CN108921602B - User purchasing behavior prediction method based on integrated neural network - Google Patents
User purchasing behavior prediction method based on integrated neural network Download PDFInfo
- Publication number
- CN108921602B CN108921602B CN201810642096.XA CN201810642096A CN108921602B CN 108921602 B CN108921602 B CN 108921602B CN 201810642096 A CN201810642096 A CN 201810642096A CN 108921602 B CN108921602 B CN 108921602B
- Authority
- CN
- China
- Prior art keywords
- neural network
- user
- classifier
- sample set
- feature
- 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.)
- Expired - Fee Related
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 62
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 24
- 230000010354 integration Effects 0.000 claims abstract description 19
- 230000002068 genetic effect Effects 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 8
- 238000005070 sampling Methods 0.000 claims abstract description 6
- 210000002569 neuron Anatomy 0.000 claims description 9
- 238000003066 decision tree Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 5
- 230000003993 interaction Effects 0.000 claims description 4
- 238000007637 random forest analysis Methods 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000007619 statistical method Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 7
- 238000000605 extraction Methods 0.000 abstract 1
- 230000006399 behavior Effects 0.000 description 30
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
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]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Development Economics (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Marketing (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- General Business, Economics & Management (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Genetics & Genomics (AREA)
- Game Theory and Decision Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a user purchasing behavior prediction method based on an integrated neural network, which comprises the following steps: 1) carrying out feature extraction and sampling on the user behavior history record to obtain a sample set T1; 2) forming a classifier C1 by using a Boosting integration method, and carrying out classification processing and feature integration on the sample set T1 to obtain a new sample set T2; 3) constructing a basic structure of the neural network, and carrying out heuristic search on parameters of the neural network by using a genetic algorithm to form an integrated neural network classifier C2; 4) classifying the sample set T2 by using a classifier C2 to obtain a new sample set T3; 5) and forming a classifier C3 by using a Bagging integration method, and classifying the sample set T3 to obtain an item list of the purchasing behavior of the user, wherein the item list is used as a prediction result of the purchasing behavior of the user. The invention solves the problems of poor classification effect, poor generalization, low efficiency under the condition of big data and the like of the traditional method.
Description
Technical Field
The invention relates to the technical field of electronic commerce, in particular to a user purchasing behavior prediction method based on an integrated neural network.
Background
With the advent of the big data era and the popularity of online shopping of users, in the big data background, the behaviors of improving the exposure rate of commodities and increasing the purchase amount of users by recommending commodities in which the users are interested through an excellent recommendation algorithm have become one of the basic functions of an electronic commerce system. And the accurate prediction of the purchasing behavior of the user is the final target of the recommendation algorithm. If the merchant can master the purchase intention of the consumer, the merchant can reasonably arrange the inventory of the commodity, and can also construct an accurate user image and feed the accurate user image back to the personnel such as market, marketing and the like to carry out targeted commodity sales. Therefore, the method has important theoretical and practical significance for predicting the purchasing behavior of the user.
Currently, in the field of e-commerce, the prediction methods of user purchasing behavior are mainly divided into three types, the first type is to use manually defined rules, for example, if a certain item is put into a shopping cart on a certain day but is not purchased, then it is likely that the purchase will be performed on the next day; methods such as statistical analysis of users using questionnaires and the like; the second is to use the traditional recommendation algorithm, such as collaborative filtering, content-based recommendation, etc., to make predictions, and these methods have all proved to have certain effects in the recommendation field; and thirdly, regarding the purchasing prediction as a two-classification problem, and predicting the purchasing behavior of the user by training a model of the user behavior by using a typical machine learning classifier such as a support vector machine, a decision tree and the like. All three methods can predict the purchasing behavior of the user to a certain extent, but all have some disadvantages. The manual method greatly depends on the labor of human beings, and has narrow analysis range and low accuracy; the traditional recommendation algorithm can only recommend commodities which a user may be interested in, scores the degree of the user interest, and still relies on manual screening and evaluation as the result cannot be obtained through the algorithm for predicting whether the user will purchase or not; the traditional classifier method is low in accuracy of prediction results and poor in model generalization, and the three methods have the problems of low efficiency, low accuracy and the like under the condition of big data.
The invention provides a user purchasing behavior prediction method based on an integrated neural network, which is characterized in that after characteristics of users and articles are extracted, the purchasing behavior of the users is predicted through various integration methods and an optimized neural network, the problems of poor classification effect, poor generalization and large dependence on manpower in the traditional method are solved by utilizing the advantages of nonlinearity, strong self-adaptability and the like of the neural network and combining an integrated learning method, and the prediction efficiency and the accuracy of a prediction result are improved.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, provides a user purchasing behavior prediction method based on an integrated neural network, solves the problems of poor classification effect, poor generalization performance, low efficiency under the big data scene and the like of the traditional method, and improves the efficiency and generalization performance of a prediction model and the accuracy under the big data scene.
In order to achieve the purpose, the technical scheme provided by the invention is as follows: a user purchasing behavior prediction method based on an integrated neural network comprises the following steps:
1) extracting and sampling characteristics of the user behavior history record to obtain a sample set T1 capable of summarizing the characteristics of the user and the article and purchasing labels of the user;
2) forming a classifier C1 by using a Boosting integration method, and carrying out classification processing and feature integration on the sample set T1 to obtain a new sample set T2;
3) constructing a basic structure of the neural network, carrying out heuristic search on parameters of the neural network by using a genetic algorithm to obtain an optimal neural network, and integrating the obtained optimal neural network to form an integrated neural network classifier C2;
4) classifying the sample set T2 by using a classifier C2, wherein each neural network processes a part of the sample set and performs result fusion and feature integration before the output of the classifier C2 to finally obtain a new sample set T3;
5) and forming a classifier C3 by using a Bagging integration method, and classifying the sample set T3 to obtain an item list of the purchasing behavior of the user, wherein the item list is used as a prediction result of the purchasing behavior of the user.
In step 1), the user behavior history record contains the user identification, the article identification, the user behavior category, the time of interaction between the user and the article and the interaction geographical location information; extracting features which can reflect the characteristics of users and articles and predict the behavior tendency of the users from the characteristics by means of statistical analysis and manual inference, and then carrying out balanced sampling and abnormal value processing on positive and negative samples to obtain a feature part of an original sample set; whether the user ultimately purchases the label portion as the original sample set; the feature portion and the tag portion together constitute a sample set T1.
In the step 2), Boosting integration is carried out on the decision tree, and a formed classifier C1 is a gradient Boosting classifier; for a sample set T1 extracted from the user behavior history record, the characteristic part is X1, and the tag part is y 1; taking X1 as the input of a classifier C1 to obtain a prediction result y2, wherein each line of the prediction result y is a two-dimensional vector and respectively represents the probability that the prediction result is not purchased and the probability that the prediction result is purchased; incorporating y2 as a new feature into the original sample feature X1, forming a new sample feature X2, the new set of samples formed by sample feature X2 and label portion y1 is referred to as T2.
In step 3), the constructed neural network consists of 1 input layer, 1 output layer and 3 hidden layers, the number of neurons of the input layer is the characteristic number of input data, the number of neurons of the output layer is 2, the two-dimensional prediction result is represented, and the hidden layers adopt a full-connection mode; the number of hidden layers of the neural network, the number of neurons in each layer, the iteration times of the neural network and the parameters of each iteration are obtained by heuristic search through a genetic algorithm, the optimal solution of each parameter is obtained through iteration of the genetic algorithm, and the optimal neural network is formed.
In step 4), after normalization processing is performed on the new sample set T2, the new sample set T2 is divided into n parts in equal proportion, which are respectively called T2_1, T2_2, T2_3, … … and T2_ n; for a neural network in the classifier C2, the training set is n-1 in the T2 set, and the test set is the rest 1; the n trained neural network models respectively predict the respective test sets to obtain n prediction results, namely y3_1, y3_2, y3_3, … … and y3_ n, and the n prediction results are combined into a new prediction result y 3; incorporating y3 as a new feature into sample feature X2, forming a new sample feature X3; the new set of samples formed by sample feature X3 and tag portion y1 is referred to as T3.
In the step 5), Bagging integration is carried out on the decision tree, and the formed classifier C3 is a random forest classifier; for a new sample set T3, taking the characteristic part X3 as the input of a classifier C3, obtaining a prediction result y4, each line of which is a two-dimensional vector and respectively represents the probability that the prediction result is not purchased and the probability that the prediction result is purchased; and after the prediction results are sorted according to the probability of purchase, an item list of the purchase behavior generated by the user is obtained and used as a final output result.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. according to the method, the computer is enabled to carry out future purchase prediction according to the historical purchase behaviors of the user through the training model, manual excessive intervention is not needed in model generation and subsequent use, manual labor force is liberated, and prediction efficiency is improved.
2. The method has excellent performance under the situation of big data, and compared with the traditional recommendation algorithm such as collaborative filtering and the classification algorithm which have low efficiency, low accuracy and the like when facing the big data, the method can well process the situation of the big data, and can smoothly process a large amount of historical data and user data needing to be predicted.
3. The invention utilizes the characteristics of neural network nonlinearity, strong self-adaptability and the like, and integrates the advantages of learning fusion characteristics and common decision, and compared with the traditional algorithm and manual processing, the accuracy rate is greatly improved.
4. The method adjusts the parameters of the neural network by using a genetic algorithm and a heuristic search method, optimizes the part which has the greatest influence on the accuracy rate in the construction of the neural network, and has better generalization and higher accuracy rate compared with the neural network which is not optimized.
5. The invention has wide application space in the field of electronic commerce, is easy to operate and customize and upgrade individually, and has wide application prospect.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
FIG. 2 is a flow chart of a genetic algorithm.
Fig. 3 is a schematic diagram of the structure of a neural network.
FIG. 4 is a schematic diagram of an integrated neural network.
Detailed Description
In order to more specifically describe the present invention, the following detailed description is provided for the technical solution of the present invention with reference to the accompanying drawings and the specific embodiments.
In the embodiment, a user purchasing behavior prediction method based on an integrated neural network is applied to the mobile recommendation data set. As shown in fig. 1, the method for predicting the purchasing behavior of the user mainly comprises the following steps:
firstly, extracting and sampling features of a user behavior historical record, and obtaining a sample set capable of summarizing the features of users, articles and the like; for the present embodiment, since the data set is derived from the real data in ali, a preprocessing operation is required to remove abnormal data in the data set, such as a user who never purchases a suspected robot, and purchase data that is greatly increased in a sales promotion day. And then, carrying out feature calculation on the preprocessed data, wherein the features of the user mainly comprise the times of clicking, purchasing and other behaviors of the user, and the times of clicking and purchasing the article and the sum and average values of the times after the clicking and purchasing of the article are combined. For the embodiment, a tag calculation day is defined, that is, a date on which whether to purchase is to be predicted, and the features are derived from data of days before the tag day, where a positive sample is a user-item pair having a purchase behavior on the tag day, and a negative sample is a user-item pair having no purchase behavior. The extracted features are combined with the labels into a sample set T1.
Then, a classifier is formed by using a Boosting integration method, and prediction is performed on a sample set T1 extracted from the user behavior history record by using the integration classifier. In this embodiment, a gradient lifting tree is used as a classifier of a Boosting integrated decision tree, a log-likelihood function is used as a loss function, the maximum depth of the tree is set to be 4, and the learning rate is 0.05, so that the extracted sample set is predicted to obtain a prediction result y 2. The prediction result y2 is combined into the feature part of the sample set T1 to form a new sample feature, and combined with the original tag into a new sample set T2.
Then, for the new sample set T2, after normalizing and forming one-hot encoding, it is used as the input of the classifier C2 formed by integrating a plurality of neural networks; the hyper-parameter setting of the neural network adopts a genetic algorithm to carry out heuristic search. FIG. 2 shows a flow chart of one iteration of the genetic algorithm of the present invention, first generating a population comprising a list of hundreds of randomly generated hyper-parameters, called population's individuals. The list includes hyper-parameters of the neural network, such as the number of hidden layers, the number of hidden layer neurons, the number of neural network iterations, the size of neural network iteration blocks, and the like.
Firstly, a neural network is constructed according to the hyper-parameters of each individual, then, the neural network is trained and classified in a sample set, the classification prediction accuracy of the neural network is obtained, and the evaluation index is the cross entropy between the prediction result and the sample label. By carrying out classification prediction on each individual of the population, a best accuracy rate and a corresponding best hyper-parameter list can be obtained. Next, we start to perform iteration of the genetic algorithm, and the specific process is as follows: two individuals are selected from the population and are crossed and mutated according to a certain probability. The crossing is that the parameters between the two individuals are fused with each other to form a new hyper-parameter list; mutation is the probability of changing the value of random parameters in the list. Then, a neural network is constructed, classified and predicted, and the effect of the changed individuals is evaluated by the same method. Comparing the effect of the changed individual with the recorded best effect, and replacing the original individual with the changed individual if the changed individual is better. After thousands of iterations, the algorithm will converge and give a best list of hyper-parameter settings. We will set the hyper-parameters of the neural network in the algorithm with this list of hyper-parameters. Fig. 3 is a schematic structural diagram of a neural network. Each neural network consists of 1 input layer, 1 output layer and a plurality of hidden layers, the number of neurons of the input layer is the dimension of X2 in a sample set T2, the number of neurons of the output layer is 2, a two-dimensional prediction result is represented, the hidden layers are in a full-connection mode, and hyper-parameters are set according to an optimal hyper-parameter list obtained by a genetic algorithm; the activation function of the neural network adopts a ReLu function, and the formula is as follows:
f(x)＝max(0,x)
the Softmax function is a function of the output layer, and the formula is:
the output function of Softmax can predict the probability of the two-dimensional output, which can be regarded as the probability of the occurrence of each tag, i.e. the probability of the purchase and the non-purchase of the user.
Fig. 4 shows a schematic diagram of an integrated neural network. In this embodiment, we divide the equal proportion into 5 parts, which are called T2_1, T2_2, T2_3, T2_4, T2_ 5; for one of the forward neural networks, its training set is 4 in the T2 set, and the test set is the remaining 1, for example, for the forward neural network 1, T2_1 is its test set, and T2_2, T2_3, T2_4, T2_5 is its training set; the 5 trained neural network models respectively predict the respective test sets to obtain 5 prediction results, namely y3_1, y3_2, y3_3, y3_4 and y3_5, and the 5 prediction results are combined into a new prediction result y 3;
y3＝y3_1∪y3_2∪y3_3∪y3_4∪y3_5
incorporating y3 as a new feature into sample feature X2, forming a new sample feature X3;
X3＝{X2,y3}
the new set of samples formed by sample feature X3 and tag portion y1 is referred to as T3.
And finally, forming a classifier by using a Bagging integration method. In this embodiment, we use a random forest algorithm as a Bagging integrated decision tree classifier. For the new sample set T3, its feature portion X3 is taken as input to classifier C3. Selecting features with large gain values of the kini indexes for further division in the random forest; after multiple times of replaced randomly extracted samples and feature training, determining the final classification of the sample data according to the voting result of the decision tree in the forest. After classification, a prediction result y4 is obtained, each line of the prediction result y4 is a two-dimensional vector and represents the probability that the prediction result is not purchased and the probability that the prediction result is purchased respectively; after the probability that the predicted result is the purchase is ranked, a list of the most likely purchased articles of the user is obtained, and the K results with the highest probability, namely the Top-K purchased articles, are used as the final output result.
For the evaluation of the output results, we used the F1 value as the evaluation criterion. The F1 value is calculated according to the precision rate and the recall rate. Their respective formulas are as follows:
wherein P is the accuracy rate, R is the recall rate, predictionSet is the final prediction result set, and preferenceSet is the result set purchased by the actual user. By continuously evaluating the results of the classifier, the parameters and settings of the classifier can be adjusted to obtain the best classification model.
For prediction, the data set is processed in a similar manner, the date needing prediction is used as the label calculation date, corresponding features are extracted and used as the input of an algorithm to obtain the final prediction result, and the final prediction result is the user-item set which is most likely to generate purchasing behaviors by the user after sequencing. The merchant can carry out corresponding scheduling and service work according to the predicted users and articles which are possibly purchased, thereby improving the purchasing efficiency and satisfaction of the users.
The above-mentioned embodiments are merely preferred embodiments of the present invention, and the scope of the present invention is not limited thereto, so that the changes in the shape and principle of the present invention should be covered within the protection scope of the present invention.
Claims (4)
1. A user purchasing behavior prediction method based on an integrated neural network is characterized by comprising the following steps:
1) extracting and sampling characteristics of the user behavior history record to obtain a sample set T1 capable of summarizing the characteristics of the user and the article and purchasing labels of the user;
2) forming a classifier C1 by using a Boosting integration method, and carrying out classification processing and feature integration on the sample set T1 to obtain a new sample set T2;
3) constructing a basic structure of the neural network, carrying out heuristic search on parameters of the neural network by using a genetic algorithm to obtain an optimal neural network, and integrating the obtained optimal neural network to form an integrated neural network classifier C2;
for the constructed neural network, the neural network consists of 1 input layer, 1 output layer and 3 hidden layers, the number of neurons of the input layer is the characteristic number of input data, the number of neurons of the output layer is 2, the two-dimensional prediction result is represented, and the hidden layers adopt a full-connection mode; the number of hidden layers of the neural network, the number of neurons in each layer, the iteration times of the neural network and the parameters of each iteration are obtained by heuristic search through a genetic algorithm, the optimal solution of each parameter is obtained through the iteration of the genetic algorithm, and the optimal neural network is formed;
4) classifying the sample set T2 by using a classifier C2, wherein each neural network processes a part of the sample set and performs result fusion and feature integration before the output of the classifier C2 to finally obtain a new sample set T3;
after normalization processing is performed on the new sample set T2, the new sample set T2 is divided into n parts in equal proportion, namely T2_1, T2_2, T2_3, … … and T2_ n; for a neural network in the classifier C2, the training set is n-1 in the T2 set, and the test set is the rest 1; the n trained neural network models respectively predict the respective test sets to obtain n prediction results, namely y3_1, y3_2, y3_3, … … and y3_ n, and the n prediction results are combined into a new prediction result y 3; incorporating y3 as a new feature into sample feature X2, forming a new sample feature X3; the new set of samples formed by sample feature X3 and tag portion y1 is referred to as T3;
5) and forming a classifier C3 by using a Bagging integration method, and classifying the sample set T3 to obtain an item list of the purchasing behavior of the user, wherein the item list is used as a prediction result of the purchasing behavior of the user.
2. The method for predicting the purchasing behavior of the user based on the integrated neural network as claimed in claim 1, wherein: in step 1), the user behavior history record contains the user identification, the article identification, the user behavior category, the time of interaction between the user and the article and the interaction geographical location information; extracting features which can reflect the characteristics of users and articles and predict the behavior tendency of the users from the characteristics by means of statistical analysis and manual inference, and then carrying out balanced sampling and abnormal value processing on positive and negative samples to obtain a feature part of an original sample set; whether the user ultimately purchases the label portion as the original sample set; the feature portion and the tag portion together constitute a sample set T1.
3. The method for predicting the purchasing behavior of the user based on the integrated neural network as claimed in claim 1, wherein: in the step 2), Boosting integration is carried out on the decision tree, and a formed classifier C1 is a gradient Boosting classifier; for a sample set T1 extracted from the user behavior history record, the characteristic part is X1, and the tag part is y 1; taking X1 as the input of a classifier C1 to obtain a prediction result y2, wherein each line of the prediction result y is a two-dimensional vector and respectively represents the probability that the prediction result is not purchased and the probability that the prediction result is purchased; incorporating y2 as a new feature into the original sample feature X1, forming a new sample feature X2, the new set of samples formed by sample feature X2 and label portion y1 is referred to as T2.
4. The method for predicting the purchasing behavior of the user based on the integrated neural network as claimed in claim 1, wherein: in the step 5), Bagging integration is carried out on the decision tree, and the formed classifier C3 is a random forest classifier; for a new sample set T3, taking the characteristic part X3 as the input of a classifier C3, obtaining a prediction result y4, each line of which is a two-dimensional vector and respectively represents the probability that the prediction result is not purchased and the probability that the prediction result is purchased; and after the prediction results are sorted according to the probability of purchase, an item list of the purchase behavior generated by the user is obtained and used as a 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 CN108921602A (en) | 2018-11-30 |
| CN108921602B true CN108921602B (en) | 2021-12-21 |
Family
ID=64419786
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810642096.XA Expired - Fee Related 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) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109784748B (en) * | 2019-01-25 | 2022-06-21 | 广东电网有限责任公司 | User electricity consumption behavior identification method and device under market competition mechanism |
| 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 |
| 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 |
| CN112232388B (en) * | 2020-09-29 | 2024-02-13 | 南京财经大学 | Shopping intention key factor identification method based on ELM-RFE |
| 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 |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
Family Cites Families (1)
| 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 |
-
2018
- 2018-06-21 CN CN201810642096.XA patent/CN108921602B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 (3)
| Title |
|---|
| 基于机器学习的推荐技术研究;孟晓龙;《中国优秀硕士学位论文全文数据库（电子期刊）》;20161215;第3.2、4.1、4.2、5.1节 * |
| 基于用户消费习惯的推荐算法研究;何博威;《中国优秀硕士学位论文全文数据库（电子期刊）》;20180415;全文 * |
| 基于遗传算法的BP神经网络的优化研究及MATLAB仿真;任谢楠;《中国优秀硕士学位论文全文数据库（电子期刊）》;20141115;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108921602A (en) | 2018-11-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108921602B (en) | User purchasing behavior prediction method based on integrated neural network | |
| CN110222272B (en) | Potential customer mining and recommending method | |
| CN110956497B (en) | Method for predicting repeated purchasing behavior of user of electronic commerce platform | |
| Ngai et al. | Application of data mining techniques in customer relationship management: A literature review and classification | |
| Solairaj et al. | Enhanced Elman spike neural network based sentiment analysis of online product recommendation | |
| Seymen et al. | Customer churn prediction using deep learning | |
| Mandala et al. | Recognition of E-Commerce through Big Data Classification and Data Mining Techniques Involving Artificial Intelligence | |
| Hwang et al. | Using genetic algorithms for personalized recommendation | |
| Mei et al. | Research on e-commerce coupon user behavior prediction technology based on decision tree algorithm | |
| Valero-Fernandez et al. | Towards accurate predictions of customer purchasing patterns | |
| Yin et al. | A network-enhanced prediction method for automobile purchase classification using deep learning | |
| CN114238758B (en) | User portrait prediction method based on multi-source cross-border data fusion | |
| Leventhal | Predictive Analytics for Marketers: Using Data Mining for Business Advantage | |
| Chashmi et al. | Predicting customer turnover using recursive neural networks | |
| Javaid et al. | Explainable artificial intelligence solution for online retail | |
| CN114997959A (en) | Electronic intelligent product marketing recommendation method | |
| Baderiya et al. | Customer buying Prediction Using Machine-Learning Techniques: A Survey | |
| Syadzali et al. | Business intelligence using the k-nearest neighbor algorithm to analyze customer behavior in online crowdfunding systems | |
| Huang | AI-based repeat buyers prediction system using deep learning | |
| Algawiaz et al. | Predicting a User’s Purchase Intention Using AdaBoost | |
| Pribil et al. | A churn analysis using data mining techniques: Case of electricity distribution company | |
| Lee et al. | A hybrid machine learning approach for customer loyalty prediction | |
| Pandey et al. | Predicting Customer Satisfaction in Brazil E-commerce: A Comparative Study of Machine Learning Techniques | |
| Feng et al. | A random forest algorithm based customer demand forecasting model for sports enterprises in the real economy | |
| Sharma | Identifying Factors Contributing to Lead Conversion Using Machine Learning to Gain Business Insights |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20211221 |