CN108921602A - A kind of user's buying behavior prediction technique based on integrated neural network - Google Patents
A kind of user's buying behavior prediction technique based on integrated neural network Download PDFInfo
- Publication number
- CN108921602A CN108921602A CN201810642096.XA CN201810642096A CN108921602A CN 108921602 A CN108921602 A CN 108921602A CN 201810642096 A CN201810642096 A CN 201810642096A CN 108921602 A CN108921602 A CN 108921602A
- Authority
- CN
- China
- Prior art keywords
- user
- neural network
- sample set
- integrated
- classifier
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- 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
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/123—DNA computing
-
- 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
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
Abstract
User's buying behavior prediction technique based on integrated neural network that the invention discloses a kind of, including step：1) feature extraction and sampling are carried out to user behavior historical record, obtains sample set T1；2) classifier C1 is formed using Boosting integrated approach, classification processing is carried out to sample set T1 and feature is integrated, obtains new sample set T2；3) basic structure for constructing neural network carries out heuristic search using parameter of the genetic algorithm to neural network, forms integrated neural network classifier C2；4) classification processing is carried out to sample set T2 using classifier C2, obtains new sample set T3；5) classifier C3 is formed using Bagging integrated approach, classification processing is carried out to sample set T3, the item lists of buying behavior can be generated by obtaining user, as the prediction result for user's buying behavior.The present invention solve conventional method classifying quality is poor, generalization is poor, under big data scene the problems such as inefficiency.
Description
Technical field
The present invention relates to the technical fields of e-commerce, refer in particular to a kind of user's purchase row based on integrated neural network
For prediction technique.
Background technique
With the arrival of big data era and the prevalence of user's shopping on the web, under big data background, by outstanding
Proposed algorithm recommends the interested commodity of user, improve commodity exposure rate, increase user's purchase volume behavior at
For one of the basic function of an e-commerce system.And for the accurate prediction of user's buying behavior, then it is proposed algorithm
Final goal.If businessman can grasp the purchase intention of consumer, businessman can arranged rational commodity inventory,
Accurate user's portrait can be constructed, the personnel such as market, marketing is fed back to, carries out targeted merchandise sales.It can be seen that
There is important theory and realistic meaning to the prediction of user's buying behavior.
Currently, the prediction technique of user's buying behavior is broadly divided into three kinds in e-commerce field, the first is to utilize people
The rule that work defines, such as in certain day, some article has been placed into shopping cart but has not bought, then it is likely used only at second day
It will do it purchase；Also the methods of statistical analysis user carried out including the use of forms such as questionnaires；Second is to utilize biography
The proposed algorithm of system, such as collaborative filtering, content-based recommendation are predicted that these methods, which all have been proven that, is recommending to lead
There is certain effect in domain；The third is will to buy prediction to be regarded as two classification problems, using typical Machine learning classifiers,
Such as support vector machines, decision tree predict user's buying behavior by the model of one user behavior of training.This three
Kind method can predict the buying behavior of user to a certain extent, but all have the shortcomings that.Artificial method largely relies on
In the labour of the mankind, and analyst coverage is narrow, and accuracy rate is low；Traditional proposed algorithm can only recommended user may be interested
Commodity score to the interested degree of user, and the prediction that whether can be bought for user can not be obtained by algorithm itself
To as a result, still relying on artificial screening and evaluation；Traditional classifier methods prediction result accuracy rate is lower, model generalization
Property it is poor, and these three methods all exist under the scene of big data inefficiency, accuracy rate decline the problems such as.
The present invention proposes a kind of user's buying behavior prediction technique based on integrated neural network, by user, the spy of article
After sign is extracted, the buying behavior of user is predicted by a variety of integrated approaches and the neural network of optimization, it is non-using neural network
Linearly, the advantages that adaptivity is strong solves that conventional method classifying quality is poor, generalization is poor, big in conjunction with the method for integrated study
Amount relies on artificial problem, improves the efficiency of prediction and the accuracy rate of prediction result.
Summary of the invention
It is an object of the invention to overcome the deficiencies in the prior art, propose a kind of user's purchase based on integrated neural network
Buy behavior prediction method, solve conventional method classifying quality is poor, generalization is poor, under big data scene the problems such as inefficiency,
So that the efficiency of prediction model, generalization and the accuracy rate under big data scene are improved.
To achieve the above object, technical solution provided by the present invention is：A kind of user's purchase based on integrated neural network
Behavior prediction method is bought, is included the following steps：
1) feature extraction and sampling are carried out to user behavior historical record, therefrom obtain can summarize user, article characteristics and
The sample set T1 of user's purchase label；
2) classifier C1 is formed using Boosting integrated approach, classification processing and feature set is carried out to sample set T1
At obtaining new sample set T2；
3) basic structure for constructing neural network carries out heuristic search using parameter of the genetic algorithm to neural network,
Optimal neural network, and the optimal Artificial neural network ensemble that will be obtained are obtained, integrated neural network classifier C2 is formed；
4) classification processing is carried out to sample set T2 using classifier C2, wherein each neural network can handle sample
A part in set, and carry out result fusion and feature before the output of classifier C2 and integrate, finally obtain new sample set
Close T3；
5) classifier C3 is formed using Bagging integrated approach, classification processing is carried out to sample set T3, obtains user's meeting
The item lists for generating buying behavior, as the prediction result for user's buying behavior.
In step 1), the user behavior historical record includes the mark, the mark of article, user behavior of user
The geographical location information of the time and interaction of classification and user and article interaction；Pass through the side for statisticalling analyze with manually inferring
Formula, user, article feature can be embodied by therefrom extracting, and the feature of prediction user behavior tendency is adopted using positive and negative sample equilibrium
Sample and outlier processing obtain the characteristic of original sample set；Whether user finally buys as original sample set
Label segment；Characteristic and label segment collectively form sample set T1.
In step 2), decision tree progress Boosting is integrated, the classifier C1 of formation is that a gradient promotes classification
Device；For the sample set T1 extracted from user behavior historical record, characteristic X1, label segment y1；By X1
As the input of classifier C1, prediction result y2 is obtained, one bivector of each behavior, respectively representing prediction result is not
The probability and prediction result of purchase are the probability of purchase；It is incorporated into original sample characteristics X1 using y2 as new feature, shape
The new sample set that the sample characteristics X2 of Cheng Xin, sample characteristics X2 and label segment y1 are formed is known as T2.
In step 3), for constructed neural network, it is made of 1 input layer, 1 output layer and 3 hidden layers,
The neuron number of input layer is the Characteristic Number of input data, and the neuron number of output layer is 2, represents two-dimensional prediction knot
Fruit, hidden layer then by the way of connecting entirely；The hidden layer number of plies of neural network, each layer of neuron number, neural network
The number of iterations and each iteration size parameter all by genetic algorithm carry out heuristic search obtain, pass through the iteration of genetic algorithm
Obtain the optimal solution of parameters, and so as to form optimal neural network.
In step 4), for new sample set T2, after it is normalized, its equal proportion is divided into n
Part, be referred to as T2_1, T2_2, T2_3 ..., T2_n；For a neural network in classifier C2, its training set is
N-1 part in T2 set, test set are remaining 1 part；Neural network model after n training respectively to respective test set into
Row prediction, obtains n prediction result, referred to as y3_1, y3_2, y3_3 ..., y3_n, this n prediction result is merged into new
Prediction result y3；It is incorporated to y3 as new feature in sample characteristics X2, forms new sample characteristics X3；Sample characteristics X3 and mark
The new sample set that label part y1 is formed is known as T3.
In step 5), decision tree progress Bagging is integrated, the classifier C3 of formation is a random forest classification
Device；Prediction result y4 is obtained using its characteristic X3 as the input of classifier C3 for new sample set T3, it is each
One bivector of behavior, respectively represent prediction result be the probability that do not buy and prediction result be purchase probability；To prediction
As a result after being ranked up according to the probability of purchase, the item lists of buying behavior can be generated by just obtaining user, as final defeated
Result out.
Compared with prior art, the present invention having the following advantages that and beneficial effect：
1, the purchase that the present invention allows computer to carry out future according to the history buying behavior of user by training pattern is pre-
It surveys, generated in model and does not need artificial excessive intervention in subsequent use, liberated manual labor, improve prediction effect
Rate.
2, the present invention has outstanding performance under the scene of big data, compared to proposed algorithms such as traditional collaborative filterings
The problems such as inefficiency, accuracy rate decline when facing big data with two sorting algorithms, the present invention can be very good to handle big number
According to scene, for a large amount of historical data and need the user data predicted, the present invention can be handled successfully.
3, present invention utilizes Neural Network Based Nonlinear, adaptivity is strong the features such as and integrated study fusion feature, altogether
, there is very big promotion in the advantages of with decision compared to traditional algorithm and artificial treatment in accuracy rate.
4, the present invention is adjusted the parameter of neural network using genetic algorithm and the method for heuristic search, optimizes
The best part is influenced for accuracy rate in neural network building, compared to not optimized neural network, the present invention is used
Have better generalization and higher accuracy rate.
5, the present invention has a wide range of applications space in e-commerce field, and easily operated and personalized customization and
Upgrading, has broad application prospects.
Detailed description of the invention
Fig. 1 is the method for the present invention flow chart.
Fig. 2 is genetic algorithm flow chart.
Fig. 3 is the structural schematic diagram of neural network.
Fig. 4 is the schematic diagram of integrated neural network.
Specific embodiment
In order to more specifically describe the present invention, with reference to the accompanying drawing and specific embodiment is to technical solution of the present invention
It is described in detail.
In the present embodiment, a kind of user's buying behavior prediction technique based on integrated neural network is applied and is being moved by we
On dynamic recommending data collection.As shown in Figure 1, user's buying behavior prediction technique mainly comprises the following steps：
Firstly, carrying out feature extraction and sampling to user behavior historical record, the spies such as user, article can be summarized by therefrom obtaining
The sample set of sign；For the present embodiment, since data set derives from the truthful data of Ali, it is therefore desirable to first carry out pre-
Processing operation removes the abnormal data in data set, such as never buys a large amount of in the user and advertising campaign day of doubtful robot
Purchase data of growth etc..Feature calculation then is carried out to by pretreated data, the feature of user mainly includes user
The behaviors such as click, purchase number, the number for being clicked, being purchased of article and number after combining between the two it
With the statistical informations such as average value.For the present embodiment, defines a label and calculate day, as need to predict whether purchase
Date, then the data of a couple of days of feature before label day, positive sample are in the user-object for having buying behavior label day
Product pair, negative sample are user-article pair of not buying behavior.By the feature extracted together with label, it is combined into sample set
Close T1.
Then, classifier is formed using Boosting integrated approach, for the sample extracted from user behavior historical record
This set T1, is predicted using integrated classifier.In the present embodiment, we use gradient boosted tree as Boosting
The classifier of Integrated Decision tree, loss function use log-likelihood function, and the depth capacity of tree is set as 4, learning rate 0.05,
Sample set of extraction is predicted with this, obtains prediction result y2.Prediction result y2 is merged into the spy of sample set T1
Part is levied, new sample characteristics, and the sample set T2 with original tag combination Cheng Xin are formed.
Then, for new sample set T2, after being normalized and being formed one-hot coding, as multiple minds
The input of classifier C2 is formed by through system integrating；The hyper parameter setting of neural network will be carried out heuristic using genetic algorithm
Search.Fig. 2 illustrates the flow chart of genetic algorithm an iteration of the present invention, firstly generates a population, contain it is hundreds of with
The list for the hyper parameter that machine generates, the referred to as individual of population.Contained in list neural network hyper parameter have the hidden layer number of plies,
Hidden layer neuron number, neural network the number of iterations, neural network iteration block size etc..
We according to the hyper parameter of each individual, construct a neural network first, then carry out sample to the neural network
The training and classification of this set, obtain the classification predictablity rate of the neural network, and evaluation index is prediction result and sample mark
Cross entropy between label.Carry out classification prediction by each individual to population, an available best accuracy rate and
Corresponding best hyper parameter list.Then, we start to carry out the iteration of genetic algorithm, and detailed process is：It is selected from population
Two individuals are taken, it is intersected and be mutated according to certain probability.The parameter intersected between as two individuals is melted mutually
It closes, forms new hyper parameter list；Mutation is then to change the value of random several parameters in list with certain probability.Later with same
The method building neural network of sample, classification are predicted and assess the effect of the individual after changing.By after change individual effect with
The best effect recorded is compared, if the individual after changing is more excellent, the individual of script is replaced with it.Passing through
After thousands of secondary iteration, algorithm, which will restrain and provide a best hyper parameter, is arranged list.We will be with this hyper parameter list
The hyper parameter of neural network in algorithm is arranged.Fig. 3 is the structural schematic diagram of neural network.Each neural network is by 1
Input layer, 1 output layer and multiple hidden layers composition, the neuron number of input layer is the dimension of X2 in sample set T2, defeated
The neuron number of layer is 2 out, represents two-dimensional prediction result, hidden layer then by the way of connecting entirely, hyper parameter according to
The optimal hyper parameter list that genetic algorithm obtains is configured；The activation primitive of neural network uses ReLu function, and formula is：
F (x)=max (0, x)
Function of the Softmax function as output layer, formula are：
The output function of Softmax can carry out probabilistic forecasting to two-dimensional output, can be regarded as each label and occur
Probability, i.e. user buys and the probability do not bought.
Fig. 4 illustrates the schematic diagram of integrated neural network.In the present embodiment, its equal proportion is divided into 5 parts by us, point
It is also known as T2_1, T2_2, T2_3, T2_4, T2_5；For one of feedforward neural network, its training set is in T2 set
4 parts, test set is remaining 1 part, for example, for feedforward neural network 1, T2_1 as its test set, T2_2, T2_3,
The training set of T2_4, T2_5 as it；Neural network model after 5 training respectively predicts respective test set,
5 prediction results, referred to as y3_1, y3_2, y3_3, y3_4, y3_5 are obtained, this 5 prediction results are merged into new prediction knot
Fruit y3；
Y3=y3_1 ∪ y3_2 ∪ y3_3 ∪ y3_4 ∪ y3_5
It is incorporated to y3 as new feature in sample characteristics X2, forms new sample characteristics X3；
X3={ X2, y3 }
The new sample set that sample characteristics X3 and label segment y1 is formed is known as T3.
Finally, forming classifier using Bagging integrated approach.In the present embodiment, we use random forests algorithm
The decision tree classifier integrated as Bagging.For new sample set T3, using its characteristic X3 as classifier C3's
Input.The big feature of random forest selection gini index yield value is further divided；By repeatedly have put back to it is random
After sample drawn and feature training, final classification is determined to sample data according to the voting results of decision tree in forest.By
After classification, prediction result y4 is obtained, one bivector of each behavior, having respectively represented prediction result is the probability that do not buy
It is the probability of purchase with prediction result；After being ranked up to the probability that prediction result is purchase, it is most possible user has just been obtained
The list of the article of purchase, and using probability highest K as a result, i.e. Top-K buys article as final output result.
For exporting the assessment of result, we use F1 value as evaluation criterion.F1 value is according to accurate rate and to recall
What rate was calculated jointly.Their own formula is as follows：
Wherein P is accurate rate, and R is recall rate, and predictionSet is final prediction result set,
ReferenceSet is the results set of actual user's purchase.By constantly assessing the result of classifier, so that it may adjust
The parameter and setting of whole classifier, to obtain best disaggregated model.
For prediction, data set is handled in a similar way, the date predicted will be needed as label meter
Calculate day, extract corresponding feature, as the input of algorithm, obtain prediction result to the end, after sequence be user most
It is possible that generating user-article set of buying behavior.Businessman can be according to the user that can be carried out purchase predicted and object
Product, are dispatched and services accordingly, to improve the efficiency and satisfaction of user's purchase.
Embodiment described above is only the preferred embodiments of the invention, and but not intended to limit the scope of the present invention, therefore
All shapes according to the present invention change made by principle, should all be included within the scope of protection of the present invention.
Claims (6)
1. a kind of user's buying behavior prediction technique based on integrated neural network, which is characterized in that include the following steps：
1) feature extraction and sampling are carried out to user behavior historical record, user, article characteristics and user can be summarized by therefrom obtaining
Buy the sample set T1 of label；
2) classifier C1 is formed using Boosting integrated approach, classification processing is carried out to sample set T1 and feature is integrated, is obtained
To new sample set T2；
3) basic structure for constructing neural network carries out heuristic search using parameter of the genetic algorithm to neural network, obtains
Optimal neural network, and the optimal Artificial neural network ensemble that will be obtained form integrated neural network classifier C2；
4) classification processing is carried out to sample set T2 using classifier C2, wherein each neural network can handle sample set
In a part, and carry out result fusion before the output of classifier C2 and feature is integrated, finally obtain new sample set T3；
5) classifier C3 is formed using Bagging integrated approach, classification processing is carried out to sample set T3, obtaining user can generate
The item lists of buying behavior, as the prediction result for user's buying behavior.
2. a kind of user's buying behavior prediction technique based on integrated neural network according to claim 1, feature exist
In：In step 1), the user behavior historical record includes mark, the mark of article, the classification of user behavior of user
And the geographical location information of the time and interaction of user and article interaction；By way of statisticalling analyze and manually inferring, from
In extract and can embody user, article feature, the feature of prediction user behavior tendency, using positive and negative sample aligned sample and different
Constant value processing, obtains the characteristic of original sample set；Whether user finally buys the label portion as original sample set
Point；Characteristic and label segment collectively form sample set T1.
3. a kind of user's buying behavior prediction technique based on integrated neural network according to claim 1, feature exist
In：In step 2), decision tree progress Boosting is integrated, the classifier C1 of formation is that a gradient promotes classifier；It is right
In the sample set T1 extracted from user behavior historical record, characteristic X1, label segment y1；Using X1 as point
The input of class device C1, obtains prediction result y2, and one bivector of each behavior respectively represents prediction result and do not buy
Probability and prediction result are the probability of purchase；It is incorporated into original sample characteristics X1, is formed new using y2 as new feature
The new sample set that sample characteristics X2, sample characteristics X2 and label segment y1 are formed is known as T2.
4. a kind of user's buying behavior prediction technique based on integrated neural network according to claim 1, feature exist
In：It in step 3), for constructed neural network, is made of, inputs 1 input layer, 1 output layer and 3 hidden layers
The neuron number of layer is the Characteristic Number of input data, and the neuron number of output layer is 2, represents two-dimensional prediction result,
Hidden layer then by the way of connecting entirely；The hidden layer number of plies of neural network, each layer of neuron number, neural network repeatedly
Generation number and each iteration size parameter all carry out heuristic search by genetic algorithm and obtain, by the iteration of genetic algorithm come
To the optimal solution of parameters, and so as to form optimal neural network.
5. a kind of user's buying behavior prediction technique based on integrated neural network according to claim 1, feature exist
In：In step 4), for new sample set T2, after it is normalized, its equal proportion is divided into n parts, point
Also known as be T2_1, T2_2, T2_3 ..., T2_n；For a neural network in classifier C2, its training set is T2 collection
N-1 part in conjunction, test set are remaining 1 part；Neural network model after n training respectively carries out respective test set pre-
Survey, obtain n prediction result, referred to as y3_1, y3_2, y3_3 ..., y3_n, this n prediction result is merged into new prediction
As a result y3；It is incorporated to y3 as new feature in sample characteristics X2, forms new sample characteristics X3；Sample characteristics X3 and label portion
The new sample set for dividing y1 to be formed is known as T3.
6. a kind of user's buying behavior prediction technique based on integrated neural network according to claim 1, feature exist
In：In step 5), decision tree progress Bagging is integrated, the classifier C3 of formation is a random forest grader；For
New sample set T3 obtains prediction result y4, each behavior one using its characteristic X3 as the input of classifier C3
Bivector, respectively represent prediction result be the probability that do not buy and prediction result be purchase probability；To prediction result according to
After the probability of purchase is ranked up, the item lists of buying behavior can be generated by just obtaining user, as final output result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810642096.XA CN108921602B (en) | 2018-06-21 | 2018-06-21 | User purchasing behavior prediction method based on integrated neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810642096.XA CN108921602B (en) | 2018-06-21 | 2018-06-21 | User purchasing behavior prediction method based on integrated neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108921602A true CN108921602A (en) | 2018-11-30 |
| CN108921602B CN108921602B (en) | 2021-12-21 |
Family
ID=64419786
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810642096.XA Active CN108921602B (en) | 2018-06-21 | 2018-06-21 | User purchasing behavior prediction method based on integrated neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108921602B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109784748A (en) * | 2019-01-25 | 2019-05-21 | 广东电网有限责任公司 | User power utilization behavior discrimination method and device under a kind of market and competitive mechanisms |
| CN110033309A (en) * | 2019-03-08 | 2019-07-19 | 平安科技（深圳）有限公司 | Method, apparatus and terminal device based on historical data prediction result |
| CN110415085A (en) * | 2019-07-31 | 2019-11-05 | 孙海龙 | A kind of commodity screening, methods of exhibiting and device based on geographical location information |
| CN111709766A (en) * | 2020-04-14 | 2020-09-25 | 中国农业银行股份有限公司 | User behavior prediction method and device, storage medium and electronic equipment |
| CN112184315A (en) * | 2020-09-29 | 2021-01-05 | 深圳市尊信网络科技有限公司 | Method, device, equipment and storage medium for identifying abnormal lottery purchasing behavior |
| CN112232388A (en) * | 2020-09-29 | 2021-01-15 | 南京财经大学 | ELM-RFE-based shopping intention key factor identification method |
| CN112950239A (en) * | 2019-11-26 | 2021-06-11 | 多点(深圳)数字科技有限公司 | Method, apparatus, device and computer readable medium for generating user information |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080306830A1 (en) * | 2007-06-07 | 2008-12-11 | Cliquality, Llc | System for rating quality of online visitors |
| CN103324690A (en) * | 2013-06-03 | 2013-09-25 | 焦点科技股份有限公司 | Mixed recommendation method based on factorization condition limitation Boltzmann machine |
| CN105868334A (en) * | 2016-03-28 | 2016-08-17 | 云南财经大学 | Personalized film recommendation method and system based on feature augmentation |
| CN106897744A (en) * | 2017-02-27 | 2017-06-27 | 郑州云海信息技术有限公司 | A kind of self adaptation sets the method and system of depth confidence network parameter |
-
2018
- 2018-06-21 CN CN201810642096.XA patent/CN108921602B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080306830A1 (en) * | 2007-06-07 | 2008-12-11 | Cliquality, Llc | System for rating quality of online visitors |
| CN103324690A (en) * | 2013-06-03 | 2013-09-25 | 焦点科技股份有限公司 | Mixed recommendation method based on factorization condition limitation Boltzmann machine |
| CN105868334A (en) * | 2016-03-28 | 2016-08-17 | 云南财经大学 | Personalized film recommendation method and system based on feature augmentation |
| CN106897744A (en) * | 2017-02-27 | 2017-06-27 | 郑州云海信息技术有限公司 | A kind of self adaptation sets the method and system of depth confidence network parameter |
Non-Patent Citations (5)
| Title |
|---|
| SHOUXIN SUN等: "The Research of the Network Security Situation Prediction Mechanism Based on the Complex Network", 《2015 INTERNATIONAL CONFERENCE ON COMPUTATIONAL INTELLIGENCE AND COMMUNICATION NETWORKS (CICN)》 * |
| 任谢楠: "基于遗传算法的BP神经网络的优化研究及MATLAB仿真", 《中国优秀硕士学位论文全文数据库（电子期刊）》 * |
| 何博威: "基于用户消费习惯的推荐算法研究", 《中国优秀硕士学位论文全文数据库（电子期刊）》 * |
| 孟晓龙: "基于机器学习的推荐技术研究", 《中国优秀硕士学位论文全文数据库（电子期刊）》 * |
| 马倩: "基于机器学习的电子商务平台重复购买客户预测", 《中国优秀硕士学位论文全文数据库（电子期刊）》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109784748A (en) * | 2019-01-25 | 2019-05-21 | 广东电网有限责任公司 | User power utilization behavior discrimination method and device under a kind of market and competitive mechanisms |
| CN110033309A (en) * | 2019-03-08 | 2019-07-19 | 平安科技（深圳）有限公司 | Method, apparatus and terminal device based on historical data prediction result |
| CN110033309B (en) * | 2019-03-08 | 2022-05-20 | 平安科技（深圳）有限公司 | Method and device for predicting result based on historical data and terminal equipment |
| CN110415085A (en) * | 2019-07-31 | 2019-11-05 | 孙海龙 | A kind of commodity screening, methods of exhibiting and device based on geographical location information |
| CN112950239A (en) * | 2019-11-26 | 2021-06-11 | 多点(深圳)数字科技有限公司 | Method, apparatus, device and computer readable medium for generating user information |
| CN111709766A (en) * | 2020-04-14 | 2020-09-25 | 中国农业银行股份有限公司 | User behavior prediction method and device, storage medium and electronic equipment |
| CN111709766B (en) * | 2020-04-14 | 2023-08-18 | 中国农业银行股份有限公司 | User behavior prediction method and device, storage medium and electronic equipment |
| CN112184315A (en) * | 2020-09-29 | 2021-01-05 | 深圳市尊信网络科技有限公司 | Method, device, equipment and storage medium for identifying abnormal lottery purchasing behavior |
| CN112232388A (en) * | 2020-09-29 | 2021-01-15 | 南京财经大学 | ELM-RFE-based shopping intention key factor identification method |
| CN112232388B (en) * | 2020-09-29 | 2024-02-13 | 南京财经大学 | Shopping intention key factor identification method based on ELM-RFE |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108921602B (en) | 2021-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108921602A (en) | A kind of user's buying behavior prediction technique based on integrated neural network | |
| Ngai et al. | Application of data mining techniques in customer relationship management: A literature review and classification | |
| Chattopadhyay et al. | Application of artificial neural network in market segmentation: A review on recent trends | |
| Tamilselvi et al. | An overview of data mining techniques and applications | |
| Oliveira | Analytical customer relationship management in retailing supported by data mining techniques | |
| Gangurde et al. | Building prediction model using market basket analysis | |
| Fatoni et al. | Online store product recommendation system uses apriori method | |
| Allegue et al. | RFMC: a spending-category segmentation | |
| Chiang | To establish online shoppers’ markets and rules for dynamic CRM systems: an empirical case study in Taiwan | |
| Mandala et al. | Recognition of E-Commerce through Big Data Classification and Data Mining Techniques Involving Artificial Intelligence | |
| Mei et al. | Research on e-commerce coupon user behavior prediction technology based on decision tree algorithm | |
| CN114723491A (en) | Accurate marketing method and system based on user portrait and data mining | |
| Mallik et al. | A comprehensive survey on sales forecasting models using machine learning algorithms | |
| Butler et al. | Customer behaviour classification using simulated transactional data | |
| Arivazhagan et al. | Pattern Discovery and Analysis of Customer Buying Behavior Using Association Rules Mining Algorithm in E-Commerce | |
| Javaid et al. | Explainable artificial intelligence solution for online retail | |
| CN114092123A (en) | Satisfaction intelligent analysis system | |
| Hervert-Escobar et al. | Optimal pricing model based on reduction dimension: A case of study for convenience stores | |
| Purnamasari et al. | Consumer Behavior Analysis of Leathercraft Small and Medium-Sized Enterprises (SME) Using Market Basket Analysis and Clustering Algorithms | |
| Idowu et al. | Customer Segmentation Based on RFM Model Using K-Means, Hierarchical and Fuzzy C-Means Clustering Algorithms | |
| Rajesh et al. | Customer Behavior Prediction for E-Commerce Sites Using Machine Learning Techniques: An Investigation | |
| Huang | AI-based Repeat Buyers Prediction System using Deep Learning | |
| Jatain | Performance Optimization of an Enterprise using Data-Driven Strategy | |
| Malik et al. | Applying data mining for clustering shoppers based on store loyalty | |
| Metilda | A Study on Customer Segmentation Using K-Means Clustering for Online Shoppers |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |