CN110443374A - A kind of resource information processing method, device and equipment - Google Patents
A kind of resource information processing method, device and equipment Download PDFInfo
- Publication number
- CN110443374A CN110443374A CN201910748139.7A CN201910748139A CN110443374A CN 110443374 A CN110443374 A CN 110443374A CN 201910748139 A CN201910748139 A CN 201910748139A CN 110443374 A CN110443374 A CN 110443374A
- Authority
- CN
- China
- Prior art keywords
- resource information
- prediction
- item
- information
- target
- 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.)
- Granted
Links
- 230000010365 information processing Effects 0.000 title claims abstract description 19
- 238000003672 processing method Methods 0.000 title claims abstract description 15
- 238000012549 training Methods 0.000 claims description 75
- 238000000034 method Methods 0.000 claims description 70
- 238000010801 machine learning Methods 0.000 claims description 46
- 238000012545 processing Methods 0.000 claims description 9
- 230000004044 response Effects 0.000 claims description 4
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 description 16
- 238000012360 testing method Methods 0.000 description 14
- 230000008569 process Effects 0.000 description 13
- 238000010586 diagram Methods 0.000 description 11
- 230000000694 effects Effects 0.000 description 8
- 238000009828 non-uniform distribution Methods 0.000 description 7
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 6
- 239000010931 gold Substances 0.000 description 6
- 229910052737 gold Inorganic materials 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 230000003203 everyday effect Effects 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 238000010187 selection method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 101100257124 Caenorhabditis elegans sma-10 gene Proteins 0.000 description 1
- 101100263503 Neurospora crassa (strain ATCC 24698 / 74-OR23-1A / CBS 708.71 / DSM 1257 / FGSC 987) vma-10 gene Proteins 0.000 description 1
- 101100351804 Schizosaccharomyces pombe (strain 972 / ATCC 24843) pfl8 gene Proteins 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
This application involves a kind of resource information processing method, device and equipment, which comprises obtains at least one target resource information and the current data of each target resource information；The current data of each target resource information is separately input at least two resource information prediction models to predict, obtains at least two resources information corresponding with each target resource information；Wherein at least two resource informations prediction model is corresponding at least two preset time periods；Based on the resources information corresponding with the objectives resource information, the target prediction information of the objectives resource information is determined；The application can predict respectively target resource information by different prediction models corresponding to different time sections, and the predictive information of each prediction model is integrated to determine the final prediction result of target resource information, accurate prediction result can be obtained for the data of sample non-identity distribution.
Description
Technical Field
The present application relates to the field of machine learning technologies, and in particular, to a method, an apparatus, and a device for processing resource information.
Background
Machine Learning (ML) is a multi-domain cross discipline, which relates to multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, and the like, and is used for specially researching how a computer simulates or realizes human Learning behaviors to acquire new knowledge or skills and reorganizes an existing knowledge structure to continuously improve the performance of the computer.
The traditional machine learning modeling method is to divide all sample data into a training set and a testing set, then input the data of the training set into an algorithm, obtain a model through parameter adjustment and optimization, and finally predict the testing set through the model. The premise that the machine learning algorithm is effective is that data samples are necessarily distributed identically, but for some types of data, such as financial data, the 'sample non-uniform distribution' cannot be avoided, and for the types of data, if a traditional machine learning modeling method is adopted for modeling, an overfitting problem exists, so that the accuracy and the practical application value of a prediction model are influenced.
Disclosure of Invention
The technical problem to be solved by the present application is to provide a method, an apparatus, and a device for processing resource information, which can respectively predict target resource information through different prediction models corresponding to different time periods, and determine a final prediction result of the target resource information by integrating prediction information of each prediction model, so as to obtain an accurate prediction result for data of non-uniform distribution of samples.
In order to solve the above technical problem, in one aspect, the present application provides a resource information processing method, where the method includes:
acquiring at least one item of target resource information and current data of each item of target resource information;
respectively inputting the current data of each item of resource marking information into at least two resource information prediction models for prediction to obtain at least two items of resource prediction information corresponding to each item of resource marking information; wherein the at least two resource information prediction models correspond to at least two preset time periods, respectively;
and determining target prediction information of each item of target resource information based on the resource prediction information corresponding to each item of target resource information.
In another aspect, the present application provides a resource information processing apparatus, including:
the resource information acquisition module is used for acquiring at least one item of target resource information and the current data of each item of target resource information;
the prediction module is used for respectively inputting the current data of each item of target resource information into the at least two resource information prediction models for prediction to obtain at least two items of resource prediction information corresponding to each item of target resource information; wherein the at least two resource information prediction models correspond to at least two preset time periods, respectively;
and the prediction information determining module is used for determining target prediction information of each item of target resource information based on the resource prediction information corresponding to each item of target resource information.
In another aspect, the present application provides an apparatus comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a set of codes, or a set of instructions, and the at least one instruction, the at least one program, the set of codes, or the set of instructions is loaded and executed by the processor to implement the resource information processing method as described above.
In another aspect, the present application provides a computer storage medium having at least one instruction, at least one program, a set of codes, or a set of instructions stored therein, which is loaded by a processor and executes the resource information processing method as described above.
The embodiment of the application has the following beneficial effects:
the method comprises the steps that at least two resource information prediction models corresponding to different preset time periods are adopted, data of a next time node of target resource information at a current time node are predicted respectively, and target prediction information corresponding to the target resource information is determined based on the obtained at least two pieces of prediction information; the method adopts the idea of integrated learning, the preset model is trained independently on the basis of historical data in different time periods to obtain a plurality of prediction models, when information prediction is carried out, target resource information is predicted respectively through the prediction models corresponding to the time periods, prediction results of the prediction models are integrated to obtain a final prediction result, the problem of over-fitting caused by the fact that a traditional modeling method is adopted for data of 'sample non-uniform distribution' is solved, and the prediction accuracy and the practical application value of the prediction models are improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a schematic view of an application scenario provided in an embodiment of the present application;
fig. 2 is a flowchart of a resource information processing method provided in an embodiment of the present application;
FIG. 3 is a flowchart of a resource information prediction model generation method according to an embodiment of the present disclosure;
FIG. 4 is a flowchart of a training sample creation method provided by an embodiment of the present application;
FIG. 5 is a flowchart of a resource information prediction model training method according to an embodiment of the present disclosure;
FIG. 6 is a flowchart of another resource information prediction model training method provided in an embodiment of the present application;
FIG. 7 is a flowchart of another resource information processing method provided in an embodiment of the present application;
FIG. 8 is a flowchart of a method for determining target prediction information according to an embodiment of the present disclosure;
fig. 9 is a flowchart of an information pushing method provided in an embodiment of the present application;
FIG. 10 is a schematic diagram of a prediction process provided by an embodiment of the present application;
FIG. 11 is a schematic diagram of a resource information processing apparatus according to an embodiment of the present application;
FIG. 12 is a schematic diagram of a predictive model generation module provided by an embodiment of the present application;
FIG. 13 is a schematic diagram of a training sample acquisition module provided in an embodiment of the present application;
FIG. 14 is a schematic diagram of a predictive model training module provided by an embodiment of the present application;
FIG. 15 is a schematic diagram of a first training module provided by an embodiment of the present application;
FIG. 16 is a schematic diagram of a prediction information determination module provided in an embodiment of the present application;
fig. 17 is a schematic diagram of an information push module provided in an embodiment of the present application;
fig. 18 is a schematic structural diagram of an apparatus according to an embodiment of the present application.
Detailed Description
To make the objects, technical solutions and advantages of the present application more clear, the present application will be further described in detail with reference to the accompanying drawings. It is to be understood that the described embodiments are merely a few embodiments of the present application and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or server that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Please refer to fig. 1, which shows a schematic diagram of an application scenario provided in an embodiment of the present application, where the application scenario includes: at least one terminal 110 and a server 120, the terminal 110 and the server 120 being in data communication via a network. Specifically, the server 120 may generate a prediction model for the historical data based on each item of resource information, for predicting data of a time node next to a current time node of the target resource information; the server 120 may respond to the information acquisition request sent by the terminal 110 and return corresponding prediction information to the terminal 110, or corresponding processing information obtained after processing the prediction information.
The terminal 110 may communicate with the Server 120 based on a Browser/Server mode (Browser/Server, B/S) or a Client/Server mode (Client/Server, C/S). The terminal 110 may include: the physical devices may also include software running in the physical devices, such as application programs and the like. The operating system running on the terminal 110 in the embodiment of the present application may include, but is not limited to, an android system, an IOS system, linux, windows, and the like.
The server 120 and the terminal 110 may establish a communication connection through a wired or wireless connection, and the server 120 may include a server operating independently, or a distributed server, or a server cluster composed of a plurality of servers, where the server may be a cloud server.
For solving the problem that the practical application value of a model is affected due to the fact that an overfitting condition exists when a traditional machine learning algorithm is modeled based on data of 'sample non-uniform distribution' in the prior art, an embodiment of the present application provides a resource information processing method, an execution subject of which may be a server shown in fig. 1, specifically, please refer to fig. 2, and the method includes:
s210, at least one item of target resource information and current data of each item of target resource information are obtained.
The target resource information in the embodiment of the present application may refer to an object of which the corresponding characteristic parameters change at preset time intervals, that is, for the same object, the corresponding characteristic parameters at different time nodes may have differences; for example, the above features are all available in related business objects in the financial field, such as various financial products in the security industry.
The related characteristic parameters of the object often have the characteristic of non-uniform distribution of samples, and the traditional modeling method for learning by a camera is difficult to obtain a good effect and has a poor general effect.
S220, respectively inputting the current data of each item of resource marking information into at least two resource information prediction models for prediction to obtain at least two resource prediction information corresponding to each item of resource marking information; wherein the at least two resource information prediction models correspond to at least two preset time periods, respectively.
For data of 'sample non-uniform distribution', the server adopts a method of constructing a prediction model based on cross-section data to realize prediction of data of target resource information, wherein the cross-section data is data of different subjects at the same time point or the same time period. The preset time period in the present embodiment may refer to one month, one year, and the like.
In this embodiment, resource information prediction models corresponding to respective time periods are respectively generated by the server based on historical data of the respective time periods, and when prediction is specifically performed, for each item of resource information, current data of the item of resource information needs to be respectively input into the models corresponding to the respective preset time periods for prediction, so as to obtain multiple items of resource prediction information.
And S230, determining target prediction information of each item of target resource information based on the resource prediction information corresponding to each item of target resource information.
The server needs to synthesize the resource prediction information obtained by each resource information prediction model to determine a final prediction result corresponding to the target resource information.
For the above resource information prediction model generation process, refer to fig. 3 in particular, which shows a resource information prediction model generation method, the method includes:
s310, obtaining a training sample, wherein the training sample comprises historical data of each item of resource information in at least two preset time periods.
Each item of resource information in this embodiment refers to resource information whose duration time in the system reaches a certain duration, and for resource information whose duration time is short or new resource information needs to be removed from the historical data, so as to ensure the accuracy of the result.
Each preset time period herein may be one year, for example, the historical data in at least two preset time periods in the present embodiment may specifically include historical data of each item of resource information in 2014, historical data in 2015, historical data in 2016, and the like. The historical data includes various parameter information of each resource information at each time node.
For the creation of training samples, see fig. 4 in particular, it shows a training sample creation method, the method includes:
and S410, for each preset time period, acquiring historical data of each item of resource information in the preset time period, and generating resource information records corresponding to each item of resource information.
Taking 2016 historical data as an example, the historical data includes data of each resource information at each time node, for example, for the resource information a, the historical data includes data of the resource information a at each time node, a resource information record corresponding to the resource information a is generated based on the resource information a and the data corresponding to the resource information a, and for other resource information, corresponding resource information records are generated in the same way.
And S420, generating a sample record corresponding to the preset time period based on the resource information record corresponding to each item of resource information in the preset time period.
And integrating the resource information records of resource information in 2016 to generate a sample record corresponding to 2016. That is, in the sample record corresponding to each preset time period, each resource information record includes data of each time node of the resource information in the preset time period.
And S430, constructing the training sample based on the sample record corresponding to each preset time period.
Based on historical data in 2014 and 2015, corresponding sample records are respectively generated, and the sample records in 2016 are combined to obtain training samples in the embodiment of the application by integrating the sample records in three years.
S320, training preset machine learning models respectively based on historical data of each item of resource information in each preset time period to obtain resource information prediction models corresponding to each preset time period respectively.
The machine learning algorithm adopted in the embodiment includes, but is not limited to, algorithms such as GBDT, xgboost, random forest and the like, wherein GBDT prediction capability is strong, and accuracy is high, and as an integrated learning algorithm, the integrated learning algorithm integrates a plurality of base models for prediction, so that overfitting can be effectively reduced, meanwhile, nonlinear data can be processed, and feature screening can be automatically performed.
The number of the preset time periods in the training sample is the number of the resource information prediction models to be generated, that is, for each preset time period, the resource information prediction model corresponding to the time period needs to be generated, and specifically, the number of the preset time periods in the training sample may be: the year 2014 corresponds to one resource information prediction model, the year 2015 corresponds to one resource information prediction model, the year 2016 corresponds to one resource information prediction model, and the three years 2014 to 2016 are taken as one time period and can also correspond to one resource information prediction model.
Specifically, please refer to fig. 5, which illustrates a resource information prediction model training method, the method includes:
and S510, traversing all resource information records in the sample record corresponding to each preset time period.
The server traverses each resource information record in the sample record of each preset time period in a certain sequence, and the specific traversal sequence is not limited as long as all resource information records in the sample record can be traversed.
S520, training the current machine learning model based on the data of each time node in the current resource information record.
S530, judging whether all resource information records are traversed.
S540, when the resource information records are not traversed, determining that the next resource information record of the current resource information record is the current resource information record, and executing the step S520.
And S550, when all resource information records are traversed, determining that the current machine learning model is the resource information prediction model corresponding to the preset time period.
For the above step S520, it is trained on the current machine learning model based on each time node data in one resource information record, and the specific process can be seen in fig. 6, which shows another resource information prediction model training method, where the method includes:
s610, traversing each time node in the current resource information record from the starting time node.
For each time node in the current resource information record, traversal is performed in a certain order, and the specific traversal order is not limited as long as each time node in the current resource information record can be traversed.
And S620, taking the data of the current time node as the input of the current machine learning model, taking the data of the next time node of the current time node as the output of the current machine learning model, and training the current machine learning model.
In this embodiment, the data of each time node may include a plurality of parameters, and when training is specifically performed, the server may use the plurality of parameters of the current time node as input of the current machine learning model, and use at least one of all parameters of the next time node of the current time node as output of the current machine learning model, so as to perform supervised training on the machine learning model.
And S630, judging whether the current time node is the termination time node.
And S640, if the current time node is not the termination time node, determining that the next time node of the current time node is the current time node, and executing the step S620.
S650, if the current time node is the termination time node, determining that the training process based on the current resource information record is completed.
The resource information prediction model in this embodiment predicts at least one parameter of the target resource information at a time node next to the current time node based on data of the target resource information at the current time node; namely, the prediction target in the embodiment of the present application is at least one parameter of the target resource information at the next time node.
Referring to fig. 7, another resource information processing method is shown, which includes:
and S710, acquiring at least one item of target resource information and current data of each item of target resource information.
S720, respectively inputting the current data of each item of resource marking information into at least two resource information prediction models for prediction to obtain at least two resource prediction information corresponding to each item of resource marking information; wherein the at least two resource information prediction models correspond to at least two preset time periods, respectively.
And S730, inputting the current data of each item of resource information into the comprehensive prediction model for prediction to obtain a resource prediction information.
The comprehensive prediction model is obtained by training a preset machine learning model based on historical data of each resource information in each preset time period.
The comprehensive prediction model is a prediction model obtained by training a preset machine learning model based on integral sample data by a server, and data of each preset time period is not distinguished in the training process of the model. And traversing each resource information record in the training sample and each time node in each resource information record by adopting a training method similar to the model training, and training the model according to the data of the corresponding time node, wherein specific implementation details can refer to the method in the embodiment and are not repeated herein.
S740, determining target prediction information of each item of target resource information based on the resource prediction information corresponding to each item of target resource information.
For processing each item of prediction information obtained based on each prediction model in the above steps to obtain final target prediction information, specifically, refer to fig. 8, which shows a target prediction information determining method, including:
and S810, calculating the arithmetic mean value of each item of resource prediction information corresponding to the target resource information for each item of resource information.
Here, each item of resource prediction information may include only resource prediction information obtained by each resource information prediction model, or may include resource prediction information obtained by each resource information prediction model and resource prediction information obtained by a comprehensive prediction model, and after each item of resource prediction information is obtained, an arithmetic average of each item of resource prediction information is calculated.
And S820, determining the arithmetic mean value as target prediction information of the target resource information.
According to the resource information processing method, the prediction models for predicting the resource information can be independently constructed based on the section data of different time periods, when the server predicts the information, the target resource information is respectively predicted through the prediction models corresponding to the time periods, then the prediction results of the prediction models are integrated to obtain the final prediction result, the over-fitting problem caused by the fact that the traditional modeling method is adopted for the data of 'sample non-uniform distribution' is solved, and the prediction accuracy and the practical application value of the prediction models are improved.
Referring to fig. 9, an information pushing method is shown, the method comprising:
s910, in response to the information acquisition request, ranking the target resource information based on the target prediction information of the target resource information.
S920, based on the sequencing result of the target resource information, a resource information list including the target resource information is generated.
S930, pushing the resource information list.
The information pushing method is realized based on the prediction result of the target resource, the target prediction information of each target resource is used as an index, the target resource information is sequenced, and the sequenced resource information list is pushed to the client and displayed on a user interface.
Through the information pushing method, the user can directly obtain the sequencing result of each item of resource information obtained based on the prediction information without manually searching, screening and predicting a large amount of resource information.
The specific implementation process of the application is described below by taking a specific example, and the stock in the financial field is taken as an example for description, the prediction target is the rise and fall amplitude of the next day of the stock, the supervised learning method is specifically adopted for training, the regression prediction training process is a regression prediction training process, and in the selection of the machine learning algorithm, an algorithm with strong prediction capability and high precision, such as GBDT, is selected as an integrated learning algorithm, and the integrated learning algorithm integrates a plurality of base models for prediction, so that overfitting can be effectively reduced, meanwhile, nonlinear data can be processed, and feature screening can be automatically performed. In addition to GBDT, other similar algorithms can be tried to be selected, such as xgboost, randomforest, etc.
1. Construction of individual strand characteristics: the characteristics of each stock every day and the rise and fall amplitude of the next day constitute one input of an algorithm; because the forecast target is the rise and fall amplitude of the next day of the stock, which is the target of a short line, the characteristics of the stock are selected from data following the market quotation aging, and the characteristics comprise volume factors (such as opening price, closing price, highest price, lowest price, rise and fall amplitude, volume of transaction and the like), technical indexes (such as KDJ, BOLL index, RSI, MACD, VMACD, MA, BIAS, WR, MACD cross, KDJ cross, RSI cross and the like), technical forms (jump-sky-up, super-fall rebound, multi-head breakthrough, average line multi-head arrangement, early light appearance, early morning stars and the like), and the like, and 42 total characteristics; the target variable is the next day rise and fall of each stock (the next day rise and fall is the next day closing price/today closing price-1). The 42 features are used as an input vector X, a target variable Y corresponding to the X is chg, namely the next day rise and fall amplitude, the features of each stock in each day are used as the input of a model, the next day rise and fall amplitude is used as the output of the model, and the model is trained.
2. Training of models
(1) The input data is constructed in the manner described above, one for each stock per day. And constructing a data set, wherein the data set for general training needs to contain a complete ox bear period data of N years. In this embodiment, data from year 2013 to year 2016 are taken as a training set, and data from year 2017 to year 2018 are taken as a test set, where N is 4. There are more than 3 thousand stocks on the market, nearly 250 trading days each year, the actual data set is constructed, the scale of the training set is about 330 ten thousand, the test set is about 160 ten thousand, and the training set VS test set: 2VS 1. Because the stock A has 'new welfare', the noise data of the stock on the new market is removed in the modeling process.
(2) Respectively and independently training the annual data and the overall data of the training set to a prediction model fi(X),i∈[1,N+1]Wherein N is the number of years of the training set, generally not less than one bear period, and in the training set, N is 4, specifically, the section model f is trained based on data in 2013, 2014, 2015 and 2016 separately1(X)、f2(X)、f3(X)、f4(X), then training an overall model f based on overall data from 2013 to 20165(X)。
3. Model testing
For each record X of the test set, it is input separately to the model f1(X)、f2(X)、f3(X)、f4(X) and f5(X) obtaining its predicted value Y in each model1，Y2，Y3，Y4，Y5And comprehensively scoring by using the prediction results of the 5 models, wherein the scoring mode is arithmetic mean:
4. obtaining the comprehensive prediction result of the test sample XThen, the absolute value of the error between the predicted value and the actual fluctuation range chg can be calculated:
error＝|r-chg|(2)
the optimization goal of the model is the sum of the squares of the absolute values of the errors of all the test samples, error _ square _ sum, so that the error _ square _ sum is minimized.
5. Comprehensive predictive scoring
When the model is used for prediction, in order to predict the rise and fall of the next day, the input of the prediction model is a feature vector X _ NEW ═ consisting of 42 features of the current day of each stock in a NEW sample (openPrice, closed price, highPrice, lowPrice, volume, rise, kdj _ k, kdj _ d, kdj _ j, blgb _ up, blgb _ mid, blgb _ down, rsi6, rsi12, macd, vmacd, sma5, sma10, sma20, vma5, vma10, vma20, bias5, bias10, bias20, ward, macd _ gold, kdj _ top _ gold _ text, bias _ big _ bell _ middle, bias _ bow _ noise, bias _ noise _. The specific form of the prediction result is to output a continuous variable Y _ NEW for each NEW sample, the prediction value represents the next-day fluctuation amplitude predicted by the model, wherein each NEW sample comprises the current-day eigenvector of at least one stock, and the output Y _ NEW is the next-day fluctuation amplitude prediction value respectively corresponding to each stock in the NEW sample.
A specific schematic diagram can be seen in FIG. 10, which shows the process of predicting a stock based on the prediction model of the embodiment of the present application, wherein XaIs the characteristic input of each stock per day, y, over a period of timeaCorresponding rise and fall amplitude of the next day, so as to train the model and obtain the model fa(ii) a Thereby obtaining each prediction model, and based on each prediction model, selecting new samplesThe next day rise and fall of each stock in the stock is predicted.
The data covered by the training, four section models and the whole model described above, is data of all stocks of the year on which the model is based. For example, model training is performed based on 2013 data, all stock data of each day of the year are taken out to construct a data set for training, the number of trading days in a year is about 250, the total number of stocks on the market is 3500, and the size of the data set is about 250 × 3500 to 87. For each year of data, the data of each day in the year needs to be taken out to form a data set for training.
A modeling method of multi-section model comprehensive scoring different from the traditional machine learning modeling method is adopted for the next day rise and fall amplitude of the stock, a GBDT algorithm is used for separately training models for section data in different time periods, the idea of integrated learning is used for reference, new samples are comprehensively scored by using different section models, and the comprehensive scoring is used as a prediction result of each stock. The reason is that the premise of machine learning is that the samples are distributed in the same way, but the financial data are usually distributed in different ways, and the traditional modeling method of machine learning is difficult to obtain good effect, so that a model construction method based on multiple sections is provided.
Comparing the next day rise and fall amplitude prediction model proposed in the present application with a traditional machine learning model:
the next-day rise and fall amplitude of the stocks predicted by the model is an index for sorting the stocks during stock selection, the larger the index value is, the larger the possible rise and fall amplitude of the stocks in the next day is, and the better return on investment income can be obtained by selecting the stocks with larger rise and fall amplitude. Wherein the rise and fall amplitude is generally between-10% and 10%.
1. And (4) sorting the stock samples of each day in the test set in a descending order based on the prediction result (the next day rise and fall amplitude) of the constructed prediction model. When the predicted values are the same and cannot be distinguished, the two indexes of the rise and fall amplitude, the volume of successful transactions and the volume of successful transactions in the previous day are used for descending and sorting. The stocks in each day are sorted according to the above rule, and then divided into 20 groups (the stocks in each group are equal in number), and the fluctuation performance of the stocks in the 20 groups on the next day is observed.
And traversing the test data set, and calculating the actual next day rise and fall amplitude of each stock. And grouping the model performance evaluation indexes according to the model prediction results. The mean rise and fall amplitude of group j was:
wherein chgi,j,kRepresents the rise and fall of the kth stock of the ith group on the ith day (the predicted ith day, the actual rise and fall appear on i +1 days), wherein i belongs to N+(positive integer), j ∈ [1,20]]，k∈N+(positive integer), chgi,j,k(stock k closing price on i +1 day/stock k closing price on i day) -1.
If the rise and fall amplitude is greater than 0, the win ratio win _ rate of the j group on the ith dayi,jThe calculation method is as follows: if the number of the j groups of stocks with the rise and fall amplitude larger than 0 on the ith day is T and the total number of the groups of stocks is T, the number of the j groups of stocks is T
win_ratei,j＝(t/T)*100％ (4)
Then the average win rate for the jth group of the full history is:
based on the calculation mode, the average fluctuation range and the average win ratio of each group can be obtained by respectively calculating each group j epsilon [1,20 ]. The investment performance of each group is shown in the following table:
TABLE 1 average Sunday odds for each group
TABLE 2 average benefit of each group
| Grouping | Prediction model proposed by the application | Traditional machine learning model |
| 0 | 0.18％ | -0.05％ |
| 1 | 0.09％ | -0.04％ |
| 2 | 0.07％ | 0.01％ |
| 3 | 0.06％ | 0.01％ |
| 4 | 0.03％ | -0.05％ |
| 5 | -0.06％ | 0.01％ |
| 6 | -0.06％ | -0.12％ |
| 7 | -0.07％ | -0.07％ |
| 8 | -0.06％ | -0.06％ |
| 9 | -0.06％ | -0.11％ |
| 10 | -0.07％ | 0.05％ |
| 11 | -0.07％ | -0.09％ |
| 12 | -0.12％ | -0.15％ |
| 13 | -0.14％ | -0.16％ |
| 14 | -0.13％ | -0.15％ |
| 15 | -0.14％ | -0.22％ |
| 16 | -0.16％ | -0.03％ |
| 17 | -0.17％ | 0.03％ |
| 18 | -0.19％ | -0.15％ |
| 19 | -0.48％ | 0.01％ |
The average win rate and the average profit rate in the graph are evaluation indexes of the model and are performances of the constructed prediction model on the test data set.
According to the predicted rising and falling amplitude descending order of the models every day, 5 stocks at the top are obtained, the stock selection capacity of different models is sequentially evaluated, and the winning rate of each model is shown in table 3:
TABLE 3 results of comparison of the odds for each model
The results show that the success rate of the prediction model provided in the application is obviously superior to that of other models, and is improved by 33.76% compared with the traditional GBDT and is improved by 43.35% compared with a linear model. Wherein, the victory ratio is the average victory ratio of 5 stocks selected next day, and the specific calculation mode of the victory ratio is as follows:
assuming that 5 selected stocks are selected, the actual change and fall of the next day is chgi,1......,chgi,j......,chgi,5Where i in the subscript indicates the ith day (the ith day on which the prediction is made, and the actual rise and fall occur on the (i + 1) th day), and j ═ {1,2,3,4,5} in the subscript indicates the j th stocks ranked in descending order of the predicted rise and fall, respectively. Wherein,
chgi,j(stock j ith +1 th day closing price/stock j ith day closing price) -1 (6)
The stocks with the top 5 ranks selected by the model every day are taken out from the test data set, and the actual next-day rise-fall amplitude chg of all the stocks is calculatedi,j，i∈N+(positive integer) j ∈ [1,5 ]]. When the fluctuation range is larger than 0, the stock is counted as winning, and if the number of winning is t in all the selected stocks in the test set, the winning rate is:
win_rate＝(t/(max(i)*5))*100％ (7)
therefore, the stock next-day fluctuation amplitude-based prediction method has better distinguishing capability and stock selection effect. Stock selection is a common function of stock software and is a decision-making auxiliary tool commonly used by vast investors in stock investment activities. The stock-selecting strategy is an important component of the stock-selecting function, and the effectiveness of the stock-selecting strategy is the most fundamental power for selecting the function. Therefore, the prediction model of the application can be used for predicting the related information of the stocks in the related stock software. It should be noted that, in this embodiment, the next day rise and fall of the stock is taken as a prediction target for description, and other relevant features of the stock can be predicted in the actual application process.
The historical return performance of the prediction model provided by the application can be seen in the following table:
TABLE 4 historical return survey Performance evaluation results
As can be seen from Table 4, the effect of holding 5 before and 5 after holding is distinct; the strategy effect of the first 5 is greatly increased to the 300 th index of the wins and wins.
In a short-line stock selection scene with granularity of day, hour and the like as a time unit, the current mainstream stock selection method is mainly a stock selection method based on technical indexes, such as MACD gold crosses, KDJ gold crosses, RSI gold crosses and the like, but the winning rate and the profitability are generally poor; the resource information processing method provided by the embodiment of the application can be applied to the scene of short-line stock selection, and the effect performance of the method is superior to other short-line stock selection methods such as technical index stock selection and capital face stock selection. A comprehensive scoring model for predicting the next-day rise and fall amplitude of the stock is constructed based on section data of different time periods, the next-day rise and fall amplitude and the yield of the stock are well distinguished, the winning rate performance is excellent and reaches 72.1%, and the method is obviously superior to classical technical stock selection strategies such as MACD gold cross and the like.
Referring to fig. 11, the present embodiment further provides a resource information processing apparatus, including:
a resource information obtaining module 1110, configured to obtain at least one item of target resource information and current data of each item of target resource information;
the prediction module 1120 is configured to input current data of each item of resource information into the at least two resource information prediction models respectively for prediction, so as to obtain at least two resource prediction information corresponding to each item of resource information; wherein the at least two resource information prediction models correspond to at least two preset time periods, respectively;
a prediction information determination module 1130, configured to determine target prediction information of each item of target resource information based on the resource prediction information corresponding to each item of target resource information.
Referring to fig. 12, the apparatus further includes a prediction model generation module 1200, and the prediction model generation module 1200 includes:
a training sample obtaining module 1210, configured to obtain a training sample, where the training sample includes historical data of each item of resource information in at least two preset time periods;
the prediction model training module 1220 is configured to train preset machine learning models respectively based on the historical data of each item of resource information in each preset time period, so as to obtain resource information prediction models respectively corresponding to each preset time period.
Referring to fig. 13, the training sample acquiring module 1210 includes:
a history data obtaining module 1310, configured to obtain, for each preset time period, history data of each item of resource information in the preset time period, and generate resource information records corresponding to each item of resource information;
a sample record generating module 1320, configured to generate a sample record corresponding to the preset time period based on the resource information record corresponding to each item of resource information in the preset time period;
a training sample construction module 1330 configured to construct the training sample based on the sample record corresponding to each preset time period.
In the sample record corresponding to each preset time period, each resource information record comprises data of each time node of the resource information in the preset time period; each preset time period comprises a start time node and an end time node.
Referring to fig. 14, the prediction model training module 1220 includes:
the first traversal module 1410 is configured to traverse each resource information record in the sample record corresponding to each preset time period;
a first training module 1420, configured to train the current machine learning model based on data of each time node in the current resource information record;
a first repeating module 1430, configured to determine that a next resource information record of the current resource information record is the current resource information record, and repeat the step of training the current machine learning model until all resource information records are traversed;
and a prediction model determining module 1440 configured to determine that the current machine learning model is the resource information prediction model corresponding to the preset time period.
Referring to fig. 15, the first training module 1420 includes:
a second traversal module 1510, configured to traverse each time node in the current resource information record from the start time node;
the second training module 1520, configured to train the current machine learning model by using data of a current time node as an input of the current machine learning model and using data of a next time node of the current time node as an output of the current machine learning model;
the second repeating module 1530 is configured to determine that a next time node of the current time node is the current time node, and repeat the step of training the current machine learning model until the current time node is the end time node.
The device further comprises:
the comprehensive prediction module is used for inputting the current data of each item target resource information into the comprehensive prediction model to obtain one item of resource prediction information corresponding to each item target resource information;
the comprehensive prediction model is obtained by training a preset machine learning model based on historical data of each item of resource information in each preset time period.
Referring to fig. 16, the prediction information determination module 1130 includes:
an average calculating module 1610, configured to calculate, for each item of target resource information, an arithmetic average of each item of resource prediction information corresponding to the target resource information;
a first determining module 1620, configured to determine the arithmetic mean as the target prediction information of the target resource information.
Referring to fig. 17, the apparatus further includes an information pushing module 1700, where the information pushing module 1700 includes:
a request response module 1710, configured to, in response to the information acquisition request, rank the target resource information items based on target prediction information of the target resource information items;
a resource information list generating module 1720, configured to generate a resource information list including each item of target resource information based on a result of sorting each item of target resource information;
a first pushing module 1730, configured to push the resource information list.
The device provided in the above embodiments can execute the method provided in any embodiment of the present application, and has corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in the above embodiments may be referred to a method provided in any of the embodiments of the present application.
The present embodiments also provide a computer-readable storage medium having stored therein at least one instruction, at least one program, set of codes, or set of instructions that is loaded by a processor and performs any of the methods described above in the present embodiments.
Referring to fig. 18, the apparatus 1800 may have a large difference due to different configurations or performances, and may include one or more Central Processing Units (CPUs) 1822 (e.g., one or more processors) and a memory 1832, and one or more storage media 1830 (e.g., one or more mass storage devices) for storing applications 1842 or data 1844. The memory 1832 and the storage medium 1830 may be, among other things, transient storage or persistent storage. The program stored on the storage medium 1830 may include one or more modules (not shown), each of which may include a sequence of instructions operating on a device. Still further, a central processor 1822 may be provided in communication with the storage medium 1830 to execute a series of instruction operations on the device 1800 within the storage medium 1830. The apparatus 1800 may also include one or more power supplies 1826, one or more wired or wireless network interfaces 1850, one or more input-output interfaces 1858, and/or one or more operating systems 1841, such as Windows Server, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, and so forth. Any of the methods described above in this embodiment can be implemented based on the apparatus shown in fig. 18.
The present specification provides method steps as described in the examples or flowcharts, but may include more or fewer steps based on routine or non-inventive labor. The steps and sequences recited in the embodiments are but one manner of performing the steps in a multitude of sequences and do not represent a unique order of performance. In the actual system or interrupted product execution, it may be performed sequentially or in parallel (e.g., in the context of parallel processors or multi-threaded processing) according to the embodiments or methods shown in the figures.
The configurations shown in the present embodiment are only partial configurations related to the present application, and do not constitute a limitation on the devices to which the present application is applied, and a specific device may include more or less components than those shown, or combine some components, or have an arrangement of different components. It should be understood that the methods, apparatuses, and the like disclosed in the embodiments may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is merely a division of one logic function, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or unit modules.
Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
Those of skill would further appreciate that the various illustrative components and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The above embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (10)
1. A resource information processing method is characterized by comprising the following steps:
acquiring at least one item of target resource information and current data of each item of target resource information;
respectively inputting the current data of each item of resource marking information into at least two resource information prediction models for prediction to obtain at least two items of resource prediction information corresponding to each item of resource marking information; wherein the at least two resource information prediction models correspond to at least two preset time periods, respectively;
and determining target prediction information of each item of target resource information based on the resource prediction information corresponding to each item of target resource information.
2. The method according to claim 1, further comprising a step of generating the resource information prediction model, wherein the step of generating the resource information prediction model comprises:
acquiring a training sample, wherein the training sample comprises historical data of each item of resource information in at least two preset time periods;
and training preset machine learning models respectively based on the historical data of each item of resource information in each preset time period to obtain resource information prediction models respectively corresponding to each preset time period.
3. The method according to claim 2, wherein the obtaining training samples includes:
for each preset time period, acquiring historical data of each item of resource information in the preset time period, and generating resource information records corresponding to each item of resource information;
generating a sample record corresponding to the preset time period based on the resource information record corresponding to each item of resource information in the preset time period;
and constructing the training sample based on the sample record corresponding to each preset time period.
4. The method according to claim 3, wherein in the sample record corresponding to each preset time period, each resource information record includes data of each time node of the resource information in the preset time period;
the training of the preset machine learning model is respectively carried out on the basis of the historical data of each item of resource information in each preset time period, and the obtaining of at least two resource information prediction models respectively corresponding to each preset time period comprises the following steps:
traversing all resource information records in the sample record corresponding to each preset time period;
training a current machine learning model based on data of each time node in the current resource information record;
determining the next resource information record of the current resource information record as the current resource information record, and repeating the step of training the current machine learning model until all resource information records are traversed;
and determining that the current machine learning model is the resource information prediction model corresponding to the preset time period.
5. The resource information processing method according to claim 4, wherein each preset time period includes a start time node and an end time node;
the training of the current machine learning model based on the data of each time node in the current item resource information record comprises:
traversing each time node in the current resource information record from the starting time node;
taking data of a current time node as input of a current machine learning model, taking data of a next time node of the current time node as output of the current machine learning model, and training the current machine learning model;
and determining the next time node of the current time node as the current time node, and repeating the step of training the current machine learning model until the current time node is the termination time node.
6. The method for processing resource information according to claim 2, further comprising:
inputting the current data of each item target resource information into a comprehensive prediction model to obtain one item of resource prediction information corresponding to each item target resource information;
the comprehensive prediction model is obtained by training a preset machine learning model based on historical data of each item of resource information in each preset time period.
7. The method according to claim 6, wherein said determining target prediction information of each item of target resource information based on the resource prediction information corresponding to each item of target resource information comprises:
calculating the arithmetic mean value of each item of resource prediction information corresponding to the target resource information for each item of resource information;
and determining the arithmetic mean as target prediction information of the target resource information.
8. The method for processing resource information according to claim 1, further comprising:
in response to the information acquisition request, sequencing each item of target resource information based on target prediction information of each item of target resource information;
generating a resource information list comprising the target resource information on the basis of the sequencing result of the target resource information;
and pushing the resource information list.
9. A resource information processing apparatus characterized by comprising:
the resource information acquisition module is used for acquiring at least one item of target resource information and the current data of each item of target resource information;
the prediction module is used for respectively inputting the current data of each item of target resource information into the at least two resource information prediction models for prediction to obtain at least two items of resource prediction information corresponding to each item of target resource information; wherein the at least two resource information prediction models correspond to at least two preset time periods, respectively;
and the prediction information determining module is used for determining target prediction information of each item of target resource information based on the resource prediction information corresponding to each item of target resource information.
10. An apparatus comprising a processor and a memory, the memory having stored therein at least one instruction, at least one program, a set of codes, or a set of instructions, the at least one instruction, the at least one program, the set of codes, or the set of instructions being loaded and executed by the processor to implement the resource information processing method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910748139.7A CN110443374B (en) | 2019-08-14 | 2019-08-14 | Resource information processing method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910748139.7A CN110443374B (en) | 2019-08-14 | 2019-08-14 | Resource information processing method, device and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110443374A true CN110443374A (en) | 2019-11-12 |
| CN110443374B CN110443374B (en) | 2024-07-12 |
Family
ID=68435266
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910748139.7A Active CN110443374B (en) | 2019-08-14 | 2019-08-14 | Resource information processing method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110443374B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111445010A (en) * | 2020-03-26 | 2020-07-24 | 南京工程学院 | Distribution network voltage trend early warning method based on evidence theory fusion quantum network |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107633332A (en) * | 2017-09-30 | 2018-01-26 | 国信优易数据有限公司 | A kind of Electronic Finance resource prediction method and system |
| CN108596765A (en) * | 2018-04-28 | 2018-09-28 | 国信优易数据有限公司 | A kind of Electronic Finance resource recommendation method and device |
| CN108596393A (en) * | 2018-04-28 | 2018-09-28 | 国信优易数据有限公司 | A kind of training method and device of Electronic Finance resources model |
| CN108846695A (en) * | 2018-06-07 | 2018-11-20 | 中国联合网络通信集团有限公司 | The prediction technique and device of terminal replacement cycle |
| CN109615106A (en) * | 2018-10-16 | 2019-04-12 | 平安科技（深圳）有限公司 | Stock yield method for pushing, device, storage medium and server |
| CN109658643A (en) * | 2018-12-06 | 2019-04-19 | 中国工商银行股份有限公司 | Payment risk alarm method, server and intelligence POS |
| CN109697522A (en) * | 2017-10-20 | 2019-04-30 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus of data prediction |
| CN110033312A (en) * | 2019-03-13 | 2019-07-19 | 平安城市建设科技(深圳)有限公司 | Generation method, device, equipment and the storage medium of room rate prediction model |
-
2019
- 2019-08-14 CN CN201910748139.7A patent/CN110443374B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107633332A (en) * | 2017-09-30 | 2018-01-26 | 国信优易数据有限公司 | A kind of Electronic Finance resource prediction method and system |
| CN109697522A (en) * | 2017-10-20 | 2019-04-30 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus of data prediction |
| CN108596765A (en) * | 2018-04-28 | 2018-09-28 | 国信优易数据有限公司 | A kind of Electronic Finance resource recommendation method and device |
| CN108596393A (en) * | 2018-04-28 | 2018-09-28 | 国信优易数据有限公司 | A kind of training method and device of Electronic Finance resources model |
| CN108846695A (en) * | 2018-06-07 | 2018-11-20 | 中国联合网络通信集团有限公司 | The prediction technique and device of terminal replacement cycle |
| CN109615106A (en) * | 2018-10-16 | 2019-04-12 | 平安科技（深圳）有限公司 | Stock yield method for pushing, device, storage medium and server |
| CN109658643A (en) * | 2018-12-06 | 2019-04-19 | 中国工商银行股份有限公司 | Payment risk alarm method, server and intelligence POS |
| CN110033312A (en) * | 2019-03-13 | 2019-07-19 | 平安城市建设科技(深圳)有限公司 | Generation method, device, equipment and the storage medium of room rate prediction model |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111445010A (en) * | 2020-03-26 | 2020-07-24 | 南京工程学院 | Distribution network voltage trend early warning method based on evidence theory fusion quantum network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110443374B (en) | 2024-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Becker-Peth et al. | Behavioral inventory decisions | |
| CN111242310B (en) | Feature validity evaluation method and device, electronic equipment and storage medium | |
| US20210103858A1 (en) | Method and system for model auto-selection using an ensemble of machine learning models | |
| Chitra et al. | Customer retention in banking sector using predictive data mining technique | |
| CN109711424B (en) | Behavior rule acquisition method, device and equipment based on decision tree | |
| Kiss et al. | Predicting dropout using high school and first-semester academic achievement measures | |
| US10191985B1 (en) | System and method for auto-curation of Q and A websites for search engine optimization | |
| CN110222838B (en) | Document sorting method and device, electronic equipment and storage medium | |
| CN110175689A (en) | A kind of method of probabilistic forecasting, the method and device of model training | |
| Viroonluecha et al. | Salary predictor system for thailand labour workforce using deep learning | |
| CN115423538A (en) | Method and device for predicting new product sales data, storage medium and electronic equipment | |
| CN114519508A (en) | Credit risk assessment method based on time sequence deep learning and legal document information | |
| Lakshmi et al. | An identification of better engineering college with conflicting criteria using adaptive TOPSIS | |
| Ribeiro et al. | Importance of statistics for data mining and data science | |
| Parkhimenka et al. | Heuristic approach to online purchase prediction based on internet store visitors classification using data mining methods | |
| CN110443374A (en) | A kind of resource information processing method, device and equipment | |
| Kanwal et al. | An attribute weight estimation using particle swarm optimization and machine learning approaches for customer churn prediction | |
| Gerard et al. | Combining multivariate density forecasts using predictive criteria | |
| CN111222993A (en) | Fund recommendation method and device | |
| Prihatmono et al. | Application of the KNN Algorithm for Predicting Data Card Sales at PT. XL Axiata Makassar | |
| KR20120103310A (en) | Calculation system and the method of a highest bid price using by database | |
| Bhardwaj et al. | Forecasting GDP per capita of OECD countries using machine learning and deep learning models | |
| Singh et al. | Data mining: consumer behavior analysis | |
| CN111913940B (en) | Temperature membership tag prediction method and device, electronic equipment and storage medium | |
| Kulakou | Exploration of time-series models on time series data |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TG01 | Patent term adjustment | ||
| TG01 | Patent term adjustment |