CN109816158A - Combined method, device, equipment and the readable storage medium storing program for executing of prediction model - Google Patents
Combined method, device, equipment and the readable storage medium storing program for executing of prediction model Download PDFInfo
- Publication number
- CN109816158A CN109816158A CN201910014124.8A CN201910014124A CN109816158A CN 109816158 A CN109816158 A CN 109816158A CN 201910014124 A CN201910014124 A CN 201910014124A CN 109816158 A CN109816158 A CN 109816158A
- Authority
- CN
- China
- Prior art keywords
- prediction model
- prediction
- data
- data set
- built
- 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 invention discloses a kind of combined method of prediction model, device, equipment and readable storage medium storing program for executing, the method comprising the steps of: after obtaining prediction model by training dataset training, obtain test data set, the test data set is input in each prediction model, the corresponding output result of each prediction model is obtained;The first predictablity rate of each prediction model is calculated according to the corresponding legitimate reading of the test data set and the output result;Prediction model of at least two first predictablity rates greater than preset threshold is chosen as object module, combines the object module according to preset combination, carries out data prediction to obtain prediction built-up pattern.The present invention realizes the combination of prediction model, is predicted by built-up pattern data, avoid the limitation and unicity of single prediction model by intelligent decision, improves the accuracy rate of data prediction.
Description
Technical field
The present invention relates to intelligent Decision Technology field more particularly to a kind of combined method of prediction model, device, equipment and
Readable storage medium storing program for executing.
Background technique
Prediction model refers to the quantitative relation for prediction, between the things described in mathematical linguistics or formula, it
The inherent law between things is disclosed to a certain extent.Current prediction model is being adopted based on single prediction model buildings
During being predicted with single prediction model data, due to the limitation and unicity of single prediction model, lead to single prediction
Model logarithm it is predicted that accuracy rate it is low.
Summary of the invention
The main purpose of the present invention is to provide a kind of combined method of prediction model, device, equipment and readable storage mediums
Matter, it is intended to solve existing single prediction model logarithm it is predicted that the low technical problem of accuracy rate.
To achieve the above object, the present invention provides a kind of combined method of prediction model, the combination side of the prediction model
Method comprising steps of
After obtaining prediction model by training dataset training, test data set is obtained, the test data set is defeated
Enter into each prediction model, obtains the corresponding output result of each prediction model;
Each prediction model is calculated according to the corresponding legitimate reading of the test data set and the output result
First predictablity rate;
Prediction model of at least two first predictablity rates greater than preset threshold is chosen as object module, according to
Preset combination combines the object module, carries out data prediction to obtain prediction built-up pattern.
Further, the prediction model conduct chosen at least two first predictablity rates and be greater than preset threshold
Object module combines the object module according to preset combination, carries out data prediction to obtain prediction built-up pattern
After step, further includes:
After getting data set to be predicted, the data volume of the data set to be predicted is calculated;
When determining the data volume less than the first preset quantity, the data set to be predicted is input to single prediction model
In, obtain prediction result;
It, will when determining that the data volume is more than or equal to first preset quantity, and when less than the second preset quantity
The data set to be predicted is input in the prediction built-up pattern obtained by weighted average combination, obtains prediction result;
When determining that the data volume is more than or equal to second preset quantity, the data set to be predicted is inputted
Into the prediction built-up pattern combined by integration mode, prediction result is obtained.
Further, described to work as the determining data volume more than or equal to first preset quantity, and less than second
When preset quantity, the data set to be predicted is input in the prediction built-up pattern obtained by weighted average combination, is obtained
The step of prediction result includes:
When determining that the data volume is more than or equal to first preset quantity, and when less than the second preset quantity, inspection
Survey whether the data concentrated to training data are time series data;
If the data concentrated to training data are time series data, the prediction obtained by weighted average combination is detected
Whether built-up pattern is made of time series data prediction model and scattered data being prediction model;
If detecting, the prediction built-up pattern is made of time series data prediction model and scattered data being prediction model,
When the weight for detecting the time series data prediction model is less than or equal to the weight of the scattered data being prediction model, adjust
The whole time series data prediction model and the corresponding weight of the scattered data being prediction model, so that the time series data predicts mould
The weight of type is greater than the weight of the scattered data being prediction model, the prediction built-up pattern after being adjusted weight;
By in the prediction built-up pattern after the data set input adjustment weight to be predicted, prediction result is obtained.
Further, described when determining the data volume less than the first preset quantity, the data set to be predicted is defeated
The step of entering into single prediction model, obtaining prediction result include:
When determining the data volume less than the first preset quantity, the dimension of the data intensive data to be predicted is calculated
Number;
If the number of dimensions is less than default number of dimensions, the data set to be predicted is input to first kind single prediction model
In, obtain prediction result;
If the number of dimensions is more than or equal to the default number of dimensions, the data set to be predicted is input to second
In class single prediction model, prediction result is obtained.
Further, the prediction model conduct chosen at least two first predictablity rates and be greater than preset threshold
Object module combines the object module according to preset combination, carries out data prediction to obtain prediction built-up pattern
After step, further includes:
When detecting the presence of the objective cross mode of at least corresponding two predictions built-up pattern, the objective cross is calculated
In mode, corresponding second predictablity rate of each prediction built-up pattern;
The corresponding prediction built-up pattern of maximum second predictablity rate is chosen as default built-up pattern, and is being got
After data set to be predicted, the data set to be predicted is input in default built-up pattern, prediction result is obtained.
Further, described that each institute is calculated according to the corresponding legitimate reading of the test data set and the output result
The step of stating the first predictablity rate of prediction model include:
After obtaining output result legitimate reading corresponding with the test data set is got, detect whether exist
The prediction model of multiple output results;
The prediction model of multiple output results if it exists then calculates the corresponding output of prediction model there are multiple output results
As a result in, fruiting quantities identical with the legitimate reading;
It is corresponding to obtain that there are multiple output results by the corresponding total quantity divided by the legitimate reading of the fruiting quantities
First predictablity rate of prediction model.
Further, described to obtain output result legitimate reading corresponding with the test data set is got
Afterwards, detect whether there are it is multiple output results prediction model the step of after, further includes:
If it does not exist it is multiple output results prediction models, then detect each prediction model output result whether be
Character string;
If detecting, the output result is character string, calculates the phase between the output result and the legitimate reading
Like degree, using the similarity as the first predictablity rate of corresponding prediction model.
In addition, to achieve the above object, the present invention also provides a kind of combination unit of prediction model, the prediction model
Combination unit includes:
Module is obtained, for obtaining test data set after obtaining prediction model by training dataset training;
Input module obtains each prediction mould for the test data set to be input in each prediction model
The corresponding output result of type;
Computing module, for calculating each institute according to the corresponding legitimate reading of the test data set and the output result
State the first predictablity rate of prediction model;
Module is chosen, the prediction model conduct for being greater than preset threshold for choosing at least two first predictablity rates
Object module;
Composite module, for combining the object module according to preset combination, with obtain prediction built-up pattern into
Line number it is predicted that.
In addition, to achieve the above object, the present invention also provides a kind of unit equipment of prediction model, the prediction model
Unit equipment includes memory, processor and is stored in the prediction model that can be run on the memory and on the processor
Combinator, the combinator of the prediction model realizes the group of prediction model as described above when being executed by the processor
The step of conjunction method.
In addition, to achieve the above object, it is described computer-readable the present invention also provides a kind of computer readable storage medium
It is stored with the combinator of prediction model on storage medium, is realized such as when the combinator of the prediction model is executed by processor
The step of combined method of the upper prediction model.
The present invention by the way that after obtaining prediction model, acquired test data set is input in each prediction model,
The first prediction of each prediction model is calculated according to the output result of each prediction model legitimate reading corresponding with test data set
Accuracy rate chooses prediction model of at least two first predictablity rates greater than preset threshold as object module, according to prediction
Combination composite object model, carry out number it was predicted that realize the combination of prediction model to obtain prediction built-up pattern, lead to
It crosses built-up pattern to predict data, avoids the limitation and unicity of single prediction model, improve the accurate of data prediction
Rate.
Detailed description of the invention
Fig. 1 is the flow diagram of the combined method first embodiment of prediction model of the present invention;
Fig. 2 is the flow diagram of the combined method second embodiment of prediction model of the present invention;
Fig. 3 is the functional schematic module map of the combination unit preferred embodiment of prediction model of the present invention;
Fig. 4 is the structural schematic diagram for the hardware running environment that the embodiment of the present invention is related to.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
The present invention provides a kind of combined method of prediction model, and referring to Fig.1, Fig. 1 is the combination side of prediction model of the present invention
The flow diagram of method first embodiment.
The embodiment of the invention provides the embodiments of the combined method of prediction model, it should be noted that although in process
Logical order is shown in figure, but in some cases, it can be to be different from shown or described by sequence execution herein
Step.
The combined method of prediction model is applied to server and is perhaps equipped with prediction in the server or terminal in terminal
System, terminal may include such as mobile phone, tablet computer, laptop, palm PC, personal digital assistant (Personal
Digital Assistant, PDA) etc. the fixed terminals such as mobile terminals, and number TV, desktop computer.In prediction mould
In each embodiment of the combined method of type, for ease of description, omits executing subject and be illustrated each embodiment.Predict mould
The combined method of type includes:
Step S10 obtains test data set, by the test after obtaining prediction model by training dataset training
Data set is input in each prediction model, obtains the corresponding output result of each prediction model.
After getting training dataset, using training dataset as the input of prediction model, training dataset is corresponding
Output of the data result as machine learning model and/or deep learning model obtains prediction model with training.Wherein, machine
Learning model includes but is not limited to support vector machines (SVM, Support VectorMachine), naive Bayesian (NB, Naive
Bayes), k nearest neighbor (KNN, k-Nearest Neighbor), decision tree (DT, Decision tree) and integrated model are (random
Forest RF/ gradient promoted decision tree GDBT), deep learning model include but is not limited to convolutional neural networks (CNN,
Convolutional NeuralNetwork), Recognition with Recurrent Neural Network (RNN, Recurrent Neural Networks) and pass
Return neural network (RNN, Recursive Neural Networks).
After obtaining prediction model, test data set is obtained, test data set is input in each prediction model, is obtained
The corresponding output result of each prediction model.It should be noted that test data set and training dataset are same kind of numbers
According to such as when training dataset is temperature data, test data set is also temperature data;When test data set is rainfall product data
When, test data set is also rainfall product data.Training dataset and the corresponding time dimension of test data set be it is different, i.e.,
Training dataset and test data set are the data got in different time sections.The data volume and test number that training data is concentrated
Can be identical according to the data volume of concentration, it can not also be identical.
Step S20 is calculated each described pre- according to the corresponding legitimate reading of the test data set and the output result
Survey the first predictablity rate of model.
After getting test data set, the corresponding legitimate reading of test data set is obtained.Such as on October 1st, 2018 is arrived
Daily mean temperature is input in prediction model in this 10 days October 10 in 2018, if the output result of prediction model is
Daily prediction mean temperature in this 10 days on October 20, of 11 days to 2018 October in 2018, then test data set is corresponding true
In fact the result is that true mean temperature daily in this 10 days on October 20, of 11 days to 2018 October in 2018.
After getting the corresponding legitimate reading of test data set and output result, according to the legitimate reading of test data set
The predictablity rate of each prediction model is calculated with output result.In order to by the predictablity rate and combined prediction of single prediction model
The predictablity rate of model distinguishes, each prediction that will be calculated according to legitimate reading and output result in the embodiment of the present invention
The predictablity rate of model is denoted as the first predictablity rate.It should be understood that since the characteristic of each prediction model is different, because
This, the corresponding output result of each prediction model may be the same, it is also possible to different.
Further, step S20 includes:
Step a, after obtaining output result legitimate reading corresponding with the test data set is got, detection is
It is no that there are the prediction models of multiple output results.
Specifically, the output of each prediction model is being obtained as a result, and getting the corresponding true knot of test data set
After fruit, detect whether that there are the prediction models of multiple output results.It should be noted that when test data set is input to prediction
After in model, according to the different of prediction model realization of functions, the output result of prediction model be also it is different, such as when pre-
Surveying model is when predicting mean temperature daily in one week following, and the output result of prediction model is 7, that is, exports 7 and put down
Equal temperature;When prediction model is the semanteme that user to be predicted inputs sentence, the output result of prediction model one shows sentence
A subsemantic character string.
Step b, if it exists it is multiple output results prediction models, then calculate there are it is multiple output results prediction models pair
It should export in result, fruiting quantities identical with the legitimate reading.
If detecting the presence of the prediction model of multiple output results, the prediction model that multiple output results will be present is denoted as
Target prediction model, and the output result fruiting quantities identical with legitimate reading for calculating target prediction model.
Step c, it is corresponding to obtain that there are multiple outputs by the corresponding total quantity divided by the legitimate reading of the fruiting quantities
First predictablity rate of prediction model as a result.
It is after the output result for calculating target prediction model fruiting quantities identical with legitimate reading, each target is pre-
The corresponding fruiting quantities of model are surveyed divided by the total quantity of legitimate reading, it is corresponding to obtain that there are the prediction models of multiple output results
First predictablity rate.It is understood that the output result due to each prediction model may be different, it is each pre-
It surveys in the output result of model, fruiting quantities identical with legitimate reading may also be different, eventually results in each prediction mould
First predictablity rate of type is also different.The quantity of legitimate reading and the quantity of output result are consistent.It is as pre- in worked as some
When the output result for surveying model is 8, if there are 6 results are identical as legitimate reading in output result, the prediction model
First predictablity rate are as follows: 6 ÷ 8=0.75.
Further, step S20 further include:
Step d, the prediction model of multiple output results, then detect the output result of each prediction model if it does not exist
It whether is character string.
If detect there is no it is multiple output results prediction models, detect each prediction model output result whether
It is character string.It is understood that the number of characters at least two in character string.When there is no the prediction moulds of multiple output results
When type, then show each prediction model only exists an output as a result, to calculate predictablity rate by character string applicable
In there are the prediction models that one exports result.
Step e, if detecting, the output result is character string, calculate the output result and the legitimate reading it
Between similarity, using the similarity as the first predictablity rate of corresponding prediction model.
If detecting, output result is character string, the similarity between output result and legitimate reading is calculated, by the phase
The first predictablity rate like degree as corresponding prediction model.Specifically, if output result is a text sentence, can pass through
Hamming distance calculates the similarity between output result and legitimate reading, and the similarity is pre- as correspond to prediction model first
Survey accuracy rate;If output result is the character string of a string of codes, English character and/or number composition, output result can be calculated
It is with the character number of identical characters in legitimate reading, the character number of identical characters is a divided by total character in legitimate reading
Number obtains the similarity between output result and legitimate reading to get the first predictablity rate of corresponding prediction model is arrived.
Step S30 chooses at least two first predictablity rates and is greater than the prediction model of preset threshold as target
Model combines the object module according to preset combination, carries out data prediction to obtain prediction built-up pattern.
After the first predictablity rate of each prediction model is calculated, chooses the first predictablity rate and be greater than default threshold
The prediction model of value, wherein preset threshold is arranged according to specific needs, does not limit preset threshold in embodiments of the present invention
Specific size.It is elected to after getting the first predictablity rate greater than the prediction model of preset threshold, in selected prediction model
It is middle to select at least two prediction models as object module, and according to preset combination composite object model, it is pre- to obtain
It surveys built-up pattern and carries out data prediction.The combination of prediction includes but is not limited to weighted average mode and integration mode.Specifically
Ground, weighted average mode is that corresponding weight is arranged for each submodel in prediction built-up pattern, by the output of each submodel
As a result multiplied by being added after respective weights, final output result is obtained;Integration mode is the son that will be predicted in built-up pattern
Input of the corresponding output result of model as other submodel, that is, each submodel string in built-up pattern will be predicted
Connection gets up.Further, preset combination further includes Kalman filtering mode and mode error, and error combination is with EWA
Model prediction result and the difference of legitimate reading realize error combination model as pre- target.
Below by weighted average mode obtain prediction built-up pattern for be illustrated: as when there are 4 object module A,
B, when C and D, by A and B, A and C, A and D, A, B, C and D, A, B and C, A, B and D, C and B, B and D, C and D, B, C and D are added
Weight average combination, in being weighted and averaged anabolic process, adjusts the weight of each object module, is input to until by test set
In model after combination, under the weight, corresponding predictablity rate is that the corresponding accuracy rate of all weights is highest, at this time should
Model after combination is the prediction built-up pattern obtained according to object module.
The present embodiment is by being input to each prediction model for acquired test data set after obtaining prediction model
In, the first pre- of each prediction model is calculated according to the output result of each prediction model legitimate reading corresponding with test data set
Accuracy rate is surveyed, chooses prediction model of at least two first predictablity rates greater than preset threshold as object module, according to pre-
The combination composite object model of survey carries out several it was predicted that realizing the combination of prediction model to obtain prediction built-up pattern,
Data are predicted by built-up pattern, avoid the limitation and unicity of single prediction model, improve the standard of data prediction
True rate.
Further, the combined method second embodiment of prediction model of the present invention is proposed.
The combined method first embodiment of the combined method second embodiment and prediction model of the prediction model
Difference is, referring to Fig. 2, the combined method of prediction model further include:
Step f calculates the data volume of the data set to be predicted after getting data set to be predicted.
After obtaining prediction built-up pattern, detect whether to get data set to be predicted.If getting data set to be predicted,
Then calculate the data volume of data set to be predicted.The data volume of data set to be predicted is arranged according to specific needs, as some is waited for
The data volume of predictive data set can be 20,100 or 500 etc..
The data set to be predicted is input to single pre- by step g when determining the data volume less than the first preset quantity
It surveys in model, obtains prediction result.
After the data volume of data set to be predicted is calculated, judge data volume whether less than the first preset quantity.When true
When the data volume of fixed data set to be predicted is less than the first preset quantity, data set to be predicted is input in single prediction model, is obtained
To prediction result, it is to be understood that the output of single prediction model is prediction result.Such as in the data volume of data set to be predicted
When less than the first preset quantity, the data in data set to be predicted are inputted in the prediction models such as SVM or RNN.Wherein, this implementation
Example is not particularly limited the size of the first preset quantity.
Further, step g includes:
Step g1 calculates the data intensive data to be predicted when determining the data volume less than the first preset quantity
Number of dimensions.
When determining the data volume of data set to be predicted less than the first preset quantity, data intensive data to be predicted is calculated
Number of dimensions.It include daily mean temperature, mean temperature weekly, the side of all temperature if some weather data is concentrated
Difference and mean value, then can determine that the weather data concentrates the number of dimensions of temperature is 4, that is, there are 4 kinds of forms to indicate temperature.
The data set to be predicted is input to first kind list if the number of dimensions is less than default number of dimensions by step g2
In prediction model, prediction result is obtained.
After the number of dimensions that data intensive data to be predicted is calculated, judge whether the number of dimensions is less than default dimension
Number.If it is determined that the number of dimensions is less than default number of dimensions, then data set to be predicted is input in first kind single prediction model, is obtained
Prediction result.Wherein, default number of dimensions may be configured as 3,4 or 7 etc..
Step g3, it is if the number of dimensions is more than or equal to the default number of dimensions, the data set to be predicted is defeated
Enter into the second class single prediction model, obtains prediction result.
If it is determined that number of dimensions is more than or equal to default number of dimensions, then data set to be predicted is input to the second class single prediction
In model, prediction result is obtained.Wherein, relative to first kind single prediction model, the second class single prediction model is more applicable and data
Measure big data set.First kind single prediction model and the second class single prediction model are pre-set, such as first kind single prediction
Model is SVM and SARIMA (SeasonalAutoregressive Integrated Moving Average, seasonal difference
Autoregressive moving-average model) etc. simple prediction model, the second class single prediction model be xgboost (eXtreme
Gradient Boosting), LSTM (Long Short-Term Memory, shot and long term memory network) etc. it is more complicated pre-
Survey model.
Step h, when determining that the data volume is more than or equal to first preset quantity, and less than the second preset quantity
When, the data set to be predicted is input in the prediction built-up pattern obtained by weighted average combination, prediction result is obtained.
After determining that data volume is more than or equal to the first preset quantity, judge whether the data volume to training dataset is small
In the second preset quantity.It is understood that the second preset quantity is greater than the first preset quantity.When determining to training dataset
When data volume is less than the second preset quantity, data set to be predicted is input to the prediction combination die obtained by weighted average combination
In type, prediction result is obtained.
Further, in order to improve to obtain the accuracy rate of prediction result, step h includes:
Step h1, when determining that the data volume is more than or equal to first preset quantity, and less than the second present count
When amount, whether the detection data concentrated to training data are time series data.
When the data volume for determining data set to be predicted is more than or equal to the first preset quantity, and less than the second preset quantity
When, detect whether the data concentrated to training data are time series data.Wherein, time series data is regular data, opposite,
Scattered data being is random data.
Step h2, if the data concentrated to training data are time series data, detection is obtained by weighted average combination
To prediction built-up pattern whether be made of time series data prediction model and scattered data being prediction model.
If detecting, the data concentrated to training data are time series data, and detection is predicted by weighted average combination
Whether built-up pattern is made of time series data prediction model and scattered data being prediction model.In embodiments of the present invention, ordinal number when
It is predicted that model carries timing mark, scattered data being prediction model carries mark at random, therefore, is carried by prediction model
Mark i.e. can determine whether prediction group molding type is made of time series data prediction model and scattered data being prediction model.It can manage
Solution, time series data prediction model are that have preferable prediction effect to time series data, and scattered data being prediction model is at random
The specific preferable prediction effect of data.Further, if detect the prediction built-up pattern that weighted average combination obtains be not by
Time series data model and scattered data being prediction model composition, then be directly input to prediction combination for the data concentrated to training data
In model, prediction result is obtained.
Step h3, if detecting, the prediction built-up pattern is by time series data prediction model and scattered data being prediction model
Composition is then less than or equal to the power of the scattered data being prediction model in the weight for detecting the time series data prediction model
When weight, the time series data prediction model and the corresponding weight of the scattered data being prediction model are adjusted, so that ordinal number when described
It is predicted that the weight of model is greater than the weight of the scattered data being prediction model, the prediction built-up pattern after being adjusted weight.
If detecting, prediction built-up pattern is made of time series data prediction model and scattered data being prediction model, is detected
Whether the weight of time series data prediction model is less than or equal to the weight of scattered data being prediction model.If it is pre- to detect time series data
The weight for surveying model is less than or equal to the weight of scattered data being prediction model, then improves the weight of time series data prediction model,
The weight of scattered data being prediction model is reduced, so that the weight of time series data prediction model is greater than the power of scattered data being prediction model
Weight, the prediction built-up pattern after being adjusted weight.It should be noted that the increase rate of time series data prediction model weight and
The reduction amplitude of scattered data being prediction model weight can be preset.
Further.If detecting, the weight of time series data prediction model is greater than the weight of scattered data being prediction model,
Directly the data in data set to be predicted are input in prediction built-up pattern, obtain prediction result.
Step h4 obtains prediction result in the prediction built-up pattern after the data set input adjustment weight to be predicted.
After prediction built-up pattern after being adjusted weight, after the data input adjustment weight in data set to be predicted
Prediction built-up pattern in, obtain prediction result.
It is understood that predicting that built-up pattern is detecting when the data of data set to be predicted are scattered data being
It is made of time series data prediction model and scattered data being prediction model, and the weight of time series data prediction model is more than or equal to
When the weight of scattered data being prediction model, it is also desirable to adjust the power between time series data prediction model and scattered data being prediction model
Weight makes the weight of time series data prediction model be less than the weight of scattered data being prediction model, improves the accurate of gained prediction result
Rate.
Step i, when determining that the data volume is more than or equal to second preset quantity, by the data to be predicted
Collection is input in the prediction built-up pattern combined by integration mode, obtains prediction result.
When determining that data volume is more than or equal to the second preset data amount, data set to be predicted is input to by integrated
In the prediction built-up pattern that mode combines, prediction result is obtained.It is understood that when combining to obtain by integration mode
Prediction built-up pattern when having multiple, data set to be predicted can be arbitrarily input in one of prediction built-up pattern.
Further, when determining that data volume is more than or equal to the first preset quantity, and less than the second preset quantity, or
When determining that data volume is more than or equal to the second preset quantity, it can also be divided according to the number of dimensions of data intensive data to be predicted
With corresponding prediction built-up pattern, specifically process and step g1, step g2 is similar with step g3, and in this not go into detail.
It further, can also will be to be predicted according to the number of dimensions of data set to be predicted after getting data set to be predicted
Data set inputs in corresponding prediction built-up pattern, obtains prediction result.Further, it can also be determined according to the size of data volume
It is to be input to data set to be predicted in the prediction built-up pattern combined by mode error, or by data set to be predicted
Category is input in the prediction built-up pattern combined by Kalman filtering mode.The present embodiment passes through according to data to be predicted
The data volume of concentration realizes the number according to data set to be predicted in the prediction model of data set to be predicted input corresponding types
According to amount smart allocation prediction model, on the basis of the accuracy rate of the prediction result obtained by guarantee, the effect for obtaining prediction result is improved
Rate.
Further, the combined method 3rd embodiment of prediction model of the present invention is proposed.
The combined method 3rd embodiment of the prediction model and the combined method first or second of the prediction model are real
The difference for applying example is, referring to Fig. 2, the combined method of prediction model further include:
Step S40, when detecting the presence of the objective cross mode of at least corresponding two predictions built-up pattern, described in calculating
In objective cross mode, corresponding second predictablity rate of each prediction built-up pattern.
When obtaining prediction built-up pattern, the objective cross side that there is at least corresponding two predictions built-up pattern is detected whether
Formula.If detecting the presence of objective cross mode, calculate in objective cross mode, the corresponding prediction of each prediction built-up pattern is quasi-
True rate, and the corresponding predictablity rate of prediction built-up pattern is denoted as the second predictablity rate.It should be noted that calculating pre-
It is identical as the principle for calculating corresponding first predictablity rate of single prediction model to survey corresponding second predictablity rate of built-up pattern,
In this not go into detail.
Step S50 chooses the corresponding prediction built-up pattern of maximum second predictablity rate as default built-up pattern, and
After getting data set to be predicted, the data set to be predicted is input in default built-up pattern, prediction result is obtained.
After corresponding second predictablity rate of each prediction built-up pattern in objective cross mode is calculated, is determined
Maximum value in two predictablity rates is combined the corresponding prediction built-up pattern of the second predictablity rate of maximum value as default
Model, and after getting data set to be predicted, data set to be predicted is input in default built-up pattern, prediction knot is obtained
Fruit.Further, after getting data set to be predicted, the corresponding user of data set to be predicted can also manually select prediction group
Data set to be predicted is input in the prediction built-up pattern of user's selection by molding type.
It further, can be corresponding according to prediction built-up pattern by data set input prediction built-up pattern to be predicted
The height of second predictablity rate, from high to low prediction model after combining select the prediction built-up pattern of certain data as
Final prediction built-up pattern.
The present embodiment according to corresponding second predictablity rate of prediction built-up pattern in prediction built-up pattern by determining
Default built-up pattern, after getting data set to be predicted, data set to be predicted is input in default built-up pattern, is obtained pre-
Data set to be predicted is preferentially input in default built-up pattern by survey as a result, in order to after getting data set to be predicted, into
Improve to one step the accuracy rate of gained prediction result.
In addition, the present invention also provides a kind of combination unit of prediction model, the group of the prediction model attaches together referring to Fig. 3
It sets and includes:
Module 10 is obtained, for obtaining test data set after obtaining prediction model by training dataset training;
Input module 20 obtains each prediction for the test data set to be input in each prediction model
The corresponding output result of model;
Computing module 30, it is each for being calculated according to the corresponding legitimate reading of the test data set and the output result
First predictablity rate of the prediction model;
Module 40 is chosen, is made for choosing at least two first predictablity rates greater than the prediction model of preset threshold
For object module;
Composite module 50, for combining the object module according to preset combination, to obtain prediction built-up pattern
Carry out data prediction.
Further, the computing module 30 is also used to after getting data set to be predicted, calculates the number to be predicted
According to the data volume of collection;
The input module 20 is also used to when determining the data volume less than the first preset quantity, by the number to be predicted
It is input in single prediction model according to collection, obtains prediction result;When determining that it is default that the data volume is more than or equal to described first
Quantity, and when less than the second preset quantity, the data set to be predicted is input to the prediction obtained by weighted average combination
In built-up pattern, prediction result is obtained;It, will be described when determining that the data volume is more than or equal to second preset quantity
Data set to be predicted is input in the prediction built-up pattern combined by integration mode, obtains prediction result.
Further, the input module 20 includes:
First detection unit for being more than or equal to first preset quantity when the determining data volume, and is less than
When the second preset quantity, whether the detection data concentrated to training data are time series data;If described to training dataset
In data be time series data, then whether the prediction built-up pattern that is obtained by weighted average combination of detection is predicted by time series data
Model and scattered data being prediction model composition;
Adjustment unit, if for detecting that the prediction built-up pattern is pre- by time series data prediction model and scattered data being
Model composition is surveyed, then is less than or equal to the scattered data being in the weight for detecting the time series data prediction model and predicts mould
When the weight of type, the time series data prediction model and the corresponding weight of the scattered data being prediction model are adjusted, so that described
The weight of time series data prediction model is greater than the weight of the scattered data being prediction model, the prediction combination after being adjusted weight
Model;
Input unit, for obtaining in the prediction built-up pattern after the data set input adjustment weight to be predicted pre-
Survey result.
Further, the input module 20 includes:
First computing unit, for calculating the number to be predicted when determining the data volume less than the first preset quantity
According to the number of dimensions of intensive data;
If the input unit, which is also used to the number of dimensions, is less than default number of dimensions, the data set to be predicted is inputted
Into first kind single prediction model, prediction result is obtained;It, will if the number of dimensions is more than or equal to the default number of dimensions
The data set to be predicted is input in the second class single prediction model, obtains prediction result.
Further, the computing module 30 is also used to when the mesh for detecting the presence of at least corresponding two predictions built-up pattern
It when marking combination, calculates in the objective cross mode, corresponding second predictablity rate of each prediction built-up pattern;
The selection module 40 is also used to choose the corresponding prediction built-up pattern of maximum second predictablity rate as silent
Recognize built-up pattern;
The input module 20 is also used to after getting data set to be predicted, the data set to be predicted is input to silent
Recognize in built-up pattern, obtains prediction result.
Further, the computing module 30 includes:
Second detection unit, for obtaining the output result true knot corresponding with the test data set is got
After fruit, detect whether that there are the prediction models of multiple output results;
Second computing unit, for the prediction model of multiple output results if it exists, then there are multiple output results for calculating
The corresponding output result of prediction model in, fruiting quantities identical with the legitimate reading;By the fruiting quantities it is corresponding divided by
The total quantity of the legitimate reading, it is corresponding to obtain that there are the first predictablity rates of the prediction models of multiple output results.
Further, the second detection unit is also used to the prediction model of multiple output results if it does not exist, then detects
Whether the output result of each prediction model is character string;
If second computing unit is also used to detect that the output result is character string, the output result is calculated
With the similarity between the legitimate reading, using the similarity as the first predictablity rate of corresponding prediction model.
It should be noted that each embodiment of the combination unit of prediction model and the combined method of above-mentioned prediction model
Each embodiment is essentially identical, and in this not go into detail.
In addition, the present invention also provides a kind of unit equipments of prediction model.As shown in figure 4, Fig. 4 is embodiment of the present invention side
The structural schematic diagram for the hardware running environment that case is related to.
It should be noted that Fig. 4 can be the structural schematic diagram of the hardware running environment of the unit equipment of prediction model.This
The unit equipment of inventive embodiments prediction model can be PC, the terminal devices such as portable computer.
As shown in figure 4, the unit equipment of the prediction model may include: processor 1001, such as CPU, memory 1005,
User interface 1003, network interface 1004, communication bus 1002.Wherein, communication bus 1002 is for realizing between these components
Connection communication.User interface 1003 may include display screen (Display), input unit such as keyboard (Keyboard), can
Selecting user interface 1003 can also include standard wireline interface and wireless interface.Network interface 1004 optionally may include mark
Wireline interface, the wireless interface (such as WI-FI interface) of standard.Memory 1005 can be high speed RAM memory, be also possible to stablize
Memory (non-volatile memory), such as magnetic disk storage.Memory 1005 optionally can also be independently of preceding
State the storage device of processor 1001.
Optionally, the unit equipment of prediction model can also include camera, RF (Radio Frequency, radio frequency) electricity
Road, sensor, voicefrequency circuit, WiFi module etc..
It will be understood by those skilled in the art that the unit equipment structure of prediction model shown in Fig. 4 is not constituted to pre-
The restriction for surveying the unit equipment of model may include perhaps combining certain components or not than illustrating more or fewer components
Same component layout.
As shown in figure 4, as may include that operating system, network are logical in a kind of memory 1005 of computer storage medium
Believe the combinator of module, Subscriber Interface Module SIM and prediction model.Wherein, operating system manages and controls prediction model
The program of unit equipment hardware and software resource supports the operation of the combinator and other softwares or program of prediction model.
In the unit equipment of prediction model shown in Fig. 4, user interface 1003 can be used for receiving training dataset, test
Data set and to training dataset etc.;Network interface 1004 is mainly used for background server, and it is logical to carry out data with background server
Letter;Processor 1001 can be used for calling the combinator of the prediction model stored in memory 1005, and execute as described above
Prediction model combined method the step of.
The unit equipment specific embodiment of prediction model of the present invention and each embodiment of combined method of above-mentioned prediction model
Essentially identical, details are not described herein.
In addition, the embodiment of the present invention also proposes a kind of computer readable storage medium, the computer readable storage medium
On be stored with the combinator of prediction model, realized when the combinator of the prediction model is executed by processor as described above
The step of combined method of prediction model.
Each embodiment of combined method of computer readable storage medium specific embodiment of the present invention and above-mentioned prediction model
Essentially identical, details are not described herein.
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, method, article or the device 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, method, article or device 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, method of element, article or device.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in a storage medium
In (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that a terminal device (can be mobile phone, computer, clothes
Business device, air conditioner or the network equipment etc.) 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 combined method of prediction model, which is characterized in that the combined method of the prediction model the following steps are included:
After obtaining prediction model by training dataset training, test data set is obtained, the test data set is input to
In each prediction model, the corresponding output result of each prediction model is obtained;
The first of each prediction model is calculated according to the corresponding legitimate reading of the test data set and the output result
Predictablity rate;
Prediction model of at least two first predictablity rates greater than preset threshold is chosen as object module, according to default
Combination combine the object module, carry out data prediction to obtain prediction built-up pattern.
2. the combined method of prediction model as described in claim 1, which is characterized in that described to choose at least two described first
Predictablity rate is greater than the prediction model of preset threshold as object module, combines the target mould according to preset combination
Type, with obtain prediction built-up pattern carry out data prediction the step of after, further includes:
After getting data set to be predicted, the data volume of the data set to be predicted is calculated;
When determining the data volume less than the first preset quantity, the data set to be predicted is input in single prediction model,
Obtain prediction result;
It, will be described when determining that the data volume is more than or equal to first preset quantity, and when less than the second preset quantity
Data set to be predicted is input in the prediction built-up pattern obtained by weighted average combination, obtains prediction result;
When determining that the data volume is more than or equal to second preset quantity, the data set to be predicted is input to logical
It crosses in the prediction built-up pattern that integration mode combines, obtains prediction result.
3. the combined method of prediction model as claimed in claim 2, which is characterized in that described when the determining data volume is greater than
Or it is equal to first preset quantity, and when less than the second preset quantity, the data set to be predicted is input to by adding
In the prediction built-up pattern that weight average combines, the step of obtaining prediction result, includes:
When determining that the data volume is more than or equal to first preset quantity, and when less than the second preset quantity, detect institute
State whether the data concentrated to training data are time series data;
If the data concentrated to training data are time series data, detection is combined by the prediction that weighted average combination obtains
Whether model is made of time series data prediction model and scattered data being prediction model;
If detecting, the prediction built-up pattern is made of time series data prediction model and scattered data being prediction model, is being examined
Measure the time series data prediction model weight be less than or equal to the scattered data being prediction model weight when, adjust institute
Time series data prediction model and the corresponding weight of the scattered data being prediction model are stated, so that the time series data prediction model
Weight is greater than the weight of the scattered data being prediction model, the prediction built-up pattern after being adjusted weight;
By in the prediction built-up pattern after the data set input adjustment weight to be predicted, prediction result is obtained.
4. the combined method of prediction model as claimed in claim 2, which is characterized in that described when the determining data volume is less than
When the first preset quantity, the step of being input in single prediction model, obtain prediction result the data set to be predicted, includes:
When determining the data volume less than the first preset quantity, the number of dimensions of the data intensive data to be predicted is calculated;
If the number of dimensions is less than default number of dimensions, the data set to be predicted is input in first kind single prediction model,
Obtain prediction result;
If the number of dimensions is more than or equal to the default number of dimensions, the data set to be predicted is input to the second class list
In prediction model, prediction result is obtained.
5. the combined method of prediction model as described in claim 1, which is characterized in that described to choose at least two described first
Predictablity rate is greater than the prediction model of preset threshold as object module, combines the target mould according to preset combination
Type, with obtain prediction built-up pattern carry out data prediction the step of after, further includes:
When detecting the presence of the objective cross mode of at least corresponding two predictions built-up pattern, the objective cross mode is calculated
In, corresponding second predictablity rate of each prediction built-up pattern;
The corresponding prediction built-up pattern of maximum second predictablity rate is chosen as default built-up pattern, and is being got to pre-
After measured data collection, the data set to be predicted is input in default built-up pattern, prediction result is obtained.
6. such as the combined method of prediction model described in any one of claim 1 to 5, which is characterized in that described according to the survey
The corresponding legitimate reading of examination data set and the output result calculate the step of the first predictablity rate of each prediction model
Suddenly include:
After obtaining output result legitimate reading corresponding with the test data set is got, detect whether that there are multiple
Export the prediction model of result;
The prediction model of multiple output results if it exists then calculates the corresponding output result of prediction model there are multiple output results
In, fruiting quantities identical with the legitimate reading;
It is corresponding to obtain that there are the predictions of multiple output results by the corresponding total quantity divided by the legitimate reading of the fruiting quantities
First predictablity rate of model.
7. the combined method of prediction model as claimed in claim 6, which is characterized in that it is described obtain the output result and
After getting the corresponding legitimate reading of the test data set, the step of there are the prediction models of multiple output results is detected whether
Later, further includes:
The prediction model of multiple output results if it does not exist, then whether the output result for detecting each prediction model is character
String;
If detecting, the output result is character string, is calculated similar between the output result and the legitimate reading
Degree, using the similarity as the first predictablity rate of corresponding prediction model.
8. a kind of combination unit of prediction model, which is characterized in that the combination unit of the prediction model includes:
Module is obtained, for obtaining test data set after obtaining prediction model by training dataset training;
Input module obtains each prediction model pair for the test data set to be input in each prediction model
The output result answered;
Computing module, it is each described pre- for being calculated according to the corresponding legitimate reading of the test data set and the output result
Survey the first predictablity rate of model;
Module is chosen, for choosing prediction model of at least two first predictablity rates greater than preset threshold as target
Model;
Composite module is counted for combining the object module according to preset combination with obtaining prediction built-up pattern
It is predicted that.
9. a kind of unit equipment of prediction model, which is characterized in that the unit equipment of the prediction model includes memory, processing
Device and the combinator for being stored in the prediction model that can be run on the memory and on the processor, the prediction model
Combinator the combination side of the prediction model as described in any one of claims 1 to 7 is realized when being executed by the processor
The step of method.
10. a kind of computer readable storage medium, which is characterized in that be stored with prediction mould on the computer readable storage medium
It realizes when the combinator of the combinator of type, the prediction model is executed by processor such as any one of claims 1 to 7 institute
The step of combined method for the prediction model stated.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910014124.8A CN109816158A (en) | 2019-01-04 | 2019-01-04 | Combined method, device, equipment and the readable storage medium storing program for executing of prediction model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910014124.8A CN109816158A (en) | 2019-01-04 | 2019-01-04 | Combined method, device, equipment and the readable storage medium storing program for executing of prediction model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109816158A true CN109816158A (en) | 2019-05-28 |
Family
ID=66604103
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910014124.8A Pending CN109816158A (en) | 2019-01-04 | 2019-01-04 | Combined method, device, equipment and the readable storage medium storing program for executing of prediction model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109816158A (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110390425A (en) * | 2019-06-20 | 2019-10-29 | 阿里巴巴集团控股有限公司 | Prediction technique and device |
CN111000544A (en) * | 2019-11-22 | 2020-04-14 | 北京航空航天大学 | Method and system for constructing hybrid continuous blood pressure measurement model based on PPG waveform |
CN111045907A (en) * | 2019-12-12 | 2020-04-21 | 苏州博纳讯动软件有限公司 | System capacity prediction method based on traffic |
CN111328401A (en) * | 2017-11-08 | 2020-06-23 | 西门子股份公司 | Method and apparatus for machine learning in a computing unit |
CN111582632A (en) * | 2020-03-26 | 2020-08-25 | 西南交通大学 | Multi-factor safety stage prediction method for whole process of underground large space construction |
CN111639798A (en) * | 2020-05-26 | 2020-09-08 | 华青融天(北京)软件股份有限公司 | Intelligent prediction model selection method and device |
CN111973165A (en) * | 2020-08-14 | 2020-11-24 | 北京航空航天大学 | Linear and nonlinear mixed non-invasive continuous blood pressure measuring system based on PPG |
CN112683923A (en) * | 2019-10-17 | 2021-04-20 | 神讯电脑(昆山)有限公司 | Method for screening surface form of object based on artificial neural network |
CN113362179A (en) * | 2021-06-30 | 2021-09-07 | 中国农业银行股份有限公司 | Prediction method, device, equipment, storage medium and program product of transaction data |
CN113866638A (en) * | 2021-08-24 | 2021-12-31 | 陈九廷 | Battery parameter inference method, device, equipment and medium |
CN116092680A (en) * | 2023-03-08 | 2023-05-09 | 成都工业学院 | Abdominal aortic aneurysm early prediction method and system based on random forest algorithm |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104732299A (en) * | 2015-04-03 | 2015-06-24 | 中国农业科学院农业信息研究所 | Maize yield combined prediction system and method |
CN107146015A (en) * | 2017-05-02 | 2017-09-08 | 联想(北京)有限公司 | Multivariate Time Series Forecasting Methodology and system |
CN107590569A (en) * | 2017-09-25 | 2018-01-16 | 山东浪潮云服务信息科技有限公司 | A kind of data predication method and device |
CN107766929A (en) * | 2017-05-05 | 2018-03-06 | 平安科技(深圳)有限公司 | model analysis method and device |
CN108364106A (en) * | 2018-02-27 | 2018-08-03 | 平安科技(深圳)有限公司 | A kind of expense report Risk Forecast Method, device, terminal device and storage medium |
US20180365525A1 (en) * | 2016-02-26 | 2018-12-20 | Alibaba Group Holding Limited | Multi-sampling model training method and device |
-
2019
- 2019-01-04 CN CN201910014124.8A patent/CN109816158A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104732299A (en) * | 2015-04-03 | 2015-06-24 | 中国农业科学院农业信息研究所 | Maize yield combined prediction system and method |
US20180365525A1 (en) * | 2016-02-26 | 2018-12-20 | Alibaba Group Holding Limited | Multi-sampling model training method and device |
CN107146015A (en) * | 2017-05-02 | 2017-09-08 | 联想(北京)有限公司 | Multivariate Time Series Forecasting Methodology and system |
CN107766929A (en) * | 2017-05-05 | 2018-03-06 | 平安科技(深圳)有限公司 | model analysis method and device |
CN107590569A (en) * | 2017-09-25 | 2018-01-16 | 山东浪潮云服务信息科技有限公司 | A kind of data predication method and device |
CN108364106A (en) * | 2018-02-27 | 2018-08-03 | 平安科技(深圳)有限公司 | A kind of expense report Risk Forecast Method, device, terminal device and storage medium |
Non-Patent Citations (1)
Title |
---|
于振明等: "一种基于专家理性预期的组合预测方法及应用", 《东北大学学报(自然科学版)》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111328401B (en) * | 2017-11-08 | 2023-10-10 | 西门子股份公司 | Method and device for machine learning in a computing unit |
CN111328401A (en) * | 2017-11-08 | 2020-06-23 | 西门子股份公司 | Method and apparatus for machine learning in a computing unit |
CN110390425A (en) * | 2019-06-20 | 2019-10-29 | 阿里巴巴集团控股有限公司 | Prediction technique and device |
CN112683923A (en) * | 2019-10-17 | 2021-04-20 | 神讯电脑(昆山)有限公司 | Method for screening surface form of object based on artificial neural network |
CN111000544A (en) * | 2019-11-22 | 2020-04-14 | 北京航空航天大学 | Method and system for constructing hybrid continuous blood pressure measurement model based on PPG waveform |
CN111045907A (en) * | 2019-12-12 | 2020-04-21 | 苏州博纳讯动软件有限公司 | System capacity prediction method based on traffic |
CN111582632A (en) * | 2020-03-26 | 2020-08-25 | 西南交通大学 | Multi-factor safety stage prediction method for whole process of underground large space construction |
CN111639798A (en) * | 2020-05-26 | 2020-09-08 | 华青融天(北京)软件股份有限公司 | Intelligent prediction model selection method and device |
CN111973165A (en) * | 2020-08-14 | 2020-11-24 | 北京航空航天大学 | Linear and nonlinear mixed non-invasive continuous blood pressure measuring system based on PPG |
CN111973165B (en) * | 2020-08-14 | 2021-12-24 | 北京航空航天大学 | Linear and nonlinear mixed non-invasive continuous blood pressure measuring system based on PPG |
CN113362179A (en) * | 2021-06-30 | 2021-09-07 | 中国农业银行股份有限公司 | Prediction method, device, equipment, storage medium and program product of transaction data |
CN113362179B (en) * | 2021-06-30 | 2024-01-30 | 中国农业银行股份有限公司 | Method, apparatus, device, storage medium and program product for predicting transaction data |
CN113866638A (en) * | 2021-08-24 | 2021-12-31 | 陈九廷 | Battery parameter inference method, device, equipment and medium |
CN116092680A (en) * | 2023-03-08 | 2023-05-09 | 成都工业学院 | Abdominal aortic aneurysm early prediction method and system based on random forest algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109816158A (en) | Combined method, device, equipment and the readable storage medium storing program for executing of prediction model | |
CN111209977B (en) | Classification model training and using method, device, equipment and medium | |
EP3757900A1 (en) | Time series prediction with confidence estimates using sparse recurrent mixture density networks | |
US11830033B2 (en) | Theme recommendation method and apparatus | |
CN106845731A (en) | A kind of potential renewal user based on multi-model fusion has found method | |
US20090307176A1 (en) | Clustering-based interest computation | |
CN108921624B (en) | Advertisement fusion method and device, storage medium and terminal equipment | |
CN110287775B (en) | Palm image clipping method, palm image clipping device, computer equipment and storage medium | |
CN107424007A (en) | A kind of method and apparatus for building electronic ticket susceptibility identification model | |
CN109658124A (en) | Effect prediction method, apparatus, medium and electronic equipment based on big data | |
CN109388674A (en) | Data processing method, device, equipment and readable storage medium storing program for executing | |
CN104221019A (en) | Method and apparatus for enhancing context intelligence in random index based system | |
CN103761254A (en) | Method for matching and recommending service themes in various fields | |
CN110019790A (en) | Text identification, text monitoring, data object identification, data processing method | |
CN111737473B (en) | Text classification method, device and equipment | |
CN110598120A (en) | Behavior data based financing recommendation method, device and equipment | |
CN108513259A (en) | Electronic device, floor location method and computer readable storage medium | |
CN111222557A (en) | Image classification method and device, storage medium and electronic equipment | |
US20230229896A1 (en) | Method and computing device for determining optimal parameter | |
CN112766402A (en) | Algorithm selection method and device and electronic equipment | |
US20220253307A1 (en) | Miniprogram classification method, apparatus, and device, and computer-readable storage medium | |
CN108197825A (en) | System scheduling method and device | |
CN107966678B (en) | Localization method, electronic device and storage medium based on signal data screening | |
CN117115596A (en) | Training method, device, equipment and medium of object action classification model | |
CN109817342A (en) | Parameter regulation means, device, equipment and the storage medium of popular season prediction model |
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: 20190528 |
|
RJ01 | Rejection of invention patent application after publication |