CN110363244A - A kind of method and apparatus of marketing data processing - Google Patents
A kind of method and apparatus of marketing data processing Download PDFInfo
- Publication number
- CN110363244A CN110363244A CN201910639120.9A CN201910639120A CN110363244A CN 110363244 A CN110363244 A CN 110363244A CN 201910639120 A CN201910639120 A CN 201910639120A CN 110363244 A CN110363244 A CN 110363244A
- Authority
- CN
- China
- Prior art keywords
- marketing
- data
- training
- positive
- negative sample
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of method and apparatus of marketing data processing, the method for marketing data processing includes: that the positive and negative sample set of marketing training model is established according to sales service basic data；Marketing training model is established according to positive and negative sample set and random forests algorithm；Initial marketing data is determined according to sales service data and marketing training model to be predicted；Marketing training model is updated using initial marketing data；Marketing data processing is carried out using updated marketing training model, generates marketing listings data.The method of marketing data processing provided by the invention, is become marking automatically from manual markings, is effectively reduced business personnel and is marked workload, and mark accuracy rate is promoted；And model modification strategy is updated by single cover type, becomes updating according to intelligent strategy, the difficulty that the business of reduction sets model parameter realizes intelligent updating, improves marketing success rate.
Description
Technical field
The present invention relates to artificial intelligence fields, more particularly to a kind of method and apparatus of marketing data processing.
Background technique
Traditional marketing mode is usually present certain difficulty when facing magnanimity client and numerous products.Finance simultaneously
Horizontal competition pressure is growing day by day, and business bank needs by new technical tool, method and platform, more towards corporate client
Precisely, more intelligently recommended products.
With the technologies such as big data, artificial intelligence, cloud computing continue to develop with it is mature, based on bank client representation data,
The continuous unified integration of transaction data, product data, can be realized the foundation of intelligent recommendation model, effectively improve marketing success rate,
Promotion business value.However, there are also universal problems: before 1. marketing, the modelling phase needs business to participate in label
For the positive negative sample of modeling, the mistake of label will have a direct impact on the accuracy rate of model training.2. in marketing, recommending client's number
It cannot such as timely update after model is online simultaneously according to continually changing, prediction result has time delay, influences marketing experience.
3. after marketing, result and the practical successful result that model is recommended can be variant, need to make intersection statistics, lack real-time system
Meter feedback cannot will recommend success or failure experience to share out.Therefore manual mark can be reduced, improve marketing by being first badly in need of one kind
The method of success rate.
Summary of the invention
In order to reduce manual mark, improve modeling efficiency, and the sample data set according to continuous renewal is updated intelligence
Marketing Model promotes business value, and the present invention provides a kind of method and apparatus of marketing data processing.
In a first aspect, the present invention provides a kind of method of marketing data processing, the method packet of the marketing data processing
It includes:
The positive and negative sample set of marketing training model is established according to sales service basic data；
Marketing training model is established according to the positive and negative sample set and random forests algorithm；
Initial marketing data is determined according to sales service data and marketing training model to be predicted；
The marketing training model is updated using the initial marketing data；
Marketing data processing is carried out using updated marketing training model, generates marketing listings data.
Further, described to include: using the initial marketing data update marketing training model
Using the data for failure of marketing in initial marketing data as negative sample, it is added in positive and negative sample set；
According to the positive and negative sample set and random forests algorithm after addition negative sample, the marketing training model is updated.
Further, described to carry out marketing data processing using updated marketing training model, generate marketing inventory number
According to including:
The training quota of the training quota of comparative marketing training pattern and updated marketing training model；
It determines European between the training quota of marketing training model and the training quota of updated marketing training model
Distance is less than preset threshold, then carries out marketing data processing using updated marketing training model, generates marketing listings data.
Further, the training quota includes:
Accuracy rate, recall rate, F1-SCORE.
Further, the positive and negative sample set for establishing marketing training model according to sales service basic data includes:
Obtain the sales service basic data of different platform；
Sample classification is carried out to the sales service basic data according to sales service situation, establishes marketing training model
Positive and negative sample set.
Further, described that sample classification is carried out to the sales service basic data according to sales service situation, it establishes
The positive and negative sample set of marketing training model includes:
According to business rule, the sales service basic data that positive sample is mistaken for negative sample is rejected, establishes and rejects erroneous judgement
Positive and negative sample set afterwards.
Second aspect, the present invention provide a kind of device of marketing data processing, and the device of the marketing data processing includes:
Positive and negative sample set establishes module, for establishing the positive negative sample of marketing training model according to sales service basic data
Collection；
Marketing training model building module, for establishing marketing training according to the positive and negative sample set and random forests algorithm
Model；
Initial marketing data determining module, for being determined just according to sales service data to be predicted and marketing training model
Beginning marketing data；
Update module, for updating the marketing training model using the initial marketing data；
Marketing listings data generation module, it is raw for utilizing updated marketing training model to carry out marketing data processing
At marketing listings data.
Further, the update module includes:
Adding unit, for being added to positive negative sample using the data for failure of marketing in initial marketing data as negative sample
It concentrates；
Updating unit, for updating the marketing according to the positive and negative sample set and random forests algorithm after addition negative sample
Training pattern.
Further, the marketing listings data generation module includes:
The training of comparing unit, training quota and updated marketing training model for comparative marketing training pattern refers to
Mark；
Marketing listings data generation unit, for determining the training quota and updated marketing training of marketing training model
Euclidean distance between the training quota of model is less than preset threshold, then carries out marketing number using updated marketing training model
According to processing, marketing listings data is generated.
Further, the training quota includes:
Accuracy rate, recall rate, F1-SCORE.
Further, the positive and negative sample set establishes module and includes:
Data cell is obtained, for obtaining the sales service basic data of different platform；
Positive and negative sample set establishes unit, for carrying out sample to the sales service basic data according to sales service situation
Classification, establishes the positive and negative sample set of marketing training model.
Further, the positive and negative sample set establishes unit and includes:
Culling unit, for rejecting the sales service basic data that positive sample is mistaken for negative sample according to business rule,
Establish the positive and negative sample set after rejecting erroneous judgement.
The third aspect, the present invention provides a kind of electronic equipment, including memory, processor and storage are on a memory and can
The computer program run on a processor, the processor realize the marketing data that first aspect provides when executing described program
The step of method of processing.
Fourth aspect, the present invention provide a kind of non-transient computer readable storage medium, are stored thereon with computer program,
The step of method for the marketing data processing that first aspect provides is realized when the computer program is executed by processor.
The method of marketing data processing provided by the invention, is become marking automatically, effectively reduces business people from manual markings
Member's mark workload, promotes mark accuracy rate；And model modification strategy is updated by single cover type, is become according to intelligent strategy
It updates, the difficulty that the business of reduction sets model parameter realizes intelligent updating, improves marketing success rate.
For above and other objects, features and advantages of the invention can be clearer and more comprehensible, preferred embodiment is cited below particularly,
And cooperate institute's accompanying drawings, it is described in detail below.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is the flow diagram of the method for marketing data processing provided in an embodiment of the present invention；
Fig. 2 is the association schematic diagram of the intelligent marketing system provided in an embodiment of the present invention based on back-propagation algorithm；
Fig. 3 is the structural block diagram of the intelligent marketing system provided in an embodiment of the present invention based on back-propagation algorithm；
Fig. 4 is data acquisition device function structure chart provided in an embodiment of the present invention；
Fig. 5 is sample definition device function structure chart provided in an embodiment of the present invention；
Fig. 6 is model training apparatus function structure chart provided in an embodiment of the present invention；
Fig. 7 is batch forecast apparatus module structure chart provided in an embodiment of the present invention；
Fig. 8 is marketing recommendation apparatus function structure chart provided in an embodiment of the present invention；
Fig. 9 is model timing optimization apparatus module structure chart provided in an embodiment of the present invention；
Figure 10 is marketing effectiveness statistic device function structure chart provided in an embodiment of the present invention；
Figure 11 is data storage device function structure chart provided in an embodiment of the present invention；
Figure 12 is the process flow diagram of the intelligent marketing system provided in an embodiment of the present invention based on back-propagation algorithm；
Figure 13 is the block diagram of the device of marketing data processing provided in an embodiment of the present invention；
Figure 14 is electronic device block diagram provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
It is being faced as the technologies such as big data, artificial intelligence, cloud computing are continued to develop with maturation, traditional marketing mode
Magnanimity client and when numerous products, is usually present certain marketing difficulty: marked erroneous situation often occurs before marketing, will affect
The accuracy rate of model training；In marketing, cannot timely update prediction model of marketing according to the variation of sample set；Lack reality after marketing
When statistics feedback, cannot by recommend success or failure experience share out.To solve the above problems, the embodiment of the present invention provides
A kind of method of marketing data processing, as shown in Figure 1, this method comprises:
Step S101 establishes the positive and negative sample set of marketing training model according to sales service basic data.
Specifically, receiving sales service basic data, positive sample and negative sample are carried out to the sales service basic data
It divides, establishes the positive and negative sample set of marketing training model.It is understood that progress positive sample and negative sample are divided into two classification
Problem, the positive and negative sample classification of accurate definition, adds positive and negative sample label.Wherein, sales service basic data includes: storage client
Essential information, the related account essential information of client, product information, Transaction Information etc..Positive sample and negative sample are put together conjunction
At positive and negative sample set, as the training sample for generating training pattern.
Step S102 establishes marketing training model according to positive and negative sample set and random forests algorithm.
Specifically, calculating the positive negative sample using random forests algorithm using the positive and negative sample set of foundation as training sample
Collection, establishes marketing training model.
Step S103 determines initial marketing data according to sales service data and marketing training model to be predicted.
(1) specifically, obtaining in sales service data input marketing training model to be predicted marketing successfully and seek
Sell the initial marketing data of failure.It meanwhile can see the feedback result of marketing data in real time by the initial marketing data: lose
It loses or successfully；Success or failure experience can will be recommended to share out at this time, client that backstage real-time statistics are marketed, marketing at
The client of function, the client of marketing failure, failure cause, marketing result Real-time Feedback is summarized, and further promotes business value.
Step S104 updates marketing training model using initial marketing data.
Specifically, obtained initial marketing data to be added to the positive and negative sample set of marketing training model, currently to add
Positive and negative sample set after initial marketing data is as newest training sample, in conjunction with random forests algorithm, updates marketing training
Model obtains updated marketing training model.
Step S105 carries out marketing data processing using updated marketing training model, generates marketing listings data.
Data processing is carried out specifically, marketing data is input in updated marketing training model, is obtained corresponding
Marketing listings data.Wherein marketing listings data can be pushed to customer manager, carry out marketing model.
The method of marketing data processing provided in an embodiment of the present invention, is become marking automatically, be effectively reduced from manual markings
Business personnel marks workload, promotes mark accuracy rate；And model modification strategy is updated by single cover type, is become according to intelligence
Energy policy update, the difficulty that the business of reduction sets model parameter realize intelligent updating, improve marketing success rate.
Content based on the above embodiment, as a kind of alternative embodiment: updating marketing training using initial marketing data
Model includes:
Using the data for failure of marketing in initial marketing data as negative sample, it is added in positive and negative sample set；
According to the positive and negative sample set and random forests algorithm after addition negative sample, marketing training model is updated.
Specifically, the data comprising marketing failure in initial marketing data and successful data of marketing, of the invention real
It applies in example, using the data of marketing failure as negative sample, is added in positive and negative sample set.It is negative to addition using random forests algorithm
Positive and negative sample set after sample carries out model training and obtains updated marketing training model to update marketing training model.
The embodiment of the present invention provides marketing success rate by updating positive and negative sample set with real-time update training pattern.
Content based on the above embodiment, as a kind of alternative embodiment: being carried out using updated marketing training model
Marketing data processing, generating marketing listings data includes:
The training quota of the training quota of comparative marketing training pattern and updated marketing training model；
It determines European between the training quota of marketing training model and the training quota of updated marketing training model
Distance is less than preset threshold, then carries out marketing data processing using updated marketing training model, generates marketing listings data.
Specifically, the training quota of the training quota of comparative marketing training pattern and updated marketing training model, meter
The Euclidean distance between the training quota of marketing training model and the training quota of updated marketing training model is calculated, determines Europe
After formula distance values are less than the threshold value of setting, then using updated marketing training model as final marketing prediction model, carry out
The processing of data generates marketing listings data.
Wherein, training quota includes: accuracy rate, recall rate, F1-SCORE.Herein, the calculating of Euclidean distance is carried out
Citing: assuming that the accuracy rate of the A of marketing training model, recall rate, F1-SCORE are respectively (a1,b1,c1), updated marketing
The accuracy rate of training pattern B, recall rate, F1-SCORE are respectively (a2,b2,c2), then the training of the A of marketing training model refers to
The calculation formula of Euclidean distance n between the training quota of the B of mark and updated marketing training model are as follows:It should be noted that
The embodiment of the present invention is by comparing the training quota of marketing training model and the instruction of updated marketing training model
Practice index, the calculating of Euclidean distance is carried out to each training quota, and evaluation is compared with preset threshold, makes
Prediction model is optimal models, can more improve marketing success rate.
Content based on the above embodiment, as a kind of alternative embodiment: being established and marketed according to sales service basic data
The positive and negative sample set of training pattern includes:
Obtain the sales service basic data of different platform；
Sample classification is carried out to sales service basic data according to sales service situation, establishes the positive and negative of marketing training model
Sample set.
Specifically, obtaining the sales service basic data of the platforms such as Campaign Management Platform, Mobile banking, Web bank, root
Sample classification is carried out to sales service basic data according to sales service situation, establishes the positive and negative sample set of marketing training model, with
Just the generation that marketing training model is completed according to the positive and negative sample set for establishing marketing training model, implements precision marketing.
The embodiment of the present invention establishes complete and comprehensive positive negative sample by the sales service basic data of each platform of acquisition
Collection establishes subsequent training pattern and provides good sample support.
Content based on the above embodiment, as a kind of alternative embodiment: according to sales service situation to sales service base
Plinth data carry out sample classification, and the positive and negative sample set for establishing marketing training model includes:
According to business rule, the sales service basic data that positive sample is mistaken for negative sample is rejected, establishes and rejects erroneous judgement
Positive and negative sample set afterwards.
Specifically, it is for re-filtering to the sales service basic data of marked good positive sample and negative sample, according to industry
Business rule, rejects the sales service basic data that positive sample is mistaken for negative sample, and then establish more accurate marketing training mould
The positive and negative sample set of type.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in Fig. 2, Fig. 2 is that the embodiment of the present invention mentions
Supply the intelligent marketing system based on back-propagation algorithm association schematic diagram, the embodiment of the present invention based on back-propagation algorithm
Intelligent marketing system, it is flat with customer information system 201, client transaction system 202, big data platform 203, machine learning respectively
Platform 204, intelligent marketing system 205, Campaign Management Platform 206, Mobile banking 207, Web bank 208 connect.Wherein, Ke Huxin
Breath system 201, client transaction system 202 are existing banking system, and big data platform 203, machine learning platform 204 are existing
Technology platform, Campaign Management Platform 206, Mobile banking 207, Web bank 208 are existing marketing of bank channel.
Customer information system 201: storage client's essential information, the related account essential information of client.
Client transaction system 202: the Transaction Details of storage client and account.
Big data platform 203 is with machine learning platform 204: being responsible for docking customer information system and client transaction system, adopt
Collection marketing relevant information, form format use data for model training.
Intelligent marketing system 205: docking with big data platform, machine learning platform, establishes smart client marketing system, will
As a result the channels end such as Campaign Management Platform, Mobile banking, Web bank is pushed to.
Campaign Management Platform 206, Mobile banking 207, Web bank 208: existing customer handles marketing channel, for recommendation
As a result, implementing precision marketing.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in figure 3, Fig. 3 is that the embodiment of the present invention mentions
The structural block diagram of the intelligent marketing system based on back-propagation algorithm supplied, in figure the system include: data acquisition device 301,
Sample definition device 302, model training apparatus 303, batch forecast device 304, marketing recommendation apparatus 305, timing optimization device
306, marketing effectiveness statistic device 307, data storage device 308.Wherein, data acquisition device 301 and sample definition device 302
It is connected, sample definition device 302 is connected with data acquisition device 301, model training apparatus 303, model training apparatus 303 and sample
This definition device 302, batch forecast device 304 are connected, and dress is recommended in batch forecast device 304 and model training apparatus 303, marketing
305 are set to be connected.Marketing recommendation apparatus 305 is connected with batch forecast device 304, timing optimization device 306 and marketing recommendation apparatus
305 are connected with marketing effectiveness statistic device 307, and data storage device 308 is connected with marketing effectiveness statistic device 307.Each device
Concrete function is as follows:
Data acquisition device 301: it is responsible for acquisition and marketing associated traffic data, Interworking Data library and hadoop data bins
Library, the comprehensive collection model training use information of feature recommended according to business expert.
Sample definition device 302: (recommending or do not recommend) for two classification problems of marketing, the positive negative sample of accurate definition point
Class adds positive and negative sample label.The present invention is to avoid error caused by manual markings, using technical mark and business expert Xiang Jie
The mode of conjunction, is specifically defined method:
1) technical mark: to have marketed at present, successfully client does not market successfully as negative sample as positive sample,.
2) business expert marks: for there is the probability for being mistaken for positive sample in negative sample, business Expert Rules are collected, it is right
Positive sample is corrected as from negative sample label in the suitable client recommended.
3) execute above-mentioned label 1), 2) step, complete positive and negative sample classification.
Model training apparatus 303: input has been formatted containing label data, selects machine learning algorithm, expansion model training
Step.
Batch forecast device 304: having trained obtained model according to model training apparatus, and to not marketing, client's batch is pre-
It surveys, the probability that can output client contract.
Marketing recommendation apparatus 305: being arranged reasonable threshold value, and the top-tier customer for the certain threshold value that will be greater than is sought as emphasis
It sells object and carries out marketing activity.
Timing optimization device 306: for selecting preferably model as new prediction model.
Marketing effectiveness statistic device 307: statistics marketing success statistics, client's conversion ratio, failure cause.
Data storage device 308: for storing each data information.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in figure 4, Fig. 4 is that the embodiment of the present invention mentions
The data acquisition device function structure chart of confession, device include characteristic acquisition module 3011, characteristic processing module 3012.
Each module concrete function is as follows:
Characteristic acquisition module 3011: collection model input feature vector data adopt homologous ray disparate databases data
Collect unified data management platform.
Characteristic processing module 3012: on the basis of initial characteristic data, according to business meaning, characteristic is carried out
Cleaning is extracted, processing, and (can specifically be unfolded to describe) obtains and the wide table of the matched feature of business meaning.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in figure 5, Fig. 5 is that the embodiment of the present invention mentions
The sample definition device function structure chart of confession, device include positive sample definition module 3021, negative sample definition module 3022.Each mould
Block concrete function is as follows:
Positive sample definition module 3021: signed client adds positive sample label as positive sample using in system.
Negative sample definition module 3022: unsigned client is as negative sample using in system, in combination with business Expert Rules,
Reject the top-tier customer that part is mistaken for negative sample.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in fig. 6, Fig. 6 is that the embodiment of the present invention mentions
The model training apparatus function structure chart of confession, device include data access module 3031, positive and negative mark module 3032, core algorithm
Module 3033, model iteration module 3034.Each module concrete function is as follows:
Data access module 3031: the processed good wide table of feature of access is used as mode input.
Positive and negative mark module 3032: last column of the completed positive and negative label of splicing to the wide table of feature, as model prediction
Column.
Core algorithm module 3033: selecting suitable machine learning algorithm, carries out model training.
Model iteration module 3034: set algorithm stopping criterion for iteration completes models for several times iteration.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in fig. 7, Fig. 7 is that the embodiment of the present invention mentions
The batch forecast apparatus module structure chart of confession, device include prediction data AM access module 3041, model prediction module 3042, prediction
As a result output module 3043, prediction result memory module 3044.Each module concrete function is as follows:
Prediction data AM access module 3041: for client characteristics data of not marketing as prediction input.
Model prediction block module 3042: using prediction data, is input to module 3034 and has trained the model obtained, composition is worked as
Preceding prediction module.
Prediction result output module 3043: output 3042 prediction result of prediction module, each client include to predict whether meeting
The probability of signing.
Prediction result memory module 3044: above-mentioned prediction result is stored into file.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in figure 8, Fig. 8 is that the embodiment of the present invention mentions
The marketing recommendation apparatus function structure chart of confession, device include bank outlets' module 3051, customer manager's mould 3052, marketing inventory push away
It send module 3053, client to contract and monitors mould 3054.Each module concrete function is as follows:
Bank outlets' module 3051: it is responsible for each branch of maintenance, subbranch, site operational management information.
Customer manager's module 3052: it is responsible for specific client and Products Show service is provided.
Marketing inventory pushing module 3053: inventory is recommended according to prediction result memory module 3044, pushes to customer manager
52, specific marketing is unfolded.
Client's signing monitoring module 3054: the high-quality product information of whether contracting and contract of monitoring is counted by site
Summarize.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in figure 9, Fig. 9 is that the embodiment of the present invention mentions
The model timing optimization apparatus module structure chart of confession, device includes prediction model A preserving module, marketing feedback module 3062, pre-
Survey Model B update module 3063, prediction model AB comparison module 3064.Each module concrete function is as follows:
Prediction model A preserving module 3061: the model A relevant parameter for predicting to complete for the first time is saved.
Marketing feedback module 3062: according to marketing as a result, the result of marketing failure is fed back to negative sample definition module
22, it is added to negative sample set.
Prediction model B update module 3063: according to new positive and negative sample set, model training is re-started, is obtained new
Prediction model B.
Prediction model AB comparison module 3064: comparison model A and B is selected more excellent in accuracy rate, recall rate, F1-SCORE
Model as new prediction model.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in Figure 10, Figure 10 is the embodiment of the present invention
The marketing effectiveness statistic device function structure chart of offer, device include marketing success statistics module 3071, client's conversion ratio module
3072, module 3073 is sent on failure cause.Each module concrete function is as follows:
Marketing success statistics module 3071: successfully customer quantity of having marketed is counted.
Client's conversion ratio module 3072: successfully client's number accounting in always marketing customer quantity of having marketed is counted.
Module 3073: system dynamic failure client's failure cause is sent on failure cause.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in figure 11, Figure 11 is the embodiment of the present invention
The data storage device function structure chart of offer, device include training data memory module 3081, prediction data memory module
3082, marketing feedback data memory module 3083, Model Self-Learning parameter memory module 3084.Each module concrete function is as follows:
Training data memory module 3081: storage is used for the data of model training.
Prediction data memory module 3082: storage is used for the data of forecast sample.
Marketing feedback data memory module 3083: storage service marketing feedback data.
Model Self-Learning parameter memory module 3084: the model parameter after storage model training.
Content based on the above embodiment, as a kind of alternative embodiment: as shown in figure 12, Figure 12 is the embodiment of the present invention
The process flow diagram of the intelligent marketing system based on back-propagation algorithm provided, process step includes step S11 training data
Acquisition, the positive negative sample definition of step S12, step S13 random forests algorithm, step S14 model training, step S21 prediction data
Acquisition, step S22 feature machining, step S23 model prediction, step S24 marketing recommendation inventory, step S31 marketing successful result,
Step S32 marketing failure result, step S41 backpropagation.Each module concrete function is as follows:
Step S11 collecting training data: acquisition intelligent recommendation model basic data.
The positive negative sample definition of step S12: positive negative sample is divided, training label column is added.
Step S13 random forests algorithm: selection random forests algorithm carries out model training.
Step S14 model training: setting stopping criterion for iteration, iteration complete model training.
The acquisition of step S21 prediction data: the client that do not market to storage carries out data acquisition.
Step S22 feature machining: processing prediction data characteristic information is spliced to form the wide table of feature.
Step S23 model prediction: using training pattern, the wide table of predictive data set feature is predicted.
Inventory is recommended in step S24 marketing: the marketing that distribution prediction obtains recommends inventory to business marketing channel end.
Step S31 marketing successful result: the practical marketing successful result of business is defined as successfully sample.
Step S32 marketing failure result: the practical marketing failure result of business is defined as unsuccessfully sample.
Step S41 backpropagation: above-mentioned failure result is propagated back in negative sample, and re -training obtains new model.
It is specifically divided into the following four stage:
Model training stage: since step S11 collecting training data, completing to acquire to related service basic data, then
Further the positive negative sample of step S12 is defined, it is complete using the strong step S13 random forests algorithm of a kind of stabilization, noise resisting ability
At step S14 model training, intelligent recommendation model is obtained.
The model prediction stage: since the acquisition of step S21 prediction data, feature machining is carried out to step S22 data, simultaneously
In conjunction with the acquired model of step S14 training, step S23 model prediction is carried out, step S24 obtains marketing and recommends inventory.
The business marketing stage: the inventory that step S24 is obtained is split to different bank's nets according to area, site information
Point after the completion of marketing, obtains step S31 marketing successful result or step S32 marketing failure result.
Back-propagation phase: step S41 backpropagation realizes that the failure result anti-pass that will market to negative sample set, is instructed again
Model accuracy rate, recall rate, F1-SCORE are obtained after white silk.Three indexs and first model training result are calculated into Euclidean distance,
Such as less than given threshold then updates intelligent marketing model.
Herein, the present embodiments relate to backpropagation it is theoretical derived from Neurobiology, i.e., neuron is receiving
After forward direction stimulation, result is fed back into organism itself, makes stress reaction.BP algorithm (back-propagation algorithm) is suitable for more
A kind of machine learning algorithm of layer neural network is built upon on the basis of gradient decline.Similar, complete process of marketing, intelligence
After the result of Marketing Model recommends in customer manager's hand, the present invention uses BP algorithm, and failure result is fed back to model training
Initial stage, training pattern adjust negative sample input, re -training.
According to another aspect of the present invention, the embodiment of the present invention also provides a kind of device of marketing data processing, referring to
Figure 13, Figure 13 are the block diagrams of the device of marketing data processing provided in an embodiment of the present invention.The device is used in aforementioned each implementation
The processing of marketing data is carried out in example.Therefore, the description in the method for the marketing data processing in foregoing embodiments and fixed
Justice can be used for the understanding of each execution module in the embodiment of the present invention.
As shown, the device includes:
Positive and negative sample set establishes module 1301, for establishing the positive and negative of marketing training model according to sales service basic data
Sample set；
Marketing training model building module 1302, for establishing marketing training according to positive and negative sample set and random forests algorithm
Model；
Initial marketing data determining module 1303, for true according to sales service data and marketing training model to be predicted
Fixed initial marketing data；
Update module 1304, for updating marketing training model using initial marketing data；
Marketing listings data generation module 1305, for being carried out at marketing data using updated marketing training model
Reason generates marketing listings data.
The method of marketing data processing provided in an embodiment of the present invention, is become marking automatically, be effectively reduced from manual markings
Business personnel marks workload, promotes mark accuracy rate；And model modification strategy is updated by single cover type, is become according to intelligence
Energy policy update, the difficulty that the business of reduction sets model parameter realize intelligent updating, improve marketing success rate.
Content based on the above embodiment, as a kind of alternative embodiment: update module includes:
Adding unit, for being added to positive negative sample using the data for failure of marketing in initial marketing data as negative sample
It concentrates；
Updating unit, for updating marketing training according to the positive and negative sample set and random forests algorithm after addition negative sample
Model.
The embodiment of the present invention provides marketing success rate by updating positive and negative sample set with real-time update training pattern.
Content based on the above embodiment, as a kind of alternative embodiment: marketing listings data generation module includes:
The training of comparing unit, training quota and updated marketing training model for comparative marketing training pattern refers to
Mark；
Marketing listings data generation unit, for determining the training quota and updated marketing training of marketing training model
Euclidean distance between the training quota of model is less than preset threshold, then carries out marketing number using updated marketing training model
According to processing, marketing listings data is generated.
The embodiment of the present invention is by comparing the training quota of marketing training model and the instruction of updated marketing training model
Practice index, the calculating of Euclidean distance is carried out to each training quota, and evaluation is compared with preset threshold, makes
Prediction model is optimal models, can more improve marketing success rate.
Content based on the above embodiment, as a kind of alternative embodiment: positive and negative sample set establishes module and includes:
Data cell is obtained, for obtaining the sales service basic data of different platform；
Positive and negative sample set establishes unit, for carrying out sample point to sales service basic data according to sales service situation
Class establishes the positive and negative sample set of marketing training model.
The embodiment of the present invention establishes complete and comprehensive positive negative sample by the sales service basic data of each platform of acquisition
Collection establishes subsequent training pattern and provides good sample support.
Content based on the above embodiment, as a kind of alternative embodiment: positive and negative sample set establishes unit and includes:
Culling unit, for rejecting the sales service basic data that positive sample is mistaken for negative sample according to business rule,
Establish the positive and negative sample set after rejecting erroneous judgement.
Figure 14 is electronic device block diagram provided in an embodiment of the present invention, and as shown in figure 14, which includes: processor
1401, memory 1402 and bus 1403；
Wherein, processor 1401 and storage 1402 complete mutual communication by bus 1403 respectively；Processor 1401
For calling the program instruction in memory 1402, to execute the method that marketing data provided by above-described embodiment is handled, example
It such as include: the positive and negative sample set that marketing training model is established according to sales service basic data；According to positive and negative sample set and at random
Forest algorithm establishes marketing training model；Initial marketing number is determined according to sales service data to be predicted and marketing training model
According to；Marketing training model is updated using initial marketing data；Marketing data processing is carried out using updated marketing training model,
Generate marketing listings data.
The embodiment of the present invention provides a kind of non-transient computer readable storage medium, is stored thereon with computer program, should
The step of method of marketing data processing is realized when computer program is executed by processor.For example, according to sales service base
Plinth data establish the positive and negative sample set of marketing training model；Marketing training mould is established according to positive and negative sample set and random forests algorithm
Type；Initial marketing data is determined according to sales service data and marketing training model to be predicted；More using initial marketing data
New marketing training model；Marketing data processing is carried out using updated marketing training model, generates marketing listings data.
The apparatus embodiments described above are merely exemplary, wherein unit can be as illustrated by the separation member
Or may not be and be physically separated, component shown as a unit may or may not be physical unit, i.e.,
It can be located in one place, or may be distributed over multiple network units.It can select according to the actual needs therein
Some or all of the modules achieves the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creative labor
In the case where dynamic, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
The method of certain parts of example or embodiment.
Finally, applying specific embodiment in the present invention, principle and implementation of the present invention are described, above
The explanation of embodiment is merely used to help understand method and its core concept of the invention；Meanwhile for the general skill of this field
Art personnel, according to the thought of the present invention, there will be changes in the specific implementation manner and application range, in conclusion this
Description should not be construed as limiting the invention.
Claims (14)
1. a kind of method of marketing data processing, which is characterized in that the method for the marketing data processing includes:
The positive and negative sample set of marketing training model is established according to sales service basic data；
Marketing training model is established according to the positive and negative sample set and random forests algorithm；
Initial marketing data is determined according to sales service data and marketing training model to be predicted；
The marketing training model is updated using the initial marketing data；
Marketing data processing is carried out using updated marketing training model, generates marketing listings data.
2. the method for marketing data processing according to claim 1, which is characterized in that described to utilize the initial marketing number
Include: according to the marketing training model is updated
Using the data for failure of marketing in initial marketing data as negative sample, it is added in positive and negative sample set；
According to the positive and negative sample set and random forests algorithm after addition negative sample, the marketing training model is updated.
3. the method for marketing data processing according to claim 1, which is characterized in that described to be instructed using updated marketing
Practice model and carry out marketing data processing, generating marketing listings data includes:
The training quota of the training quota of comparative marketing training pattern and updated marketing training model；
Determine the Euclidean distance between the training quota of marketing training model and the training quota of updated marketing training model
Less than preset threshold, then marketing data processing is carried out using updated marketing training model, generates marketing listings data.
4. the method for marketing data according to claim 3 processing, which is characterized in that the training quota includes:
Accuracy rate, recall rate, F1-SCORE.
5. the method for marketing data processing according to claim 1, which is characterized in that described according to sales service basis number
Include: according to the positive and negative sample set for establishing marketing training model
Obtain the sales service basic data of different platform；
Sample classification is carried out to the sales service basic data according to sales service situation, establishes the positive and negative of marketing training model
Sample set.
6. the method for marketing data processing according to claim 5, which is characterized in that described according to sales service situation pair
The sales service basic data carries out sample classification, and the positive and negative sample set for establishing marketing training model includes:
According to business rule, the sales service basic data that positive sample is mistaken for negative sample is rejected, is established after rejecting erroneous judgement
Positive and negative sample set.
7. a kind of device of marketing data processing, which is characterized in that the device of the marketing data processing includes:
Positive and negative sample set establishes module, for establishing the positive and negative sample set of marketing training model according to sales service basic data；
Marketing training model building module, for establishing marketing training mould according to the positive and negative sample set and random forests algorithm
Type；
Initial marketing data determining module, for determining initial battalion according to sales service data to be predicted and marketing training model
Sell data；
Update module, for updating the marketing training model using the initial marketing data；
Listings data generation module of marketing is generated and is sought for carrying out marketing data processing using updated marketing training model
Sell listings data.
8. the device of marketing data according to claim 7 processing, which is characterized in that the update module includes:
Adding unit, for being added in positive and negative sample set using the data for failure of marketing in initial marketing data as negative sample；
Updating unit, for updating the marketing training according to the positive and negative sample set and random forests algorithm after addition negative sample
Model.
9. the device of marketing data processing according to claim 7, which is characterized in that the marketing listings data generates mould
Block includes:
Comparing unit, for the training quota of comparative marketing training pattern and the training quota of updated marketing training model；
Marketing listings data generation unit, for determining the training quota and updated marketing training model of marketing training model
Training quota between Euclidean distance be less than preset threshold, then using updated marketing training model progress marketing data at
Reason generates marketing listings data.
10. the device of marketing data according to claim 9 processing, which is characterized in that the training quota includes:
Accuracy rate, recall rate, F1-SCORE.
11. the device of marketing data processing according to claim 7, which is characterized in that the positive and negative sample set establishes mould
Block includes:
Data cell is obtained, for obtaining the sales service basic data of different platform；
Positive and negative sample set establishes unit, for carrying out sample point to the sales service basic data according to sales service situation
Class establishes the positive and negative sample set of marketing training model.
12. the device of marketing data processing according to claim 11, which is characterized in that the positive and negative sample set is established single
Member includes:
Culling unit is established for rejecting the sales service basic data that positive sample is mistaken for negative sample according to business rule
Positive and negative sample set after rejecting erroneous judgement.
13. a kind of electronic equipment including memory, processor and stores the calculating that can be run on a memory and on a processor
Machine program, which is characterized in that number of marketing as described in any one of claim 1 to 6 is realized when the processor executes described program
According to processing method the step of.
14. a kind of non-transient computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer
The step of method that the marketing data as described in any one of claim 1 to 6 is handled is realized when program is executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910639120.9A CN110363244A (en) | 2019-07-16 | 2019-07-16 | A kind of method and apparatus of marketing data processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910639120.9A CN110363244A (en) | 2019-07-16 | 2019-07-16 | A kind of method and apparatus of marketing data processing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110363244A true CN110363244A (en) | 2019-10-22 |
Family
ID=68219245
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910639120.9A Pending CN110363244A (en) | 2019-07-16 | 2019-07-16 | A kind of method and apparatus of marketing data processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110363244A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111047364A (en) * | 2019-12-23 | 2020-04-21 | 杭州数澜科技有限公司 | Method and system for implementing accurate marketing |
| CN111667024A (en) * | 2020-06-30 | 2020-09-15 | 腾讯科技（深圳）有限公司 | Content pushing method and device, computer equipment and storage medium |
| CN112927012A (en) * | 2021-02-23 | 2021-06-08 | 第四范式（北京）技术有限公司 | Marketing data processing method and device and marketing model training method and device |
| CN113902471A (en) * | 2021-09-26 | 2022-01-07 | 科大讯飞股份有限公司 | Marketing strategy updating method and device and computer readable storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107689008A (en) * | 2017-06-09 | 2018-02-13 | 平安科技（深圳）有限公司 | A kind of user insures the method and device of behavior prediction |
| CN108648011A (en) * | 2018-05-11 | 2018-10-12 | 上海赢科信息技术有限公司 | Model generates, identification client buys the method and system of vehicle insurance intention |
| CN109558892A (en) * | 2018-10-30 | 2019-04-02 | 银河水滴科技（北京）有限公司 | A kind of target identification method neural network based and system |
| CN109598264A (en) * | 2017-09-30 | 2019-04-09 | 北京猎户星空科技有限公司 | Grasping body method and device |
| CN109784966A (en) * | 2018-11-29 | 2019-05-21 | 昆明理工大学 | A kind of music website customer churn prediction method |
-
2019
- 2019-07-16 CN CN201910639120.9A patent/CN110363244A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107689008A (en) * | 2017-06-09 | 2018-02-13 | 平安科技（深圳）有限公司 | A kind of user insures the method and device of behavior prediction |
| CN109598264A (en) * | 2017-09-30 | 2019-04-09 | 北京猎户星空科技有限公司 | Grasping body method and device |
| CN108648011A (en) * | 2018-05-11 | 2018-10-12 | 上海赢科信息技术有限公司 | Model generates, identification client buys the method and system of vehicle insurance intention |
| CN109558892A (en) * | 2018-10-30 | 2019-04-02 | 银河水滴科技（北京）有限公司 | A kind of target identification method neural network based and system |
| CN109784966A (en) * | 2018-11-29 | 2019-05-21 | 昆明理工大学 | A kind of music website customer churn prediction method |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111047364A (en) * | 2019-12-23 | 2020-04-21 | 杭州数澜科技有限公司 | Method and system for implementing accurate marketing |
| CN111667024A (en) * | 2020-06-30 | 2020-09-15 | 腾讯科技（深圳）有限公司 | Content pushing method and device, computer equipment and storage medium |
| CN111667024B (en) * | 2020-06-30 | 2023-09-05 | 腾讯科技（深圳）有限公司 | Content pushing method, device, computer equipment and storage medium |
| CN112927012A (en) * | 2021-02-23 | 2021-06-08 | 第四范式（北京）技术有限公司 | Marketing data processing method and device and marketing model training method and device |
| CN113902471A (en) * | 2021-09-26 | 2022-01-07 | 科大讯飞股份有限公司 | Marketing strategy updating method and device and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110363244A (en) | A kind of method and apparatus of marketing data processing | |
| CN107038486A (en) | Financial product prediction meanss and financial product Forecasting Methodology | |
| US20210073695A1 (en) | Production scheduling system and method | |
| CN86101057A (en) | Be used for method of allocating resources and equipment effectively | |
| CN109993414A (en) | Evaluation method and device for innovation development of power enterprise and storage medium | |
| CN113159835B (en) | Power generation side electricity price quotation method and device based on artificial intelligence, storage medium and electronic equipment | |
| CN110445939B (en) | Capacity resource prediction method and device | |
| CN111581516A (en) | Investment product recommendation method and related device | |
| CN109872052A (en) | A kind of law court's case intelligence division householder method and system | |
| CN110796331A (en) | Power business collaborative classification method and system based on C4.5 decision tree algorithm | |
| CN110766272A (en) | Power business collaborative classification method and system based on ID3 decision tree algorithm | |
| CN107067282B (en) | Consumer product rebate sale marketing management system and use method thereof | |
| CN111932106B (en) | Effective and practical cloud manufacturing task and service resource matching method | |
| CN110147389A (en) | Account number treating method and apparatus, storage medium and electronic device | |
| CN110837966A (en) | SNA-based building equipment optimization method based on mixed multi-attribute group decision | |
| CN112449002A (en) | Method, device and equipment for pushing object to be pushed and storage medium | |
| CN117217820A (en) | Intelligent integrated prediction method and system for purchasing demand of supply chain | |
| Seo et al. | Auction-based federated learning using software-defined networking for resource efficiency | |
| CN111988786A (en) | Sensor network covering method and system based on high-dimensional multi-target decomposition algorithm | |
| CN106919998A (en) | A kind of physical distribution model and method based on AOE networks | |
| CN103903070B (en) | Application system resource requirement calculating system | |
| CN112884572A (en) | Multi-microgrid transaction optimization method and system under credit risk | |
| CN111915102A (en) | Load prediction-based work platform workload prediction method and system | |
| CN104036403A (en) | Supply chain network graded-assessment system based on repeated game theory | |
| CN116128247A (en) | Resource allocation optimization method and system for production equipment before scheduling |
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: 20191022 |
|
| RJ01 | Rejection of invention patent application after publication |