CN107808220A - A kind of revenue prediction method and device - Google Patents
A kind of revenue prediction method and device Download PDFInfo
- Publication number
- CN107808220A CN107808220A CN201711032481.4A CN201711032481A CN107808220A CN 107808220 A CN107808220 A CN 107808220A CN 201711032481 A CN201711032481 A CN 201711032481A CN 107808220 A CN107808220 A CN 107808220A
- Authority
- CN
- China
- Prior art keywords
- data
- network structure
- neural network
- expenditure
- channels
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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"
-
- 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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a kind of revenue prediction method and device, wherein, this method includes：Receive current expenditure data of the enterprise to each channels；By current expenditure data input into the neural network structure pre-set, output corresponds to the prediction total income data of current expenditure data；Wherein, neural network structure is obtained as follows：Obtain the historical data set of enterprise；Wherein, historical data set includes the expenditure data of each channels and the total income data corresponding to all channels；The neural network structure of structure is trained based on expenditure data and total income data using back-propagation method, until when the output of neural network structure reaches convergence, deconditioning.For the present invention by enterprise in current expenditure data input to the neural network structure for advancing with back-propagation method training of each channels, obtaining prediction total income data, the efficiency and flexibility ratio of prediction are higher.
Description
Technical field
The present invention relates to technical field of data processing, in particular to a kind of revenue prediction method and device.
Background technology
The network promotion, due to its relatively conventional way of promotion, has without region, passed as a kind of emerging way of promotion
The features such as continuous, precision is high, interactive strong, audient's level is high is broadcast, and can easily carry out feedback of the information, effect
Tracking, consumer's analysis, behavior investigation, even consumption etc. behavior and more and more paid attention to.
For enterprise product (such as website, APP), it can carry out product a surname by various network promotion channels
Pass.Wherein, the channels of website can be each search engine including 360, Baidu, search dog etc., APP popularization canal
Road can be then including App Store, using each application platform including treasured, Huawei's application market etc..Enterprise can basis
Itself need to choose one or more progress product promotions in above-mentioned channels, however, either applying any popularization
Channel, enterprise are required for putting into corresponding fund, and final promotion effect is then judged with the income of enterprise.
For to a certain degree, enterprise using each channels promoted after total income by with channels
The increase of total expenditure and increase, however, above-mentioned total income is not necessarily directly proportional to total expenditure, therefore, how to be pushed away using limited
The revenus maximization that wide fund obtains enterprise turns into a urgent problem to be solved.
The content of the invention
In view of this, it is an object of the invention to provide a kind of revenue prediction method and device, based on back-propagation method
The income of enterprise is predicted, forecasting efficiency and flexibility ratio are higher.
In a first aspect, the invention provides a kind of revenue prediction method, methods described includes：
Receive current expenditure data of the enterprise to each channels；
By the current expenditure data input into the neural network structure pre-set, output corresponds to the current branch
Go out the prediction total income data of data；
Wherein, the neural network structure is obtained as follows：
Obtain the historical data set of the enterprise；Wherein, the historical data set includes each channels
Expenditure data and the total income data corresponding to all channels；
Using back-propagation method the neural network structure of data and the total income data to structure is paid based on described
It is trained, until when the output of the neural network structure reaches convergence, deconditioning.
With reference in a first aspect, the invention provides the possible embodiment of the first of first aspect, wherein, the acquisition
After the historical data set of the enterprise expenditure data and the total income are based on described using back-propagation method
Before data are trained to the neural network structure of structure, in addition to：
The expenditure data in the historical data set and the total income data are normalized, obtain normalizing
Expenditure data and total income data after change.
With reference in a first aspect, the invention provides the possible embodiment of second of first aspect, wherein, by as follows
Step judges whether the output of the neural network structure reaches convergence：
Determine whether the frequency of training of the neural network structure reaches predetermined threshold value, obtain the first matching result；
Judge whether to reach convergence according to first matching result.
With reference in a first aspect, the invention provides the possible embodiment of the third of first aspect, wherein, by as follows
Step judges whether the output of the neural network structure reaches convergence：
Determine whether the output error between the output result of the neural network structure and the total income data is less than
Default error, obtains the second matching result；
Judge whether to reach convergence according to second matching result.
With reference to the third possible embodiment of first aspect, the invention provides the 4th of first aspect kind is possible
Embodiment, wherein, it is described to pay the god of data and the total income data to structure based on described using back-propagation method
It is trained through network structure, until when the output of the neural network structure reaches convergence, deconditioning, including：
Build neural network structure；
The neural network structure that the expenditure data input of initial training parameter and each channels is extremely built；
Judge whether the output error is less than the default error, if it is not, based on the output error to described initial
Training parameter is adjusted；
Circulate the expenditure data input of the training parameter after adjustment and each channels to the neutral net
Structure, judges whether the output error is less than the default error, until judging that it is described pre- that the output error is less than
If during error, stop circulation.
Second aspect, present invention also offers a kind of income forecast device, described device includes：
Receiving module, for receiving current expenditure data of the enterprise to each channels；
Prediction module, for by the current expenditure data input into the neural network structure pre-set, output pair
The prediction total income data of data are currently paid described in Ying Yu；
Training module, for obtaining the historical data set of the enterprise；Wherein, the historical data set includes each
The expenditure data of the channels and the total income data corresponding to all channels；Institute is based on using back-propagation method
State expenditure data and the total income data are trained to the neural network structure of structure, until in the neural network structure
Output when reaching convergence, deconditioning.
With reference to second aspect, the invention provides the possible embodiment of the first of second aspect, wherein, in addition to：
Module is normalized, for the expenditure data in the historical data set and total income data progress
Normalization, expenditure data and total income data after being normalized.
With reference to second aspect, the invention provides the possible embodiment of second of second aspect, wherein, in addition to：
First judge module, for determining whether the frequency of training of the neural network structure reaches predetermined threshold value, obtain
First matching result；Judge whether to reach convergence according to first matching result.
With reference to second aspect, the invention provides the possible embodiment of the third of second aspect, wherein, in addition to：
Second judge module, for determining between the output result of the neural network structure and the total income data
Whether output error is less than default error, obtains the second matching result；Judge whether to reach receipts according to second matching result
Hold back.
With reference to the third possible embodiment of second aspect, the invention provides the 4th of second aspect kind is possible
Embodiment, wherein, the training module includes：
Construction unit, for building neural network structure；
Input block, for the god for extremely building the expenditure data input of initial training parameter and each channels
Through network structure；
Judging unit, for judging whether the output error is less than the default error, if it is not, being missed based on the output
Difference is adjusted to the initial training parameter；
Cycling element, for circulating the expenditure data input of the training parameter after adjustment and each channels extremely
The neural network structure, judges whether the output error is less than the default error, until judging the output mistake
When difference is less than the default error, stop circulation.
Revenue prediction method and device provided by the invention, it receives current expenditure of the enterprise to each channels first
Data；Then by current expenditure data input into the neural network structure pre-set, output corresponds to current expenditure data
Prediction total income data；Wherein, neural network structure is obtained as follows：Obtain the historical data set of enterprise；Its
In, historical data set includes the expenditure data of each channels and the total income data corresponding to all channels；Profit
The neural network structure of structure is trained based on expenditure data and total income data with back-propagation method, until in nerve
When the output of network structure reaches convergence, deconditioning, its by enterprise to the current expenditure data inputs of each channels extremely
Advance with back-propagation method training neural network structure in, obtain predict total income data, the efficiency of prediction and flexibly
Degree is higher.
To enable the above objects, features and advantages of the present invention to become apparent, preferred embodiment cited below particularly, and coordinate
Appended accompanying drawing, is described in detail below.
Brief description of the drawings
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below by embodiment it is required use it is attached
Figure is briefly described, it will be appreciated that the following drawings illustrate only certain embodiments of the present invention, therefore be not construed as pair
The restriction of scope, for those of ordinary skill in the art, on the premise of not paying creative work, can also be according to this
A little accompanying drawings obtain other related accompanying drawings.
Fig. 1 shows a kind of flow chart for revenue prediction method that the embodiment of the present invention is provided；
Fig. 2 shows the flow chart for another revenue prediction method that the embodiment of the present invention is provided；
Fig. 3 shows the flow chart for another revenue prediction method that the embodiment of the present invention is provided；
Fig. 4 shows a kind of module composition schematic diagram for income forecast device that the embodiment of the present invention is provided.
Main element symbol description：
11st, receiving module；22nd, training module；33rd, prediction module.
Embodiment
To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, below in conjunction with the embodiment of the present invention
Middle accompanying drawing, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described embodiment is only
It is part of the embodiment of the present invention, rather than whole embodiments.The present invention being generally described and illustrated herein in the accompanying drawings is real
Applying the component of example can be configured to arrange and design with a variety of.Therefore, it is of the invention to what is provided in the accompanying drawings below
The detailed description of embodiment is not intended to limit the scope of claimed invention, but is merely representative of the selected reality of the present invention
Apply example.Based on embodiments of the invention, institute that those skilled in the art are obtained on the premise of creative work is not made
There is other embodiment, belong to the scope of protection of the invention.
In view of correlation technique urgently demand it is a kind of how using limited popularizations fund acquisition enterprise revenus maximization
Solution, based on this, the embodiments of the invention provide a kind of revenue prediction method and device, based on back-propagation method pair
The income of enterprise is predicted, and forecasting efficiency and flexibility ratio are higher.
The flow chart of revenue prediction method provided in an embodiment of the present invention shown in Figure 1, the above method specifically include
Following steps：
S101, receive current expenditure data of the enterprise to each channels；
S102, data input will be currently paid into the neural network structure pre-set, output corresponds to current pay
The prediction total income data of data.
Specifically, in view of the concrete application scene of revenue prediction method provided in an embodiment of the present invention, the present invention is implemented
The revenue prediction method that example provides needs to receive current expenditure data of the enterprise to each channels.Data will currently paid
Before input to neural network structure, above-mentioned current expenditure data are normalized first, it is defeated then to carry out data again
Enter, the output valve by above-mentioned neural network structure computing is to correspond to the prediction total income data of current expenditure data, enterprise
Industry operation department can optimize channel promotion input according to this data.
Specifically, with the arrival in big data epoch, only more complicated model, articulate mould in other words
Type, the abundant information contained in mass data could be fully excavated, so, the neural network structure in the embodiment of the present invention uses
Be more powerful model, to allow us to excavate more valuable information and knowledge from historical data.Wherein, exist
When carrying out total income prediction to current expenditure data, it is necessary first to it is above-mentioned to realize to obtain the good neural network structure of training in advance
Prediction, referring to Fig. 2, the training process of above-mentioned neural network structure specifically includes：
S201, the historical data set for obtaining enterprise；Wherein, historical data set includes the expenditure number of each channels
According to the total income data corresponding to all channels；
S202, using back-propagation method the neural network structure of structure is carried out based on expenditure data and total income data
Training, until when the output of neural network structure reaches convergence, deconditioning.
Specifically, revenue prediction method provided in an embodiment of the present invention before neural network structure training is carried out, it is necessary to
Historical data is obtained as sample data.The sample data includes the expenditure data of each channels and corresponding to all popularization
The total income data of channel.Wherein, above-mentioned sample data can be carried out by data-interface, web crawlers even manual type
Obtain, the embodiment of the present invention does not do specific limitation to this.
In addition, revenue prediction method provided in an embodiment of the present invention can arrange nearly 2 years of enterprise daily each channel input and
The historical datas such as same day total income, the data for randomly selecting wherein 80% are used as test data as training data, 20%, wherein
Training set is used for training pattern, determines model parameter, and test set is used for the accuracy rate of testing model, and this data is only examined in model
Used when testing.
In view of a user from knowing some enterprise product (such as website or APP) to final purchase networking products/meeting
Member, generally comprises following steps：1. certain website or APP are known by some channel；2. access website or obtained using APP
Required function；3. just purchase networking products or purchase member complete consumption if the need arises.Based on above-mentioned user behavior,
The embodiment of the present invention builds the neural network structure for including input layer, hidden layer and output layer first, wherein, above-mentioned input layer is
The each channel promoted, it is inputted as the promotion expense of corresponding channel；Hidden layer may include multilayer, it is impossible in training sample
It is directly observed their value；Output layer buys consumer behavior, the output layer such as product or purchase member and only includes a node,
Final output is the income on the same day；Then the sample data again based on above-mentioned acquisition is trained to the neural network structure,
To obtain the neural network structure for meeting user behavior, practicality is preferable.
Revenue prediction method provided by the invention, it receives current expenditure data of the enterprise to each channels first；
Then by current expenditure data input into the neural network structure pre-set, output corresponds to the prediction of current expenditure data
Total income data；Wherein, neural network structure is obtained as follows：Obtain the historical data set of enterprise；Wherein, history
Data acquisition system includes the expenditure data of each channels and the total income data corresponding to all channels；Passed using reverse
Broadcasting method is trained based on expenditure data and total income data to the neural network structure of structure, until in neural network structure
Output when reaching convergence, deconditioning, it is by enterprise to the current expenditure data inputs of each channels to advancing with
In the neural network structure of back-propagation method training, prediction total income data are obtained, the efficiency and flexibility ratio of prediction are higher.
So that the data of separate sources are unified under a referential, it is significant so to compare up.It is of the invention real
The revenue prediction method of example offer is applied after the historical data set of enterprise is obtained, also to the expenditure number in the historical data set
It is normalized according to total income data, with the expenditure data and total income data after being normalized.Wherein, it is of the invention
Preferably above-mentioned sample data is entered (such as expenditure data and total income data) using mean normalization processing method in embodiment
Row normalized, (- 1,1) section is transformed the data into, specifically using formulaWherein, μ is all samples
The average of data, σ are that the standard deviation of all sample datas is realized.
Revenue prediction method provided in an embodiment of the present invention can judge above-mentioned neural network structure in the following way
Whether output reaches convergence.The first situation, the embodiment of the present invention can be using the frequency of training of determination neural network structure
The no judgment mode for reaching predetermined threshold value (such as 5000 times), if frequency of training reaches predetermined threshold value, it is determined that above-mentioned nerve net
The output of network structure has reached convergence, if frequency of training is not up to predetermined threshold value, it is determined that not up to restrain.Second of situation,
The embodiment of the present invention can also be using the output error between the output result and total income data that determine neural network structure (such as
Error 0.0001) whether it is less than the judgment mode of default error, if output error is less than default error, it is determined that above-mentioned nerve
The output of network structure has reached convergence, if output error is less than or equal to default error, it is determined that not up to restrain.The third
Situation, the embodiment of the present invention can also use the combination of the first above-mentioned judgment mode and second of judgment mode, if
Frequency of training and when output error meets the requirements, just determines to have restrained.It can be seen that income forecast side provided in an embodiment of the present invention
Method can meet the different demands of different user, and applicability is stronger.
The training process of the neural network structure in the case of second is further illustrated below, referring to Fig. 3, this hair
The revenue prediction method that bright embodiment provides also comprises the following steps：
S301, structure neural network structure；
S302, the neural network structure for extremely building initial training parameter and the expenditure data input of each channels；
S303, judge whether output error is less than default error, if it is not, being carried out based on output error to initial training parameter
Adjustment；
S304, circulate the expenditure data input of the training parameter after adjustment and each channels to neutral net knot
Structure, judges whether output error is less than default error, until when judging that output error is less than default error, stops circulation.
The training process of above-mentioned neural network structure for convenience of description, next said with a specific example
It is bright.
By taking APP extension process as an example, the input layer of the neural network structure of structure is all channels, and quantity is general
In n=50 or so, as input layer number；Output layer is the total income on the same day, only needs a node；Hidden layer start node
Number is log2 n, wherein, n is the nodes of input layer；Therefore hidden layer can be designed as 2 layers, every layer includes 6 nodes.
Wherein, the embodiment of the present invention is entered data into before neural network structure, and the sample data of input can be entered
Row normalized, for APP, its channels includes App Store, promoted using treasured, Huawei's application market, 360
Deng, then it is corresponding expenditure data be input sample data；Total income data are to export sample data, above-mentioned input sample data
It is the data after normalized with output sample data.In addition, it is corresponding to the input layer number of neural network structure
For channels number, corresponding A pp Store is distinguished, using treasured, Huawei's application market, 360 popularizations etc., hidden layer design includes
For 2 layers, every layer includes 6 nodes, one node of output layer.
Revenue prediction method provided in an embodiment of the present invention is also by initial training parameter and the expenditure number of each channels
According to inputting into the neural network structure of above-mentioned structure to be trained to neural network structure, corresponding training parameter is obtained.
That is, the embodiment of the present invention, which inputs above-mentioned sample data (the expenditure data of i.e. each channel), arrives input layer, backpropagation is used
Method carries out adjusting training repeatedly to the weights and deviation of network structure, and detailed process is：For the input data given, from
Input layer passes to hidden layer, and the result after processing is passed to output layer by hidden layer by connection weight and excitation function, will
The output result of output layer is compared with total income data, obtains output error, then backstepping to the connection weight in neutral net
Feedback modifiers are carried out again, neutral net are carried out using training data repeatedly to train with the accuracy of sophisticated model, when network is defeated
Training is completed when the error gone out is less than the default error specified.Wherein, the activation primitive selection of above-mentioned hidden layer node is
Sigmoid functions；Activation primitive expression formula：In, the x in above-mentioned expression formula is by input data and connection weight
It is determined that.
Wherein, above-mentioned initial training parameter includes initial connection weight and initial deviation, receipts provided in an embodiment of the present invention
Enter Forecasting Methodology and above-mentioned initial value is preferably randomly generated using random function, to avoid all parameters from using identical value,
The node output function of all hidden layers is all identical, and the possibility for causing study to fail.
Revenue prediction method provided by the invention, it receives current expenditure data of the enterprise to each channels first；
Then by current expenditure data input into the neural network structure pre-set, output corresponds to the prediction of current expenditure data
Total income data；Wherein, neural network structure is obtained as follows：Obtain the historical data set of enterprise；Wherein, history
Data acquisition system includes the expenditure data of each channels and the total income data corresponding to all channels；Passed using reverse
Broadcasting method is trained based on expenditure data and total income data to the neural network structure of structure, until in neural network structure
Output when reaching convergence, deconditioning, it is by enterprise to the current expenditure data inputs of each channels to advancing with
In the neural network structure of back-propagation method training, by the income feelings on obtained prediction total income data prediction enterprise's same day
Condition, and then instruct enterprise operation department to carry out channel optimization of investment work, obtain and maximum receipts are obtained with the channel input of minimum
The effect entered, reaches profit maximization.
The embodiment of the present invention additionally provides a kind of income forecast device, and the device is used to perform above-mentioned revenue prediction method,
Referring to Fig. 4, said apparatus includes：
Receiving module 11, for receiving current expenditure data of the enterprise to each channels；
Prediction module 33, for, into the neural network structure pre-set, output to be corresponding by current expenditure data input
In the prediction total income data of currently expenditure data；
Training module 22, for obtaining the historical data set of enterprise；Wherein, historical data set includes each popularization canal
The expenditure data in road and the total income data corresponding to all channels；Using back-propagation method based on expenditure data and always
Income data is trained to the neural network structure of structure, until when the output of neural network structure reaches convergence, is stopped
Training.
So that the data of separate sources are unified under a referential, it is significant so to compare up.It is of the invention real
The income forecast device of example offer is applied after the historical data set of enterprise is obtained, also by normalizing module to the historical data
Expenditure data and total income data in set are normalized, with the expenditure data and total income number after being normalized
According to.Wherein, in the embodiment of the present invention preferably using mean normalization processing method to above-mentioned sample data (as expenditure data and
Total income data) it is normalized, (- 1,1) section is transformed the data into, specifically using formulaIts
In, μ is the average of all sample datas, and σ is that the standard deviation of all sample datas is realized.
Income forecast device provided in an embodiment of the present invention can judge above-mentioned neural network structure in the following way
Whether output reaches convergence.The first situation, the embodiment of the present invention can determine neural network structure using the first judge module
Frequency of training whether reach the judgment modes of predetermined threshold value (such as 5000 times), if frequency of training reaches predetermined threshold value, really
The output of fixed above-mentioned neural network structure has reached convergence, if frequency of training is not up to predetermined threshold value, it is determined that not up to restrain.
Second of situation, the embodiment of the present invention can also use the second judge module to determine the output result of neural network structure and total receipts
Enter the judgment mode whether output error (such as error 0.0001) between data is less than default error, if output error is less than
Default error, it is determined that the output of above-mentioned neural network structure has reached convergence, if output error is less than or equal to default error,
Then determine not up to restrain.The third situation, the embodiment of the present invention can also use the first above-mentioned judgment mode and second
The combination of judgment mode, if frequency of training and when output error meets the requirements, just determine to have restrained.It can be seen that this hair
The income forecast device that bright embodiment provides can meet the different demands of different user, and applicability is stronger.
Training module 22 in income forecast device provided in an embodiment of the present invention, is specifically included：
Construction unit, for building neural network structure；
Input block, for the nerve net by the expenditure data input of initial training parameter and each channels to structure
Network structure；
Judging unit, for judging whether output error is less than default error, if it is not, based on output error to initial training
Parameter is adjusted；
Cycling element, for circulating the expenditure data input of the training parameter after adjustment and each channels to nerve
Network structure, judges whether output error is less than default error, until when judging that output error is less than default error, stops
Circulation.
Income forecast device provided by the invention, its receiving module 11 receive current expenditure of the enterprise to each channels
Data；For prediction module 33 by current expenditure data input into the neural network structure pre-set, output corresponds to current branch
Go out the prediction total income data of data；Training module 22, then obtain the historical data set of enterprise；Wherein, historical data set
Expenditure data including each channels and the total income data corresponding to all channels；Utilize back-propagation method base
The neural network structure of structure is trained in expenditure data and total income data, until the output in neural network structure reaches
To during convergence, deconditioning, it is by enterprise to the current expenditure data inputs of each channels to advancing with backpropagation
In the neural network structure of method training, by the income situation on obtained prediction total income data prediction enterprise's same day, and then
Instruct enterprise operation department to carry out channel optimization of investment work, obtain the effect that maximum income is obtained with the channel input of minimum
Fruit, reach profit maximization.
The computer program product of the method for the progress income forecast that the embodiment of the present invention is provided, including store program
The computer-readable recording medium of code, the instruction that described program code includes can be used for performing described in previous methods embodiment
Method, specific implementation can be found in embodiment of the method, will not be repeated here.
The device for the income forecast that the embodiment of the present invention is provided can be that the specific hardware or be installed in equipment is set
Standby upper software or firmware etc..The device that the embodiment of the present invention is provided, its realization principle and caused technique effect and foregoing
Embodiment of the method is identical, and to briefly describe, device embodiment part does not refer to part, refers to corresponding in preceding method embodiment
Content.It is apparent to those skilled in the art that for convenience and simplicity of description, system described above, dress
The specific work process with unit is put, the corresponding process in above method embodiment is may be referred to, will not be repeated here.
In embodiment provided by the present invention, it should be understood that disclosed apparatus and method, can be by others side
Formula is realized.Device embodiment described above is only schematical, for example, the division of the unit, only one kind are patrolled
Function division is collected, there can be other dividing mode when actually realizing, in another example, multiple units or component can combine or can
To be integrated into another system, or some features can be ignored, or not perform.Another, shown or discussed is mutual
Coupling or direct-coupling or communication connection can be INDIRECT COUPLING or communication link by some communication interfaces, device or unit
Connect, can be electrical, mechanical or other forms.
The unit illustrated as separating component can be or may not be physically separate, show as unit
The part shown can be or may not be physical location, you can with positioned at a place, or can also be distributed to multiple
On NE.Some or all of unit therein can be selected to realize the mesh of this embodiment scheme according to the actual needs
's.
In addition, each functional unit in embodiment provided by the invention can be integrated in a processing unit, also may be used
To be that unit is individually physically present, can also two or more units it is integrated in a unit.
If the function is realized in the form of SFU software functional unit and is used as independent production marketing or in use, can be with
It is stored in a computer read/write memory medium.Based on such understanding, technical scheme is substantially in other words
The part to be contributed to prior art or the part of the technical scheme can be embodied in the form of software product, the meter
Calculation machine software product is stored in a storage medium, including some instructions are causing a computer equipment (can be
People's computer, server, or network equipment etc.) perform all or part of step of each embodiment methods described of the present invention.
And foregoing storage medium includes：USB flash disk, mobile hard disk, read-only storage (ROM, Read-Only Memory), arbitrary access are deposited
Reservoir (RAM, Random Access Memory), magnetic disc or CD etc. are various can be with the medium of store program codes.
It should be noted that：Similar label and letter represents similar terms in following accompanying drawing, therefore, once a certain Xiang Yi
It is defined, then it further need not be defined and explained in subsequent accompanying drawing in individual accompanying drawing, in addition, term " the
One ", " second ", " the 3rd " etc. are only used for distinguishing description, and it is not intended that instruction or hint relative importance.
Finally it should be noted that：Embodiment described above, it is only the embodiment of the present invention, to illustrate the present invention
Technical scheme, rather than its limitations, protection scope of the present invention is not limited thereto, although with reference to the foregoing embodiments to this hair
It is bright to be described in detail, it will be understood by those within the art that：Any one skilled in the art
The invention discloses technical scope in, it can still modify to the technical scheme described in previous embodiment or can be light
Change is readily conceivable that, or equivalent substitution is carried out to which part technical characteristic；And these modifications, change or replacement, do not make
The essence of appropriate technical solution departs from the spirit and scope of technical scheme of the embodiment of the present invention.The protection in the present invention should all be covered
Within the scope of.Therefore, protection scope of the present invention described should be defined by scope of the claims.
Claims (10)
- A kind of 1. revenue prediction method, it is characterised in that including：Receive current expenditure data of the enterprise to each channels；By the current expenditure data input into the neural network structure pre-set, output corresponds to the current expenditure number According to prediction total income data；Wherein, the neural network structure is obtained as follows：Obtain the historical data set of the enterprise；Wherein, the historical data set includes the branch of each channels Go out data and the total income data corresponding to all channels；The neural network structure of structure is carried out based on the expenditure data and the total income data using back-propagation method Training, until when the output of the neural network structure reaches convergence, deconditioning.
- 2. according to the method for claim 1, it is characterised in that after the historical data set for obtaining the enterprise and It is described that the neural network structure of structure is carried out based on the expenditure data and the total income data using back-propagation method Before training, in addition to：The expenditure data in the historical data set and the total income data are normalized, after obtaining normalization Expenditure data and total income data.
- 3. according to the method for claim 1, it is characterised in that judge the defeated of the neural network structure as follows Go out and whether reach convergence：Determine whether the frequency of training of the neural network structure reaches predetermined threshold value, obtain the first matching result；Judge whether to reach convergence according to first matching result.
- 4. according to the method for claim 1, it is characterised in that judge the defeated of the neural network structure as follows Go out and whether reach convergence：It is default to determine whether the output error between the output result of the neural network structure and the total income data is less than Error, obtain the second matching result；Judge whether to reach convergence according to second matching result.
- 5. according to the method for claim 4, it is characterised in that described to be based on the expenditure data using back-propagation method The neural network structure of structure is trained with the total income data, until the output in the neural network structure reaches During convergence, deconditioning, including：Build neural network structure；The neural network structure that the expenditure data input of initial training parameter and each channels is extremely built；Judge whether the output error is less than the default error, if it is not, based on the output error to the initial training Parameter is adjusted；Circulate the expenditure data input of the training parameter after adjustment and each channels to the neural network structure, Judge whether the output error is less than the default error, until judging that the output error is less than the default error When, stop circulation.
- A kind of 6. income forecast device, it is characterised in that including：Receiving module, for receiving current expenditure data of the enterprise to each channels；Prediction module, for, into the neural network structure pre-set, output to correspond to by the current expenditure data input The prediction total income data of the current expenditure data；Training module, for obtaining the historical data set of the enterprise；Wherein, the historical data set includes each described The expenditure data of channels and the total income data corresponding to all channels；The branch is based on using back-propagation method Go out data and the total income data are trained to the neural network structure of structure, until in the defeated of the neural network structure When going out to reach convergence, deconditioning.
- 7. device according to claim 6, it is characterised in that also include：Module is normalized, for carrying out normalizing to the expenditure data in the historical data set and the total income data Change, expenditure data and total income data after being normalized.
- 8. device according to claim 6, it is characterised in that also include：First judge module, for determining whether the frequency of training of the neural network structure reaches predetermined threshold value, obtain first Matching result；Judge whether to reach convergence according to first matching result.
- 9. device according to claim 6, it is characterised in that also include：Second judge module, for determining the output between the output result of the neural network structure and the total income data Whether error is less than default error, obtains the second matching result；Judge whether to reach convergence according to second matching result.
- 10. device according to claim 9, it is characterised in that the training module includes：Construction unit, for building neural network structure；Input block, for the nerve net for extremely building the expenditure data input of initial training parameter and each channels Network structure；Judging unit, for judging whether the output error is less than the default error, if it is not, being based on the output error pair The initial training parameter is adjusted；Cycling element, for circulating the expenditure data input of the training parameter after adjustment and each channels to described Neural network structure, judges whether the output error is less than the default error, until judging that the output error is small When the default error, stop circulation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711032481.4A CN107808220A (en) | 2017-10-30 | 2017-10-30 | A kind of revenue prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711032481.4A CN107808220A (en) | 2017-10-30 | 2017-10-30 | A kind of revenue prediction method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107808220A true CN107808220A (en) | 2018-03-16 |
Family
ID=61582527
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711032481.4A Pending CN107808220A (en) | 2017-10-30 | 2017-10-30 | A kind of revenue prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107808220A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108629632A (en) * | 2018-05-09 | 2018-10-09 | 北京京东金融科技控股有限公司 | Predict the method, apparatus and computer readable storage medium of user's income |
| CN108764863A (en) * | 2018-05-24 | 2018-11-06 | 腾讯科技（深圳）有限公司 | A kind of virtual resource transfer method, device, server and storage medium |
| CN111461872A (en) * | 2020-04-02 | 2020-07-28 | 杨九妹 | Block chain-based capital management method and system for big data enterprise |
| CN111861000A (en) * | 2020-07-21 | 2020-10-30 | 携程计算机技术（上海）有限公司 | Daily income prediction method, system, equipment and storage medium based on historical data |
| CN113313316A (en) * | 2021-06-11 | 2021-08-27 | 北京明略昭辉科技有限公司 | Method and device for outputting prediction data, storage medium and electronic equipment |
-
2017
- 2017-10-30 CN CN201711032481.4A patent/CN107808220A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108629632A (en) * | 2018-05-09 | 2018-10-09 | 北京京东金融科技控股有限公司 | Predict the method, apparatus and computer readable storage medium of user's income |
| CN108764863A (en) * | 2018-05-24 | 2018-11-06 | 腾讯科技（深圳）有限公司 | A kind of virtual resource transfer method, device, server and storage medium |
| CN108764863B (en) * | 2018-05-24 | 2021-10-29 | 腾讯科技（深圳）有限公司 | Data transfer method, device, server and storage medium |
| CN111461872A (en) * | 2020-04-02 | 2020-07-28 | 杨九妹 | Block chain-based capital management method and system for big data enterprise |
| CN111861000A (en) * | 2020-07-21 | 2020-10-30 | 携程计算机技术（上海）有限公司 | Daily income prediction method, system, equipment and storage medium based on historical data |
| CN113313316A (en) * | 2021-06-11 | 2021-08-27 | 北京明略昭辉科技有限公司 | Method and device for outputting prediction data, storage medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107808220A (en) | A kind of revenue prediction method and device | |
| KR100751965B1 (en) | method and system for predicting attrition customers | |
| CN111754336B (en) | Method and device for scheduling network points | |
| CN107220217A (en) | Characteristic coefficient training method and device that logic-based is returned | |
| CN107153874A (en) | Water quality prediction method and system | |
| CN108053314A (en) | A kind of Loan Demand Forecasting Methodology | |
| CN105897616B (en) | Resource allocation method and server | |
| CN108122122A (en) | Advertisement placement method and system | |
| CN108427708A (en) | Data processing method, device, storage medium and electronic device | |
| CN105550927A (en) | Bank credit system risk assessment method and device | |
| CN107633323A (en) | Core protects method, apparatus, computer equipment and the storage medium of difficulty prediction | |
| CN105354725A (en) | Prediction method and system of promotion effect of application | |
| CN106651097A (en) | Data collection method, data collection device and data collection server based on crowd sourcing | |
| CN108090788A (en) | Ad conversion rates predictor method based on temporal information integrated model | |
| CN107203912A (en) | Dynamic Pricing method and apparatus | |
| CN109741098A (en) | Broadband off-network prediction technique, equipment and storage medium | |
| CN106407477A (en) | Multidimensional interconnection recommendation method and system | |
| CN113674013B (en) | Advertisement bidding adjustment method and system based on merchant custom rules | |
| CN112669084B (en) | Policy determination method, device and computer readable storage medium | |
| CN110362728A (en) | Information-pushing method, device, equipment and storage medium based on big data analysis | |
| CN110288377A (en) | Optimization method and device, storage medium and electronic device are launched in advertisement | |
| CN108596765A (en) | A kind of Electronic Finance resource recommendation method and device | |
| CN112632179A (en) | Model construction method and device, storage medium and equipment | |
| CN114584406B (en) | Industrial big data privacy protection system and method for federated learning | |
| CN111062774B (en) | Activity delivery method and device, electronic equipment and computer readable 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180316 |
|
| RJ01 | Rejection of invention patent application after publication |