyzing production conditions of enterprises based on neural network | |
CN111598457A (en) | Method and device for determining quality of power wireless network | |
CN112488404B (en) | Multithreading efficient prediction method and system for large-scale power load of power distribution network | |
CN110968512A (en) | Software quality evaluation method, device, equipment and computer readable storage medium | |
Peng et al. | Demand modeling and forecasting | |
CN116245379A (en) | Index weight optimization method and device for power grid construction engineering effect evaluation | |
CN114462506B (en) | Communication network auxiliary planning method supporting preference strategy and application thereof | |
CN114971272A (en) | Medium-voltage distribution network gridding planning evaluation method and system | |
Chen | Construction and Application of Forecast Model for Eco-Tourism Market Demand Based on BP Neural Network | |
CN109978038A (en) | A kind of cluster abnormality determination method and device | |
Chang et al. | Fuzzy back-propagation network for PCB sales forecasting | |
CN110766426A (en) | Electricity price making method and device |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191001 |