CN109508461A - Order price prediction technique, terminal and medium based on Chinese natural language processing - Google Patents
Order price prediction technique, terminal and medium based on Chinese natural language processing Download PDFInfo
- Publication number
- CN109508461A CN109508461A CN201811644617.1A CN201811644617A CN109508461A CN 109508461 A CN109508461 A CN 109508461A CN 201811644617 A CN201811644617 A CN 201811644617A CN 109508461 A CN109508461 A CN 109508461A
- Authority
- CN
- China
- Prior art keywords
- convolutional neural
- neural networks
- networks model
- price
- transaction information
- 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
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000003058 natural language processing Methods 0.000 title claims abstract description 16
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 58
- 239000013598 vector Substances 0.000 claims abstract description 21
- 239000011159 matrix material Substances 0.000 claims abstract description 12
- 238000013528 artificial neural network Methods 0.000 claims abstract description 6
- 238000002203 pretreatment Methods 0.000 claims abstract description 4
- 238000004590 computer program Methods 0.000 claims description 13
- 238000003860 storage Methods 0.000 claims description 13
- 238000001914 filtration Methods 0.000 claims description 10
- 230000006870 function Effects 0.000 claims description 10
- 238000003475 lamination Methods 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 6
- 238000011156 evaluation Methods 0.000 abstract description 6
- 238000013135 deep learning Methods 0.000 abstract description 4
- 238000012549 training Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 5
- 238000003062 neural network model Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000008676 import Effects 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000010977 unit operation Methods 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 238000005498 polishing Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Software Systems (AREA)
- Strategic Management (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Economics (AREA)
- Molecular Biology (AREA)
- Human Resources & Organizations (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Machine Translation (AREA)
Abstract
Order price prediction technique, terminal and medium provided by the invention based on Chinese natural language processing, this method obtain History Order Transaction Information；Execute pre-treatment step；Construct convolutional neural networks model；Study, and the initialization weight matrix as convolutional neural networks model are indicated to preprocessed data；Initialization weight matrix is predicted using convolutional neural networks model, to obtain prediction result；Reverse train is carried out according to prediction result and legitimate reading, to obtain best convolutional neural networks model；Item description document to be predicted is received, prediction item description document is treated using best convolutional neural networks model and is predicted, obtain forecast price.This method is compared with existing manual evaluation order Price Method, natural language processing is done using deep learning neural network, indicates semantic with mathematic vector, so that convolutional neural networks model is more accurate to the identification of order semanteme, convolutional neural networks modelling effect is more preferable, fast speed.
Description
Technical field
The invention belongs to technical field of information processing, and in particular to the order price expectation based on Chinese natural language processing
Method, terminal and medium.
Background technique
The existing method for describing document forecast price according to user items, is often manually commented using business personnel
Estimate, awareness granularity is thicker, inefficiency, and accuracy rate is limited by business personnel's knowledge background and professional ability.And business personnel is very
Hardly possible reads whole History Orders, it is also difficult to remember hundreds of classification price, can come with some shortcomings in price expectation.
In the understanding that user items describe document content, Text Classification, including DNN, CNN, RNN at present, this is several
Kind of technology has been quite mature technology, but is not applied to that analysis expression way is complex, is difficult to structuring
Item description document in.
Summary of the invention
For the defects in the prior art, the present invention provides a kind of order price expectation based on Chinese natural language processing
Method, terminal and medium improve the speed and accuracy of order price expectation.
In a first aspect, a kind of order price prediction technique based on Chinese natural language processing, comprising the following steps:
Obtain History Order Transaction Information；
It executes pre-treatment step: the History Order Transaction Information being pre-processed, to obtain preprocessed data；
Construct convolutional neural networks model；
Study is indicated to the preprocessed data, and the matrix that study is obtained is as the convolutional neural networks mould
The initialization weight matrix of type；
Initialization weight matrix is predicted using convolutional neural networks model, to obtain prediction result；
According to corresponding legitimate reading in the prediction result and History Order Transaction Information to the convolutional neural networks
Model carries out reverse train, to obtain best convolutional neural networks model；
Receive item description document to be predicted, using best convolutional neural networks model treat prediction item description document into
Row prediction, obtains forecast price.
Preferably, described pre-process includes:
Construct multiple price ranges；
The concluded price of order each in History Order Transaction Information is mapped in corresponding price range, label strikes a bargain
The classification of the price range of price mapping.
Preferably, this method is after the classification of the price range of the label concluded price mapping, further includes:
Default filtering rule；
The History Order Transaction Information is filtered according to the filtering rule.
Preferably, the History Order Transaction Information includes item description document；
This method it is described the History Order Transaction Information is filtered according to the filtering rule after, also wrap
It includes:
Filtered item description document is segmented, to obtain participle phrase；
Word lists are established according to participle phrase, the number of each participle phrase is set.
Preferably, the convolutional neural networks model includes embeding layer, three-layer coil lamination, global pool layer and full connection
Layer；
The embeding layer converts expression study tensor for item description document using the number of the participle phrase, should
Indicate that study tensor is denoted as N*K dimension；
The three-layer coil lamination use S M1*K convolution kernel, respectively 0 dimension, 1 dimension ..., M1-1 tie up on to the M1*K
Dimension table dendrography practises tensor and does convolution, obtains the tensor of S N*1 dimension；The convolution kernel tieed up respectively using M2*1 and M3*1 again is to described
The tensor of N*1 dimension carries out convolution；
The global pool layer carries out global maximum pond to the tensor of the three-layer coil lamination final output, obtains text
This vector；
The text vector is connected by the full articulamentum with preset two layers full Connection Neural Network, and final output is multiple
Price.
Preferably, it is described according to corresponding legitimate reading in the prediction result and History Order Transaction Information to the volume
Product neural network model carries out reverse train, is specifically included with obtaining best convolutional neural networks model:
Cross entropy is calculated according to legitimate reading corresponding in the prediction result and History Order Transaction Information, is defined as damaging
Lose function；
Reverse train is carried out using Adam and SGD optimization method, until the loss function is minimum, to obtain best convolution
Neural network model.
Preferably, this method it is described to obtain best convolutional neural networks model after, further includes:
Delta Time is set；
When Delta Time reaches, the History Order Transaction Information in the Delta Time is obtained, returns and executes the pre- place
Step is managed, the best convolutional neural networks model is optimized.
Second aspect, a kind of terminal, including processor, input equipment, output equipment and memory, it is the processor, defeated
Enter equipment, output equipment and memory to be connected with each other, wherein the memory is for storing computer program, the computer
Program includes program instruction, and the processor is configured for calling described program instruction, executes method described in first aspect.
The third aspect, a kind of computer readable storage medium, the computer storage medium are stored with computer program, institute
Stating computer program includes program instruction, and described program instruction makes the processor execute first aspect when being executed by a processor
The method.
As shown from the above technical solution, the order price expectation side provided by the invention based on Chinese natural language processing
Method, terminal and medium, have the advantage that
1, the present invention converts classification problem for the problem, uses depth by the distribution characteristics of analysis user price data
Learning method is spent, learns training convolutional neural networks model from the item description document of user, realizes the intelligence of item price
Prediction.
2, word is indicated study and is mapped to the dense vector space of low level by the present invention, so that in similar import, semanteme
Hold similar word being closer in vector space, and distance is farther out in space for the farther away word of semantic distance.It is learning
Practise word expression study when, using Word2vec term vector indicate learning method, study to an initial feature word to
Amount, then using the vector as initial parameter, bring convolutional neural networks model into and optimize.After being indicated study to word,
It can be a tensor by each item description document representation, learn the potential language of the tensor using multichannel convolutional neural networks
Justice, and entire semantic meaning is ultimately expressed as a vector, it is done and is classified using a full articulamentum.
3, the invention does nature language using deep learning neural network compared with existing manual evaluation order Price Method
Speech processing indicates semantic with mathematic vector, so that convolutional neural networks model is more accurate to the identification of order semanteme, convolutional Neural
Network model effect is more preferable.
4, the invention is compared with existing manual evaluation order Price Method, fast speed.In existing method, manual evaluation
The average every single needs 10 minutes or so of order price, and an orders can be predicted at 200 milliseconds of single-unit operation or so in the present invention
Price, is run if disposed using large-scale cluster, and efficiency will also greatly improve.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art are briefly described.In all the appended drawings, similar element
Or part is generally identified by similar appended drawing reference.In attached drawing, each element or part might not be drawn according to actual ratio.
Fig. 1 is the flow chart for the order price prediction technique that the embodiment of the present invention one provides.
Fig. 2 is the schematic diagram of the expression study for the Skip-Gram model that the embodiment of the present invention one provides.
Fig. 3 is the module frame chart of terminal provided by Embodiment 2 of the present invention.
Specific embodiment
It is described in detail below in conjunction with embodiment of the attached drawing to technical solution of the present invention.Following embodiment is only used for
Clearly illustrate technical solution of the present invention, therefore be only used as example, and cannot be used as a limitation and limit protection model of the invention
It encloses.It should be noted that unless otherwise indicated, technical term or scientific term used in this application are should be belonging to the present invention
The ordinary meaning that field technical staff is understood.
Embodiment one:
A kind of order price prediction technique based on Chinese natural language processing, referring to Fig. 1, comprising the following steps:
S1: History Order Transaction Information is obtained；
The History Order Transaction Information includes item description document, includes Project Introduction, price in item description document
Information etc..
S2: it executes pre-treatment step: the History Order Transaction Information being pre-processed, to obtain preprocessed data；
In order to realize the association learnt from item description document with price, data are pre-processed first, are pre-processed
Including price mapping, filtering and participle.
1) price, which maps, includes:
Construct multiple price ranges；Price range can be formulated according to the distribution of price, due to the transaction value of project
The effect " by whole aggregation " is showed, therefore this method can formulate suitable price range according to this effect.Wherein referring to
When determining price range, every two can be set and lean on the median of full price lattice for the boundary line of price range.
The concluded price of order each in History Order Transaction Information is mapped in corresponding price range, label strikes a bargain
The classification of the price range of price mapping.History Order Transaction Information can be classified according to price in this way, such as: price
Section can be set to (0,100], (100,200], (200,500], (and 500,1000] etc..
2) it filters and includes:
Default filtering rule；
The History Order Transaction Information is filtered according to the filtering rule.
Specifically, since History Order Transaction Information has noise, this method filters out a batch by providing filtering rule
Data.Such as: bound or the number of words of price are set to filter out partial data, for example filter out valence now in " software development " class
Data of the lattice less than 100.Data very short, less than 20 words are either such as also described for some empty data, are filtered out,
In order to avoid the training of interference convolutional neural networks model.
3) participle includes:
Filtered item description document is segmented, to obtain participle phrase；
Word lists are established according to participle phrase, the number of each participle phrase is set.
Specifically, the participle tool that open source can be used segments the item description document of user.After participle, build
Vertical word lists give the setting of each participle phrase unique number, such as ID number.During participle, it can also be closed by addition
Key field come to participle phrase distinguish.Such as: one ' UNK ' and ' PAD ' two fields are added, are not occurred in word lists
The word crossed indicates with ' UNK ', and the inadequate phrase of length uses ' PAD ' polishing, two fields respectively indicate word be not present and
Sentence zero padding.
S3: building convolutional neural networks model；
The convolutional neural networks model includes embeding layer, three-layer coil lamination, global pool layer and full articulamentum；
The embeding layer converts expression study tensor for item description document using the number of the participle phrase, should
Indicate that study tensor is denoted as N*K dimension；
The three-layer coil lamination use S M1*K convolution kernel, respectively 0 dimension, 1 dimension ..., M1-1 tie up on to the M1*K
Dimension table dendrography practises tensor and does convolution, obtains the tensor of S N*1 dimension；The convolution kernel tieed up respectively using M2*1 and M3*1 again is to described
The tensor of N*1 dimension carries out convolution；
Specifically, if M1 is 2, the convolution kernel of S 2*K is used, is tieed up respectively in 0 peacekeeping 1 to the 2*K dimension table dendrography
It practises tensor and does convolution.M2 and M3 represents different size of convolution kernel, can extract the information of different length, it is therefore an objective in the base of M1
More details are extracted on plinth.
The global pool layer carries out global maximum pond to the tensor of the three-layer coil lamination final output, obtains text
This vector；
Specifically, unanimously, global maximum Chi Huahui is more from each for the size and tensor (multidimensional) of global maximum pond layer
Maximum one is chosen in dimension tensor, therefore becomes one-dimensional vector after exporting.
The text vector is connected by the full articulamentum with preset two layers full Connection Neural Network, final output layer
Number of nodes is identical with the interval number of price, and then by Softmax function, select probability maximum one is used as the prediction valence
Lattice.
S4: study is indicated to the preprocessed data, and the matrix that study is obtained is as the convolutional Neural net
The initialization weight matrix of network model；
Specifically, this method can be indicated to word using the method and Skip-Gram model of Word2vec
It practises, predicts context using word, as shown in Fig. 2, the word vectors expression that study one is initial, and as convolutional neural networks
The initialization weight matrix of model insertion layer.
S5: initialization weight matrix is predicted using convolutional neural networks model, to obtain prediction result；
S6: according to the prediction result Y/With legitimate reading Y corresponding in History Order Transaction Information to the convolution mind
Reverse train is carried out through network model, to obtain best convolutional neural networks model；
Specifically, it is described according to corresponding legitimate reading in the prediction result and History Order Transaction Information to the volume
Product neural network model carries out reverse train, is specifically included with obtaining best convolutional neural networks model:
Cross entropy is calculated according to legitimate reading corresponding in the prediction result and History Order Transaction Information, is defined as damaging
Function is lost, the loss of convolutional neural networks model is obtained；
Reverse train is carried out using Adam and SGD optimization method, until the loss function is minimum, to obtain best convolution
Neural network model.
Specifically, following methods can be used in direction training: pray to Gods for blessing since the loss function finally exported first through
The derivative of network parameter (weight w and biasing b), it is defeated until convolutional neural networks model along the value of the reversed undated parameter of derivative
Enter layer, then calculate loss function from input layer to output layer again, in cycles, until loss function minimum, gradient is transmitted
To each layer of convolutional neural networks model parameter, be updated study (such as:It means to having
The propagation formula derivation of w is simultaneously multiplied by a learning rate, then with current parameter wi-1It subtracts and just obtains updated parameter wi)。
S7: receiving item description document to be predicted, treats prediction item description text using best convolutional neural networks model
Shelves are predicted, forecast price is obtained.
It specifically,, will be to be predicted when needing to predict order price after convolutional neural networks model construction is good
Item description document, which is directly inputted in convolutional neural networks model, carries out price expectation.
S8: increment optimization is carried out.Include:
Delta Time is set；
When Delta Time reaches, the History Order Transaction Information in the Delta Time is obtained, returns and executes the pre- place
Step is managed, the best convolutional neural networks model is optimized.
Specifically, after convolutional neural networks model training is good, every setting time (such as Delta Time is one week), product
After tiring out certain data, increment optimization is carried out, it is ensured that convolutional neural networks model can learn to optimize with business development.
Increment optimization can use following methods: by collecting the data that business is accumulative on line, (history i.e. in Delta Time is ordered first
Single Transaction Information), the parameter and structure for then loading existing convolutional neural networks model continue training pattern, after being optimized
Convolutional neural networks model continues to treat prediction item description document progress price according to the convolutional neural networks model after optimization
Prediction.
Method of the invention is converted classification problem for the problem, is made by the distribution characteristics of analysis user price data
With deep learning method, learns training convolutional neural networks model from the item description document of user, realize item price
Intelligent predicting.Word is indicated study and is mapped to the dense vector space of low level, so that similar import, semantic content are similar
Word being closer in vector space, and distance is farther out in space for the farther away word of semantic distance.In study word
When indicating study, learning method is indicated using Word2vec term vector, is learnt to an initial feature word vectors, then should
Vector is brought convolutional neural networks model into and is optimized as initial parameter.It, can will be each after being indicated study to word
Item description document representation is a tensor, learns the potential applications of the tensor using multichannel convolutional neural networks, and will
Entire semantic meaning is ultimately expressed as a vector, is done and is classified using a full articulamentum.
This method does natural language compared with existing manual evaluation order Price Method, using deep learning neural network
Processing indicates semantic with mathematic vector, so that convolutional neural networks model is more accurate to the identification of order semanteme, convolutional Neural net
Network modelling effect is more preferable.Fast speed.In existing method, the average every single needs 10 minutes or so of manual evaluation order price, and
An order price can be predicted at 200 milliseconds of single-unit operation or so in the present invention, runs if disposed using large-scale cluster, effect
Rate will also greatly improve.
Embodiment two:
A kind of terminal, referring to Fig. 3, including processor 801, input equipment 802, output equipment 803 and memory 804, institute
It states processor 801, input equipment 802, output equipment 803 and memory 804 to be connected with each other by bus 805, wherein described to deposit
Reservoir 804 is for storing computer program, and the computer program includes program instruction, and the processor 801 is configured for
Described program instruction is called, above-mentioned method is executed.
In the specific implementation, terminal described in the embodiment of the present invention is including but not limited to such as with touch sensitive surface
The mobile phone, laptop computer or tablet computer of (for example, touch-screen display and/or touch tablet) etc it is other just
Portable device.It is to be further understood that in certain embodiments, the equipment is not portable communication device, but there is touching
Touch the desktop computer of sensing surface (for example, touch-screen display and/or touch tablet).
In following discussion, the terminal including display and touch sensitive surface is described.It is, however, to be understood that
It is that terminal may include one or more of the other physical user-interface device of such as physical keyboard, mouse and/or control-rod.
Terminal supports various application programs, such as one of the following or multiple: drawing application program, demonstration application journey
Sequence, word-processing application, website create application program, disk imprinting application program, spreadsheet applications, game application
Program, telephony application, videoconference application, email application, instant messaging applications, exercise
Support application program, photo management application program, digital camera application program, digital camera application program, web-browsing application
Program, digital music player application and/or video frequency player application program.
The various application programs that can be executed at the terminal can be used such as touch sensitive surface at least one is public
Physical user-interface device.It can adjust and/or change among applications and/or in corresponding application programs and touch sensitive table
The corresponding information shown in the one or more functions and terminal in face.In this way, the public physical structure of terminal is (for example, touch
Sensing surface) it can support the various application programs with user interface intuitive and transparent for a user.
It should be appreciated that in embodiments of the present invention, alleged processor 801 can be central processing unit (Central
Processing Unit, CPU), which can also be other general processors, digital signal processor (Digital
Signal Processor, DSP), specific integrated circuit (Application Specific Integrated Circuit,
ASIC), ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic
Device, discrete gate or transistor logic, discrete hardware components etc..General processor can be microprocessor or this at
Reason device is also possible to any conventional processor etc..
Input equipment 802 may include that Trackpad, fingerprint adopt sensor (for acquiring the finger print information and fingerprint of user
Directional information), microphone etc., output equipment 803 may include display (LCD etc.), loudspeaker etc..
The memory 804 may include read-only memory and random access memory, and to processor 801 provide instruction and
Data.The a part of of memory 804 can also include nonvolatile RAM.For example, memory 804 can also be deposited
Store up the information of device type.
Terminal provided by the embodiment of the present invention, to briefly describe, embodiment part does not refer to place, can refer to aforementioned side
Corresponding contents in method embodiment.
Embodiment three:
A kind of computer readable storage medium, the computer storage medium are stored with computer program, the computer
Program includes program instruction, and described program instruction makes the processor execute above-mentioned method when being executed by a processor.
The computer readable storage medium can be the internal storage unit of terminal described in aforementioned any embodiment, example
Such as the hard disk or memory of terminal.The computer readable storage medium is also possible to the External memory equipment of the terminal, such as
The plug-in type hard disk being equipped in the terminal, intelligent memory card (Smart Media Card, SMC), secure digital (Secure
Digital, SD) card, flash card (Flash Card) etc..Further, the computer readable storage medium can also be wrapped both
The internal storage unit for including the terminal also includes External memory equipment.The computer readable storage medium is described for storing
Other programs and data needed for computer program and the terminal.The computer readable storage medium can be also used for temporarily
When store the data that has exported or will export.
Terminal provided by the embodiment of the present invention, to briefly describe, embodiment part does not refer to place, can refer to aforementioned side
Corresponding contents in method embodiment.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme should all cover within the scope of the claims and the description of the invention.
Claims (9)
1. a kind of order price prediction technique based on Chinese natural language processing, which comprises the following steps:
Obtain History Order Transaction Information；
It executes pre-treatment step: the History Order Transaction Information being pre-processed, to obtain preprocessed data；
Construct convolutional neural networks model；
Study is indicated to the preprocessed data, and the matrix that study is obtained is as the convolutional neural networks model
Initialize weight matrix；
Initialization weight matrix is predicted using convolutional neural networks model, to obtain prediction result；
According to corresponding legitimate reading in the prediction result and History Order Transaction Information to the convolutional neural networks model
Reverse train is carried out, to obtain best convolutional neural networks model；
Item description document to be predicted is received, prediction item description document is treated using best convolutional neural networks model and carries out in advance
It surveys, obtains forecast price.
2. the order price prediction technique according to claim 1 based on Chinese natural language processing, which is characterized in that described
Pretreatment includes:
Construct multiple price ranges；
The concluded price of order each in History Order Transaction Information is mapped in corresponding price range, concluded price is marked
The classification of the price range of mapping.
3. the order price prediction technique according to claim 2 based on Chinese natural language processing, which is characterized in that the party
Method is after the classification of the price range of the label concluded price mapping, further includes:
Default filtering rule；
The History Order Transaction Information is filtered according to the filtering rule.
4. the order price prediction technique according to claim 3 based on Chinese natural language processing, which is characterized in that described
History Order Transaction Information includes item description document；
This method it is described the History Order Transaction Information is filtered according to the filtering rule after, further includes:
Filtered item description document is segmented, to obtain participle phrase；
Word lists are established according to participle phrase, the number of each participle phrase is set.
5. the order price prediction technique according to claim 4 based on Chinese natural language processing, which is characterized in that described
Convolutional neural networks model includes embeding layer, three-layer coil lamination, global pool layer and full articulamentum；
The embeding layer converts expression study tensor, the expression for item description document using the number of the participle phrase
Study tensor is denoted as N*K dimension；
The three-layer coil lamination use S M1*K convolution kernel, respectively 0 dimension, 1 dimension ..., M1-1 tie up on to the M1*K dimension table
Dendrography practises tensor and does convolution, obtains the tensor of S N*1 dimension；The N*1 is tieed up using the convolution kernel that M2*1 and M3*1 is tieed up respectively again
Tensor carry out convolution；
The global pool layer carries out global maximum pond to the tensor of the three-layer coil lamination final output, obtain text to
Amount；
The text vector is connected by the full articulamentum with preset two layers full Connection Neural Network, the multiple valences of final output
Lattice.
6. the order price prediction technique according to claim 5 based on Chinese natural language processing, which is characterized in that described
The convolutional neural networks model is carried out according to corresponding legitimate reading in the prediction result and History Order Transaction Information
Reverse train is specifically included with obtaining best convolutional neural networks model:
Cross entropy is calculated according to legitimate reading corresponding in the prediction result and History Order Transaction Information, is defined as loss letter
Number；
Reverse train is carried out using Adam and SGD optimization method, until the loss function is minimum, to obtain best convolutional Neural
Network model.
7. the order price prediction technique according to claim 1 based on Chinese natural language processing, which is characterized in that
This method it is described to obtain best convolutional neural networks model after, further includes:
Delta Time is set；
When Delta Time reaches, the History Order Transaction Information in the Delta Time is obtained, returns and executes the pretreatment step
Suddenly, the best convolutional neural networks model is optimized.
8. a kind of terminal, which is characterized in that the processor, defeated including processor, input equipment, output equipment and memory
Enter equipment, output equipment and memory to be connected with each other, wherein the memory is for storing computer program, the computer
Program includes program instruction, and the processor is configured for calling described program instruction, is executed such as any one of claim 1-7
The method.
9. a kind of computer readable storage medium, which is characterized in that the computer storage medium is stored with computer program, institute
Stating computer program includes program instruction, and described program instruction executes the processor as right is wanted
Seek the described in any item methods of 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811644617.1A CN109508461A (en) | 2018-12-29 | 2018-12-29 | Order price prediction technique, terminal and medium based on Chinese natural language processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811644617.1A CN109508461A (en) | 2018-12-29 | 2018-12-29 | Order price prediction technique, terminal and medium based on Chinese natural language processing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109508461A true CN109508461A (en) | 2019-03-22 |
Family
ID=65755902
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811644617.1A Pending CN109508461A (en) | 2018-12-29 | 2018-12-29 | Order price prediction technique, terminal and medium based on Chinese natural language processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109508461A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110162558A (en) * | 2019-04-01 | 2019-08-23 | 阿里巴巴集团控股有限公司 | Structural data processing method and processing device |
| CN110188926A (en) * | 2019-05-10 | 2019-08-30 | 重庆天蓬网络有限公司 | A kind of order information forecasting system and method |
| CN110187945A (en) * | 2019-04-26 | 2019-08-30 | 平安科技（深圳）有限公司 | Indicate information generating method, device, terminal and storage medium |
| CN110838022A (en) * | 2019-10-15 | 2020-02-25 | 北京三快在线科技有限公司 | Advertisement price adjusting method and device, electronic equipment and computer readable medium |
| CN111007918A (en) * | 2019-12-11 | 2020-04-14 | 国网辽宁省电力有限公司经济技术研究院 | Electric power engineering cost prediction model based on artificial neural network |
| CN111833137A (en) * | 2020-01-14 | 2020-10-27 | 北京嘀嘀无限科技发展有限公司 | Order processing method, order processing device, computer equipment and storage medium |
| CN112884128A (en) * | 2021-03-04 | 2021-06-01 | 深圳前海黑顿科技有限公司 | Semi-supervised lightweight expert system based on scene |
| CN113743440A (en) * | 2021-01-13 | 2021-12-03 | 北京沃东天骏信息技术有限公司 | Information processing method and device and storage medium |
| CN114626302A (en) * | 2022-03-18 | 2022-06-14 | 杭州勒贝格智能系统有限责任公司 | Channel prediction method, system, terminal and medium based on Koopman operator and machine learning |
| CN115136154B (en) * | 2020-02-26 | 2023-09-15 | 凯首公司 | Systems and methods for optimizing wearable item selection in an electronic garment ordering platform |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102982229A (en) * | 2012-09-06 | 2013-03-20 | 淮阴工学院 | Multi-assortment commodity price expectation data pre-processing method based on neural networks |
| CN104599002A (en) * | 2015-02-13 | 2015-05-06 | 北京嘀嘀无限科技发展有限公司 | Order value predicting method and equipment |
| CN107203828A (en) * | 2017-06-22 | 2017-09-26 | 中北大学 | A kind of reinforcing bar price expectation method, system and platform |
| CN107301246A (en) * | 2017-07-14 | 2017-10-27 | 河北工业大学 | Chinese Text Categorization based on ultra-deep convolutional neural networks structural model |
| CN107369043A (en) * | 2017-07-19 | 2017-11-21 | 河海大学常州校区 | A kind of used car price evaluation optimized algorithm based on BP neural network |
| CN107943860A (en) * | 2017-11-08 | 2018-04-20 | 北京奇艺世纪科技有限公司 | The recognition methods and device that the training method of model, text are intended to |
| CN108154194A (en) * | 2018-01-18 | 2018-06-12 | 北京工业大学 | A kind of method with the convolutional network extraction high dimensional feature based on tensor |
| CN108509427A (en) * | 2018-04-24 | 2018-09-07 | 北京慧闻科技发展有限公司 | The data processing method of text data and application |
| WO2018191301A1 (en) * | 2017-04-12 | 2018-10-18 | Yodlee, Inc. | Neural networks for information extraction from transaction data |
| CN108694476A (en) * | 2018-06-29 | 2018-10-23 | 山东财经大学 | A kind of convolutional neural networks Stock Price Fluctuation prediction technique of combination financial and economic news |
| CN108717439A (en) * | 2018-05-16 | 2018-10-30 | 哈尔滨理工大学 | A kind of Chinese Text Categorization merged based on attention mechanism and characteristic strengthening |
| CN109034915A (en) * | 2018-09-03 | 2018-12-18 | 王昆 | A kind of artificial intelligent type e-commerce system using digital asset or integral as the media of exchange |
-
2018
- 2018-12-29 CN CN201811644617.1A patent/CN109508461A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102982229A (en) * | 2012-09-06 | 2013-03-20 | 淮阴工学院 | Multi-assortment commodity price expectation data pre-processing method based on neural networks |
| CN104599002A (en) * | 2015-02-13 | 2015-05-06 | 北京嘀嘀无限科技发展有限公司 | Order value predicting method and equipment |
| WO2018191301A1 (en) * | 2017-04-12 | 2018-10-18 | Yodlee, Inc. | Neural networks for information extraction from transaction data |
| CN107203828A (en) * | 2017-06-22 | 2017-09-26 | 中北大学 | A kind of reinforcing bar price expectation method, system and platform |
| CN107301246A (en) * | 2017-07-14 | 2017-10-27 | 河北工业大学 | Chinese Text Categorization based on ultra-deep convolutional neural networks structural model |
| CN107369043A (en) * | 2017-07-19 | 2017-11-21 | 河海大学常州校区 | A kind of used car price evaluation optimized algorithm based on BP neural network |
| CN107943860A (en) * | 2017-11-08 | 2018-04-20 | 北京奇艺世纪科技有限公司 | The recognition methods and device that the training method of model, text are intended to |
| CN108154194A (en) * | 2018-01-18 | 2018-06-12 | 北京工业大学 | A kind of method with the convolutional network extraction high dimensional feature based on tensor |
| CN108509427A (en) * | 2018-04-24 | 2018-09-07 | 北京慧闻科技发展有限公司 | The data processing method of text data and application |
| CN108717439A (en) * | 2018-05-16 | 2018-10-30 | 哈尔滨理工大学 | A kind of Chinese Text Categorization merged based on attention mechanism and characteristic strengthening |
| CN108694476A (en) * | 2018-06-29 | 2018-10-23 | 山东财经大学 | A kind of convolutional neural networks Stock Price Fluctuation prediction technique of combination financial and economic news |
| CN109034915A (en) * | 2018-09-03 | 2018-12-18 | 王昆 | A kind of artificial intelligent type e-commerce system using digital asset or integral as the media of exchange |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110162558A (en) * | 2019-04-01 | 2019-08-23 | 阿里巴巴集团控股有限公司 | Structural data processing method and processing device |
| CN110162558B (en) * | 2019-04-01 | 2023-06-23 | 创新先进技术有限公司 | Structured data processing method and device |
| CN110187945A (en) * | 2019-04-26 | 2019-08-30 | 平安科技（深圳）有限公司 | Indicate information generating method, device, terminal and storage medium |
| CN110188926A (en) * | 2019-05-10 | 2019-08-30 | 重庆天蓬网络有限公司 | A kind of order information forecasting system and method |
| CN110188926B (en) * | 2019-05-10 | 2020-11-13 | 重庆天蓬网络有限公司 | Order information prediction system and method |
| CN110838022A (en) * | 2019-10-15 | 2020-02-25 | 北京三快在线科技有限公司 | Advertisement price adjusting method and device, electronic equipment and computer readable medium |
| CN111007918A (en) * | 2019-12-11 | 2020-04-14 | 国网辽宁省电力有限公司经济技术研究院 | Electric power engineering cost prediction model based on artificial neural network |
| CN111833137A (en) * | 2020-01-14 | 2020-10-27 | 北京嘀嘀无限科技发展有限公司 | Order processing method, order processing device, computer equipment and storage medium |
| CN115136154B (en) * | 2020-02-26 | 2023-09-15 | 凯首公司 | Systems and methods for optimizing wearable item selection in an electronic garment ordering platform |
| CN113743440A (en) * | 2021-01-13 | 2021-12-03 | 北京沃东天骏信息技术有限公司 | Information processing method and device and storage medium |
| CN112884128A (en) * | 2021-03-04 | 2021-06-01 | 深圳前海黑顿科技有限公司 | Semi-supervised lightweight expert system based on scene |
| CN114626302A (en) * | 2022-03-18 | 2022-06-14 | 杭州勒贝格智能系统有限责任公司 | Channel prediction method, system, terminal and medium based on Koopman operator and machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109508461A (en) | Order price prediction technique, terminal and medium based on Chinese natural language processing | |
| Ma et al. | Autonomous deep learning: A genetic DCNN designer for image classification | |
| CN108629687B (en) | Anti-money laundering method, device and equipment | |
| US10839790B2 (en) | Sequence-to-sequence convolutional architecture | |
| US9495345B2 (en) | Methods and systems for modeling complex taxonomies with natural language understanding | |
| CN108959246A (en) | Answer selection method, device and electronic equipment based on improved attention mechanism | |
| CN108920654A (en) | A kind of matched method and apparatus of question and answer text semantic | |
| WO2018145098A1 (en) | Systems and methods for automatic semantic token tagging | |
| CN107066537A (en) | Hot news generation method, equipment, electronic equipment | |
| CN108280112A (en) | Abstraction generating method, device and computer equipment | |
| CN109863488A (en) | The device/server of Neural Network Data input system is disposed | |
| CN107810496A (en) | User text analysis | |
| CN111950596A (en) | Training method for neural network and related equipment | |
| CN112136142A (en) | Computer system predictive machine learning model | |
| CN107273019A (en) | Input language based on collaboration posture | |
| CN107168954A (en) | Text key word generation method and device and electronic equipment and readable storage medium storing program for executing | |
| CN109313720A (en) | The strength neural network of external memory with sparse access | |
| CN109344234A (en) | Machine reads understanding method, device, computer equipment and storage medium | |
| CN108241741A (en) | A kind of file classification method, server and computer readable storage medium | |
| CN107807945A (en) | Device, method and the data handling system of data are provided | |
| CN113722438B (en) | Sentence vector generation method and device based on sentence vector model and computer equipment | |
| CN107145485A (en) | Method and apparatus for compressing topic model | |
| CN110309271A (en) | Intelligent knowledge study and question and answer technology | |
| CN112132238A (en) | Method, device, equipment and readable medium for identifying private data | |
| CN114175018A (en) | New word classification technique |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190322 |
|
| RJ01 | Rejection of invention patent application after publication |