CN114627330A - Time sequence flow prediction method and device, storage medium and electronic equipment - Google Patents
Time sequence flow prediction method and device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN114627330A CN114627330A CN202210203124.4A CN202210203124A CN114627330A CN 114627330 A CN114627330 A CN 114627330A CN 202210203124 A CN202210203124 A CN 202210203124A CN 114627330 A CN114627330 A CN 114627330A
- Authority
- CN
- China
- Prior art keywords
- time sequence
- historical
- data
- thermodynamic diagram
- image
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/147—Network analysis or design for predicting network behaviour
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a time sequence flow prediction method and device, a storage medium and an electronic device, wherein the method comprises the following steps: acquiring each historical service time sequence, and processing each historical service time sequence to obtain a historical time sequence thermodynamic diagram; extracting image characteristic data in a historical time-series thermodynamic diagram; and processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence. By converting each historical service time sequence into a historical time sequence thermodynamic diagram and extracting image characteristic data from the historical time sequence thermodynamic diagram, wherein the image characteristic data comprises various high-dimensional characteristics of the time sequence and contains global and local related characteristics of the time sequence, the predicted time sequence flow of each historical service time sequence can be obtained by processing the image characteristic data, the image characteristic data containing the high-dimensional characteristics is introduced, and the accuracy of predicting the flow of the time sequence is effectively improved.
Description
Technical Field
The present invention relates to the field of data processing technologies, and in particular, to a time sequence traffic prediction method and apparatus, a storage medium, and an electronic device.
Background
At present, with the rapid development of the internet industry, network services develop various forms, such as instant messaging, search engines, social entertainment, teleworking, online transactions, public services and the like, the scale of the network services is explosively increased, the network demand is increased, however, network resources are limited, and network congestion and service quality reduction are inevitably caused by excessive access clicks of users at the same time, so that network behaviors of the users are analyzed, and by predicting information flows, enterprises can be helped to manage, design and plan the network resources, and the enterprise cost is effectively reduced. Time-series flow prediction has developed rapidly within the industry in recent years.
The existing mainstream time sequence prediction method starts from the time sequence, usually takes the time sequence as the input of a prediction model, and predicts the data development trend based on the time domain characteristics and the frequency domain characteristics of the time sequence.
Disclosure of Invention
In view of this, the present invention provides a time series flow prediction method and apparatus, a storage medium, and an electronic device, which improve accuracy of time series prediction by introducing image feature data including high-dimensional features and associating information between time series.
In order to achieve the above purpose, the embodiments of the present invention provide the following technical solutions:
the invention discloses a time sequence flow prediction method in a first aspect, which comprises the following steps:
acquiring each historical service time sequence;
processing each historical service time sequence to obtain a historical time sequence thermodynamic diagram;
extracting image characteristic data in the historical time-series thermodynamic diagram;
and processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence.
Optionally, the foregoing method, wherein the processing the historical service time series to obtain a historical time series thermodynamic diagram includes:
constructing a data matrix based on each historical service time sequence;
and carrying out normalization processing on each numerical value in the data matrix to obtain a historical time sequence thermodynamic diagram.
The method described above, optionally, the extracting image feature data in the historical time-series thermodynamic diagram includes:
inputting the historical time sequence thermodynamic diagram into a feature extraction model which is trained in advance, so that the feature extraction model extracts high-dimensional features of each item of image from the historical time sequence thermodynamic diagram;
and taking each high-dimensional feature of the image as image feature data of the historical time-series thermal image.
Optionally, in the method, the processing the image feature data to obtain the predicted time-series flow of each historical service time series includes:
inputting the image characteristic data into a classification neural network trained in advance, so that the classification neural network outputs picture classification data of the historical time sequence thermodynamic diagram, wherein the picture classification data comprises prediction information of time sequence flow of each historical service time sequence;
calling a preset regression function to carry out regression processing on the image classification data to obtain regression data corresponding to the image classification data;
and carrying out inverse normalization processing on the regression data to obtain the predicted time sequence flow of each historical service time sequence.
The above method, optionally, further includes:
and performing risk assessment based on the predicted time sequence flow of each historical service time sequence to obtain a risk score of the service corresponding to each historical service time sequence.
A second aspect of the present invention discloses a time-series flow rate prediction apparatus, including:
the acquisition unit is used for acquiring each historical service time sequence;
the first processing unit is used for processing each historical service time sequence to obtain a historical time sequence thermodynamic diagram;
the extraction unit is used for extracting image characteristic data in the historical time series thermodynamic diagram;
and the second processing unit is used for processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence.
The above apparatus, optionally, the first processing unit includes:
the construction subunit is used for constructing a data matrix based on each historical service time sequence;
and the normalization processing subunit is used for performing normalization processing on each numerical value in the data matrix to obtain a historical time sequence thermodynamic diagram.
The above apparatus, optionally, the extracting unit includes:
the input subunit is used for inputting the historical time sequence thermodynamic diagram into a feature extraction model which is trained in advance, so that the feature extraction model extracts high-dimensional features of each item of image from the historical time sequence thermodynamic diagram;
and the determining subunit is used for taking the high-dimensional features of the images as the image feature data of the historical time-series thermal image.
The above apparatus, optionally, the second processing unit includes:
the output subunit is configured to input the image feature data into a classification neural network trained in advance, so that the classification neural network outputs picture classification data of the historical timing thermodynamic diagram, where the picture classification data includes prediction information of timing traffic of each historical service time sequence;
the calling subunit is used for calling a preset regression function to carry out regression processing on the image classification data to obtain regression data corresponding to the image classification data;
and the inverse normalization processing subunit is used for performing inverse normalization processing on the regression data to obtain the predicted time sequence flow of each historical service time sequence.
The above apparatus, optionally, further comprises:
and the risk evaluation unit is used for carrying out risk evaluation on the basis of the predicted time sequence flow of each historical service time sequence so as to obtain the risk score of the service corresponding to each historical service time sequence.
The third aspect of the present invention discloses a storage medium, where the storage medium includes stored instructions, and when the instructions are executed, the storage medium controls a device in which the storage medium is located to execute the above time-series flow prediction method.
In a fourth aspect, the present invention discloses an electronic device comprising a memory, and one or more instructions, wherein the one or more instructions are stored in the memory and configured to be executed by the one or more processors to perform the method for time series traffic prediction as described above.
The invention provides a time sequence flow prediction method and device, a storage medium and an electronic device, wherein the method comprises the following steps: acquiring each historical service time sequence, and processing each historical service time sequence to obtain a historical time sequence thermodynamic diagram; extracting image characteristic data in a historical time-series thermodynamic diagram; and processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence. By converting each historical service time sequence into a historical time sequence thermodynamic diagram and extracting image characteristic data from the historical time sequence thermodynamic diagram, wherein the image characteristic data comprises various high-dimensional characteristics of the time sequence and contains global and local related characteristics of the time sequence, the predicted time sequence flow of each historical service time sequence can be obtained by processing the image characteristic data, the image characteristic data containing the high-dimensional characteristics is introduced, and the accuracy of predicting the flow of the time sequence is effectively improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in 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 embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a flowchart of a method for predicting time-series traffic according to an embodiment of the present invention;
FIG. 2 is an exemplary diagram of a historical timing thermodynamic diagram provided by an embodiment of the invention;
FIG. 3 is a flowchart of a method for extracting image feature data according to an embodiment of the present invention;
fig. 4 is a flowchart of a method for obtaining predicted time-series traffic of a historical service time-series according to an embodiment of the present invention;
fig. 5 is a flowchart of another method of a time-series flow prediction method according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a time-series flow rate prediction apparatus according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the 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 invention.
In this application, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Currently, the application demand of the industry for time-series traffic is increasing, especially for time-series data parallel prediction of multi-service scenarios. A plurality of time sequences keep certain correlation and evolve independently from each other, and each sequence has independent characteristics and consistency. Due to the characteristics of the sequence prediction of the multi-service scene, certain requirements are required for the characteristic processing and the model structure of the prediction algorithm.
The existing mainstream time sequence prediction method starts from a time sequence, the time sequence is directly used as input of model training, the extraction of characteristics is characterized by taking a time domain or a frequency domain, and meanwhile, fitting is carried out by adopting a convolutional neural network and a linear machine learning regression model by assisting in the characteristics of external influence factors. Algorithms applied by the existing model include a time sequence algorithm, a machine learning algorithm, a neural network algorithm and the like. The time sequence algorithm is similar to algorithms such as hot windows, ARIMA, MA, AR and the like, is suitable for scenes with simple rules, easily-disassembled time sequence and few external influence factors, and is generally used for data prediction which is not easily influenced by the environment. Machine learning algorithms include gbm, xgboost, gbrt and the like, and the algorithms generally construct a plurality of regression trees by a random forest regression and gradient lifting method, achieve a final prediction target by weight distribution, are suitable for scenes with more external features, and conveniently select appropriate features according to feature contribution degrees to cut out a model. The most typical neural network model is an lstm model, and based on a long-term and short-term time memory unit, the flow can be accurately fitted, but the neural network model needs a large amount of data for training, and the flow data in a real scene is not so much, so that the training is affected, and the prediction result is inaccurate.
The structures of the currently applied models can be divided into three types; single-sequence independent prediction, hierarchical recursive prediction and parallel prediction. Single-sequence independent prediction, namely, each sequence is predicted independently, the used characteristics are only from the sequence itself and are not related to the information of other sequences, and obviously, the related information between the sequences cannot be utilized in the mode; the method can utilize information of other sequences, but the characteristic acquisition mode is single, and the information is not sufficiently utilized in a scene with a plurality of sequences having complex correlation; and parallel prediction is performed, when a model is constructed, all time sequence data are introduced into the same hidden layer and output as multiple outputs, and meanwhile, the extraction and prediction of the features of a plurality of sequences are performed.
When the time sequence is predicted in the prior art, the time domain characteristics or the frequency domain characteristics of the time sequence are usually used, and when the parallel time sequence is predicted, the information among the time sequences is difficult to be correlated by using the time domain characteristics or the frequency domain characteristics, so that the accuracy of predicting the parallel time sequence is reduced.
In view of the above problems, the present invention provides a time series flow rate prediction method for predicting a time series flow rate of each time series using image feature data by processing the time series into a time series thermodynamic diagram and extracting the image feature data from the time series thermodynamic diagram; the image characteristic data comprises global and local parallel relevant characteristics of each time sequence, and the time flow of each time sequence is predicted by using the image characteristic data, so that the prediction accuracy is effectively improved.
The invention is operational with numerous general purpose or special purpose computing device environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet-type devices, multi-processor apparatus, distributed computing environments that include any of the above devices or equipment, and the like.
Referring to fig. 1, a flowchart of a method for predicting a time-series flow according to an embodiment of the present invention is specifically described as follows:
s101, acquiring each historical service time sequence.
When obtaining each historical service time series, the historical service time series can be obtained from a time series database, furthermore, a prediction instruction sent by a user can be analyzed, and each historical service time series can be obtained from the time series database based on the sequence information in the prediction instruction.
Preferably, the historical service time series may be parallel time series, and each historical service time series may be a time series generated when a user transacts services. The historical traffic time series may be a time series of different traffic.
And S102, processing each historical service time sequence to obtain a historical time sequence thermodynamic diagram.
And converting each historical service time sequence into a historical time sequence thermodynamic diagram, wherein the historical time sequence thermodynamic diagram comprises a plurality of high-dimensional image features.
The specific process of processing each historical service time sequence to obtain the historical time sequence thermodynamic diagram is as follows:
constructing a data matrix based on each historical service time sequence;
and carrying out normalization processing on each numerical value in the data matrix to obtain a historical time sequence thermodynamic diagram.
It should be noted that, when a data matrix is constructed, each historical service time sequence is intercepted based on a preset historical data window, so that an intercepting time sequence of each historical service time sequence can be obtained, and each intercepting time sequence is used for forming the data matrix; further, when intercepting the historical service time sequence, the interception may be started from any point of the historical service time sequence, or may be performed according to actual requirements, for example, the interception is started from the beginning of the historical service time sequence. When the data length of the historical service time sequence is shorter than the window length of the historical data window, zero padding operation can be carried out on the historical time sequence so as to obtain a truncated time sequence of the historical time sequence.
After obtaining the data matrix, each numerical value in the data matrix is normalized, and further, after normalizing each numerical value in the data matrix, each normalized data may be multiplied by 65536 and then converted into RGB, thereby obtaining a historical time series thermodynamic diagram, exemplarily, referring to fig. 2, an exemplary diagram of a historical time series thermodynamic diagram provided by an embodiment of the present invention is obtained, in fig. 2, time 5 is a time point to be predicted, sequences 1 to 4 are historical time sequences constituting the historical time series thermodynamic diagram, and time 1 to time 4 are time points constituting the historical time series thermodynamic diagram.
In the method provided by the embodiment of the invention, each historical time sequence is converted into a historical sequence thermodynamic diagram, and each historical time sequence can be associated, wherein the historical sequence thermodynamic diagram comprises high-dimensional information of each historical time sequence, and after each historical time sequence is converted into the historical sequence thermodynamic diagram, the high-dimensional information of each historical time sequence can be conveniently acquired from the historical sequence thermodynamic diagram.
And S103, extracting image characteristic data in the historical time-series thermodynamic diagram.
In the method provided by the embodiment of the present invention, after obtaining the historical time-series thermodynamic diagram, it is necessary to extract specific feature data from the historical time-series thermodynamic diagram, and referring to fig. 3, an exemplary diagram of a process for extracting image feature data provided by the embodiment of the present invention is specifically described as follows:
s301, inputting the historical time sequence thermodynamic diagram into a feature extraction model which is trained in advance, and enabling the feature extraction model to extract high-dimensional features of all items of images from the historical time sequence thermodynamic diagram.
It should be noted that the feature extraction model is constructed by using a neural network capable of extracting features, such as a deep residual network, a discrete Hopfield network, and so on. The feature extraction model is trained before being put into use, and after the feature extraction model is trained, the feature extraction model is put into use.
The feature extraction model extracts various high-dimensional features of the image from the historical time-series thermodynamic diagram, wherein the high-dimensional features of the image comprise, but are not limited to, color features, texture features, shape features, space features and the like of the image. Different characteristics represent different characteristics of the flow of the time series, specifically, for example, colors represent the overall trend of the flow; the texture features represent the difference characteristics between the flows; the shape characteristics represent local correlation and similarity among the flows; the spatial features characterize the overlapping and repeating characteristics of the parallel flows over the period.
And S302, taking the high-dimensional features of each image as image feature data of the historical time sequence thermal image.
In the method provided by the embodiment of the invention, the trained feature extraction model is used for processing the historical time sequence thermodynamic diagram, so that the image high-dimensional features representing different characteristics of the time sequence can be extracted from the historical time sequence thermodynamic diagram, and each image high-dimensional feature is determined as the image feature data, so that the image feature data comprises the characteristics of each historical service time sequence in different aspects, the features of the historical service time sequences are effectively correlated, and the relationship among the historical service time sequences is tighter.
And S104, processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence.
Exemplarily, referring to fig. 4, a flowchart of a method for processing image feature data to obtain a predicted time-series flow of each historical service time series according to an embodiment of the present invention is specifically described as follows:
s401, inputting image characteristic data into a classification neural network trained in advance, and enabling the classification neural network to output picture classification data of a historical time sequence thermodynamic diagram, wherein the picture classification data comprise prediction information of time sequence flow of each historical service time sequence.
The classification neural network can be a BP neural network, a CNN neural network or a multi-classification dense convolution neural network constructed by using a dense connection mechanism, wherein the dense connection mechanism can effectively relieve the gradient problem and strengthen the feature propagation. Preferably, when the convolutional neural network is constructed by using a dense connection mechanism, the gradient problem is effectively relieved, the feature propagation is enhanced, the feature administration is encouraged to greatly reduce the number of parameters, and the requirement on training samples in the network training process is reduced; furthermore, in the application process of the invention, the classification number is not lower than the numerical value after the historical service time sequence is de-duplicated.
The image feature data is processed by using a classification neural network, and image classification data is output, wherein the output image classification data can be a character string composed of numbers, and further the image classification data is discrete data.
The process of processing image feature data using a classification neural network is also a process of substantially classifying the historical time-series thermodynamic diagrams, and by classifying the historical time-series thermodynamic diagrams, it is possible to predict the time-series flow of each historical service time-series that constructs the historical time-series thermodynamic diagrams.
S402, calling a preset regression function to carry out regression processing on the image classification data to obtain regression data corresponding to the image classification data.
Preferably, the regression function may be a linear activation function, and the regression function is used to perform regression processing on the image classification data to obtain regression data, where it is to be noted that the regression data is linear continuous data.
And S403, carrying out inverse normalization processing on the regression data to obtain the predicted time sequence flow of each historical service time sequence.
In the method provided by the embodiment of the invention, the classified neural network is used for processing the image characteristic data, and high-precision fitting can be performed, so that prediction data with high precision can be obtained, and the prediction result of time accumulation can be improved.
Preferably, after the predicted time sequence flow of each historical service time sequence is obtained, a preset risk assessment mechanism may be used to perform risk assessment operation based on the predicted time sequence flow of each historical service time sequence, so as to obtain a risk score of the service corresponding to each historical service time sequence, where it is to be noted that the risk score is a score of the service at a time point corresponding to the predicted time sequence flow, and the risk score may be used to represent a risk degree of the service at the time point corresponding to the predicted time sequence flow, and a worker may arrange service handling at the time point, maintenance work on equipment, and the like according to the predicted time flow and the risk score. The risk assessment of the business can be beneficial to the arrangement of various works by workers, so that various risks are avoided, and a good business handling environment is provided for clients.
In the method provided by the embodiment of the invention, each historical service time sequence is obtained and processed to obtain a historical time sequence thermodynamic diagram; extracting image characteristic data in a historical time-series thermodynamic diagram; and processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence. By converting each historical service time sequence into a historical time sequence thermodynamic diagram and extracting image characteristic data from the historical time sequence thermodynamic diagram, wherein the image characteristic data comprises various high-dimensional characteristics of the time sequence and contains global and local related characteristics of the time sequence, the predicted time sequence flow of each historical service time sequence can be obtained by processing the image characteristic data, the image characteristic data containing the high-dimensional characteristics is introduced, and the accuracy of predicting the flow of the time sequence is effectively improved.
Referring to fig. 5, an application example diagram of the time-series flow prediction method according to the embodiment of the present invention is specifically described as follows:
acquiring M time sequences from historical data, wherein the time sequences can be understood as historical service time sequences in the above; it should be noted that the data length of each time series is N, and each time series is normalized to obtain a normalized data matrix, and further, the data matrix may also be referred to as an M × N time sequence matrix. Performing imaging processing on data evidence, specifically, multiplying each numerical value in a data matrix by 65536, converting the numerical value into RGB (red, green and blue), so as to obtain a historical time sequence thermodynamic diagram, when performing feature processing on the historical time sequence thermodynamic diagram, processing the historical time sequence thermodynamic diagram by using a depth residual error network, so as to obtain image feature data, performing classification processing based on the image feature data, and performing classification processing by using a multi-classification dense layer, so as to realize fine classification, and performing regression processing on the classified data by using a linear planning function, so as to regress the discretely classified data into linear continuous data; and performing inverse normalization processing on the data obtained by the regression processing, thereby obtaining prediction data containing the predicted time-series flow of each time series.
It should be noted that, in the process of converting the historical service time sequence into the historical time sequence thermodynamic diagram, values of the time sequences are arranged in parallel, and the values are pixelized, that is, each pixel on the image is corresponding to a value at a different time point of each sequence, so that the historical time sequence thermodynamic diagram can be obtained, specifically, as shown in fig. 2, the historical time sequence thermodynamic diagram of fig. 2 is composed of a plurality of picture blocks, and further, each picture block has a corresponding time sequence and time point.
Extracting features from the historical time-series thermodynamic diagram by using a residual neural network, extracting overall features from each picture block, and extracting local features from the pixel distribution of the historical time-series thermodynamic diagram, so that image feature data including the overall features and the local features can be obtained, further, the image feature data includes but is not limited to color features, texture features, shape features, space features and the like of the historical time-series thermodynamic diagram, and specifically, the color features such as histograms, color distribution and other global features can represent surface properties of regions of an image; the texture features such as gray level co-occurrence matrix are used as global features, and can well resist the influence of noise; shape features such as contours, regions may describe local traits of the image information; spatial features such as overlapping of regions, orientation, can distinguish flow conditions in different regions. Further, when the residual error neural network is used for extracting features from the historical time-series thermodynamic diagram, a short connection to the output of the nonlinear layer can be directly introduced from the input, so that a better fitting classification function is realized, and higher classification accuracy is obtained.
And processing the image characteristic data by using a multi-classification dense convolutional neural network constructed by a dense connection mechanism so as to obtain image classification data containing prediction information of time sequence flow of each time sequence, wherein the multi-classification dense convolutional neural network can be subjected to fine classification so as to obtain a more accurate prediction result. After the image classification data is obtained, regression processing is carried out on the image classification data by using a linear activation function, so that the image classification data of the discrete classification is regressed into linear continuous regression data, and inverse normalization processing is carried out on the regression data, so that prediction data of the prediction time sequence flow of each time sequence is obtained.
According to the invention, the time sequence is converted into the thermal image, and then the high-dimensional relevant feature extraction is carried out on the thermal image, so that the local and global features of the time sequence are covered, the high-precision fitting prediction is carried out on the future time point data, and the prediction accuracy of the time sequence flow of the time sequence is improved; besides, the method can be applied to various scenes for prediction, and the applicability of the scenes is expanded.
Corresponding to fig. 1, an embodiment of the present invention further provides a time-series flow rate prediction apparatus, which is used to support the application of the method shown in fig. 1 in real life, and the apparatus may be disposed in an intelligent computing terminal or a distributed computing terminal. Referring to fig. 6, a schematic structural diagram of a time-series flow rate prediction apparatus provided in an embodiment of the present invention is specifically described as follows:
an obtaining unit 601, configured to obtain each historical service time series;
a first processing unit 602, configured to process each historical service time series to obtain a historical time sequence thermodynamic diagram;
an extracting unit 603 configured to extract image feature data in the history time-series thermodynamic diagram;
a second processing unit 604, configured to process the image feature data to obtain a predicted time sequence flow of each historical service time sequence.
In the device provided by the embodiment of the invention, each historical service time sequence is obtained and processed to obtain a historical time sequence thermodynamic diagram; extracting image characteristic data in a historical time-series thermodynamic diagram; and processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence. By converting each historical service time sequence into a historical time sequence thermodynamic diagram and extracting image characteristic data from the historical time sequence thermodynamic diagram, wherein the image characteristic data comprises various high-dimensional characteristics of the time sequence and contains global and local related characteristics of the time sequence, the predicted time sequence flow of each historical service time sequence can be obtained by processing the image characteristic data, the image characteristic data containing the high-dimensional characteristics is introduced, and the accuracy of predicting the flow of the time sequence is effectively improved.
In the apparatus provided in the embodiment of the present invention, the first processing unit 602 may be configured to:
the construction subunit is used for constructing a data matrix based on each historical service time sequence;
and the normalization processing subunit is used for performing normalization processing on each numerical value in the data matrix to obtain a historical time sequence thermodynamic diagram.
In the apparatus provided in the embodiment of the present invention, the extracting unit 603 may be configured to:
the input subunit is used for inputting the historical time sequence thermodynamic diagram into a feature extraction model which is trained in advance, so that the feature extraction model extracts high-dimensional features of each item of image from the historical time sequence thermodynamic diagram;
and the determining subunit is used for taking the high-dimensional features of the images as the image feature data of the historical time-series thermal image.
In the apparatus provided in the embodiment of the present invention, the second processing unit 604 may be configured to:
the output subunit is configured to input the image feature data into a classification neural network trained in advance, so that the classification neural network outputs picture classification data of the historical timing thermodynamic diagram, where the picture classification data includes prediction information of timing traffic of each historical service time sequence;
the calling subunit is used for calling a preset regression function to carry out regression processing on the image classification data to obtain regression data corresponding to the image classification data;
and the inverse normalization processing subunit is used for performing inverse normalization processing on the regression data to obtain the predicted time sequence flow of each historical service time sequence.
The device provided by the embodiment of the invention further comprises:
and the risk evaluation unit is used for carrying out risk evaluation on the basis of the predicted time sequence flow of each historical service time sequence so as to obtain the risk score of the service corresponding to each historical service time sequence.
The embodiment of the present invention further provides a storage medium, where the storage medium includes a stored instruction, where when the instruction runs, the apparatus where the storage medium is located is controlled to perform the following operations:
acquiring each historical service time sequence;
processing each historical service time sequence to obtain a historical time sequence thermodynamic diagram;
extracting image characteristic data in the historical time-series thermodynamic diagram;
and processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence.
An electronic device is provided in an embodiment of the present invention, and its structural schematic diagram is shown in fig. 7, which specifically includes a memory 701 and one or more instructions 702, where the one or more instructions 702 are stored in the memory 701, and are configured to be executed by one or more processors 603 to perform the following operations for the one or more instructions 702:
acquiring each historical service time sequence;
processing each historical service time sequence to obtain a historical time sequence thermodynamic diagram;
extracting image characteristic data in the historical time-series thermodynamic diagram;
and processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence.
The specific implementation procedures and derivatives thereof of the above embodiments are within the scope of the present invention.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the system or system embodiments, which are substantially similar to the method embodiments, are described in a relatively simple manner, and reference may be made to some descriptions of the method embodiments for relevant points. The above-described system and system embodiments are only illustrative, wherein the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement without inventive effort.
Those of skill would further appreciate that the various illustrative elements 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 invention.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for time series traffic prediction, comprising:
acquiring each historical service time sequence;
processing each historical service time sequence to obtain a historical time sequence thermodynamic diagram;
extracting image characteristic data in the historical time-series thermodynamic diagram;
and processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence.
2. The method of claim 1, wherein the processing each historical traffic time series to obtain a historical time series thermodynamic diagram comprises:
constructing a data matrix based on each historical service time sequence;
and carrying out normalization processing on each numerical value in the data matrix to obtain a historical time sequence thermodynamic diagram.
3. The method of claim 1, wherein the extracting image feature data in the historical time-series thermodynamic diagram comprises:
inputting the historical time sequence thermodynamic diagram into a feature extraction model which is trained in advance, so that the feature extraction model extracts high-dimensional features of each item of image from the historical time sequence thermodynamic diagram;
and taking each high-dimensional feature of the image as image feature data of the historical time-series thermal image.
4. The method of claim 1, wherein said processing said image characteristic data to obtain a predicted time series flow rate for each said historical traffic time series comprises:
inputting the image characteristic data into a classification neural network trained in advance, so that the classification neural network outputs picture classification data of the historical time sequence thermodynamic diagram, wherein the picture classification data comprises prediction information of time sequence flow of each historical service time sequence;
calling a preset regression function to carry out regression processing on the image classification data to obtain regression data corresponding to the image classification data;
and carrying out inverse normalization processing on the regression data to obtain the predicted time sequence flow of each historical service time sequence.
5. The method of claim 1, further comprising:
and performing risk assessment based on the predicted time sequence flow of each historical service time sequence to obtain a risk score of the service corresponding to each historical service time sequence.
6. A time series flow prediction apparatus, comprising:
the acquisition unit is used for acquiring each historical service time sequence;
the first processing unit is used for processing each historical service time sequence to obtain a historical time sequence thermodynamic diagram;
an extraction unit, configured to extract image feature data in the historical time-series thermodynamic diagram;
and the second processing unit is used for processing the image characteristic data to obtain the predicted time sequence flow of each historical service time sequence.
7. The apparatus of claim 6, wherein the first processing unit comprises:
the construction subunit is used for constructing a data matrix based on each historical service time sequence;
and the normalization processing subunit is used for performing normalization processing on each numerical value in the data matrix to obtain a historical time sequence thermodynamic diagram.
8. The apparatus of claim 6, wherein the extraction unit comprises:
the input subunit is used for inputting the historical time sequence thermodynamic diagram into a feature extraction model which is trained in advance, so that the feature extraction model extracts high-dimensional features of each item of image from the historical time sequence thermodynamic diagram;
and the determining subunit is used for taking each item of image high-dimensional feature as image feature data of the historical time-series thermal image.
9. A storage medium comprising stored instructions, wherein the instructions, when executed, control a device on which the storage medium is located to perform the time-series flow prediction method according to any one of claims 1 to 5.
10. An electronic device comprising a memory and one or more instructions, wherein the one or more instructions are stored in the memory and configured to be executed by the one or more processors to perform the method of time series flow prediction according to any one of claims 1-5.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210203124.4A CN114627330A (en) | 2022-03-02 | 2022-03-02 | Time sequence flow prediction method and device, storage medium and electronic equipment |
| PCT/CN2022/127217 WO2023165145A1 (en) | 2022-03-02 | 2022-10-25 | Time sequence traffic prediction method and apparatus, storage medium, and electronic device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210203124.4A CN114627330A (en) | 2022-03-02 | 2022-03-02 | Time sequence flow prediction method and device, storage medium and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114627330A true CN114627330A (en) | 2022-06-14 |
Family
ID=81899895
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210203124.4A Pending CN114627330A (en) | 2022-03-02 | 2022-03-02 | Time sequence flow prediction method and device, storage medium and electronic equipment |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN114627330A (en) |
| WO (1) | WO2023165145A1 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023165145A1 (en) * | 2022-03-02 | 2023-09-07 | 北京沃东天骏信息技术有限公司 | Time sequence traffic prediction method and apparatus, storage medium, and electronic device |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109995562B (en) * | 2017-12-30 | 2022-11-08 | 中国移动通信集团河北有限公司 | Network traffic prediction method, device, equipment and medium |
| US20200402665A1 (en) * | 2019-06-19 | 2020-12-24 | GE Precision Healthcare LLC | Unplanned readmission prediction using an interactive augmented intelligent (iai) system |
| CN110674019B (en) * | 2019-08-30 | 2023-02-07 | 中国人民财产保险股份有限公司 | Method and device for predicting system fault and electronic equipment |
| CN113259950B (en) * | 2021-05-13 | 2022-03-11 | 重庆邮电大学 | Low-orbit satellite spot beam closing method based on service prediction |
| CN114118622A (en) * | 2021-12-08 | 2022-03-01 | 北京北大软件工程股份有限公司 | Data trend prediction method and system based on time series |
| CN114627330A (en) * | 2022-03-02 | 2022-06-14 | 北京沃东天骏信息技术有限公司 | Time sequence flow prediction method and device, storage medium and electronic equipment |
-
2022
- 2022-03-02 CN CN202210203124.4A patent/CN114627330A/en active Pending
- 2022-10-25 WO PCT/CN2022/127217 patent/WO2023165145A1/en unknown
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023165145A1 (en) * | 2022-03-02 | 2023-09-07 | 北京沃东天骏信息技术有限公司 | Time sequence traffic prediction method and apparatus, storage medium, and electronic device |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2023165145A1 (en) | 2023-09-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108133013B (en) | Information processing method, information processing device, computer equipment and storage medium | |
| CN111614690B (en) | Abnormal behavior detection method and device | |
| CN111177473B (en) | Personnel relationship analysis method, device and readable storage medium | |
| CN109583904A (en) | Training method, impaired operation detection method and the device of abnormal operation detection model | |
| US11062240B2 (en) | Determining optimal workforce types to fulfill occupational roles in an organization based on occupational attributes | |
| US20070015125A1 (en) | Automated updating of job assessments | |
| CN112995414B (en) | Behavior quality inspection method, device, equipment and storage medium based on voice call | |
| KR102195629B1 (en) | Method for selecting workers based on capability of work in crowdsourcing based projects for artificial intelligence training data generation | |
| US10762089B2 (en) | Open ended question identification for investigations | |
| CN114627330A (en) | Time sequence flow prediction method and device, storage medium and electronic equipment | |
| US11295325B2 (en) | Benefit surrender prediction | |
| US20200380376A1 (en) | Artificial Intelligence Based System And Method For Predicting And Preventing Illicit Behavior | |
| CN114493686A (en) | Operation content generation and pushing method and device | |
| KR102155748B1 (en) | Method for automatic apdating inspection unit price using work time versus inspection time of crowdsourcing based project for artificial intelligence training data generation | |
| CN116501979A (en) | Information recommendation method, information recommendation device, computer equipment and computer readable storage medium | |
| CN114581130A (en) | Bank website number assigning method and device based on customer portrait and storage medium | |
| WO2023049280A1 (en) | Systems and methods to screen a predictive model for risks of the predictive model | |
| EP3493082A1 (en) | A method of exploring databases of time-stamped data in order to discover dependencies between the data and predict future trends | |
| CN115545088A (en) | Model construction method, classification method and device and electronic equipment | |
| Chen et al. | Social-network-assisted task recommendation algorithm in mobile crowd sensing | |
| US20200382530A1 (en) | Unequal probability sampling based on a likelihood model score to evaluate prevalence of inappropriate entities | |
| CN116431319B (en) | Task processing method and device | |
| CN116664016B (en) | Screening method and device of ESG (electronic service guide) sub-topics, electronic equipment and readable storage medium | |
| CN117829869A (en) | Potential customer determination method and device, storage medium and electronic equipment | |
| CN112308404A (en) | Project risk management method and device, electronic equipment and storage medium |
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 |