CN109787821A - A kind of Large-scale Mobile customer traffic consumption intelligent Forecasting - Google Patents
A kind of Large-scale Mobile customer traffic consumption intelligent Forecasting Download PDFInfo
- Publication number
- CN109787821A CN109787821A CN201910006654.8A CN201910006654A CN109787821A CN 109787821 A CN109787821 A CN 109787821A CN 201910006654 A CN201910006654 A CN 201910006654A CN 109787821 A CN109787821 A CN 109787821A
- Authority
- CN
- China
- Prior art keywords
- predictor
- value
- layer
- fallout predictor
- regressive
- 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
- 230000001373 regressive effect Effects 0.000 claims abstract description 36
- 238000012549 training Methods 0.000 claims abstract description 29
- 238000012800 visualization Methods 0.000 claims abstract description 6
- 230000001537 neural effect Effects 0.000 claims description 34
- 238000000034 method Methods 0.000 claims description 27
- 230000008569 process Effects 0.000 claims description 19
- 230000004913 activation Effects 0.000 claims description 18
- 239000013598 vector Substances 0.000 claims description 17
- 238000013528 artificial neural network Methods 0.000 claims description 16
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 6
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 6
- 238000003066 decision tree Methods 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 230000009471 action Effects 0.000 claims description 3
- 238000013079 data visualisation Methods 0.000 claims description 2
- 230000004907 flux Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 12
- 210000005036 nerve Anatomy 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 2
- 238000004321 preservation Methods 0.000 description 2
- 230000003542 behavioural effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of Large-scale Mobile customer traffics to consume intelligent Forecasting, comprising steps of mobile subscriber attribute feature and consumer behavior data 1) are collected, visualization, and pre-processed；2) structural classification fallout predictor and regressive predictor, are trained, and obtain the prediction model of two different scales；3) joint classification fallout predictor and regressive predictor are trainable linear combination, carry out second stage training, obtain joint forecast model；4) joint forecast model is used, the flow consumption value of user next month is predicted according to the attributive character of mobile subscriber and consumer behavior.The present invention joins together classification fallout predictor with regressive predictor, the training of dual-stage is carried out on Large-scale Mobile user data, make obtained joint flux prediction model that there is higher accuracy and robustness, to provide thinking of more precisely effectively marketing for mobile service popularization.
Description
Technical field
The present invention relates to the technical fields of data mining, and it is pre- to refer in particular to a kind of Large-scale Mobile customer traffic consumption intelligence
Survey method.
Background technique
As the universal of 4G mobile communication technology, mobile Internet flourish, the life style of user gradually occurs
Variation.The marketing emphasis of telecom operators is gradually from traditional voice service steering flow business.Accurately predict user's future
Flow consumption, operator can be made more efficiently to promote flow business, stimulate the consumption of user, improve flow battalion
It receives.
Traditional method for predicting only predicts the flow consumption value of user by homing method, is easy by data
The interference of noise, accuracy and robustness are insufficient.The present invention utilizes flow after discretization to consume the value class two that field has
As property, building classification fallout predictor and the united prediction model of regressive predictor, in the mobile subscriber attribute feature of magnanimity and consumption
The implicit rule of mobile subscriber's flow consumption is excavated on behavioral data, to predict the flow consumption feelings in mobile subscriber's future
Condition promotes flow package with then customizing to it, achievees the purpose that precision marketing, improves flow business revenue.
Summary of the invention
The purpose of the present invention is to overcome the shortcomings of the existing technology and deficiency, proposes a kind of Large-scale Mobile customer traffic
Intelligent Forecasting is consumed, classification fallout predictor is joined together with regressive predictor, is carried out on Large-scale Mobile user data
The training of dual-stage makes obtained joint flux prediction model have higher accuracy and robustness, thus for mobile industry
Business, which is promoted, provides more accurate effective marketing thinking.
To achieve the above object, a kind of technical solution provided by the present invention are as follows: Large-scale Mobile customer traffic consumption intelligence
Energy prediction technique, comprising the following steps:
1) mobile subscriber attribute feature and consumer behavior data, visualization are collected, and is pre-processed；
2) structural classification fallout predictor and regressive predictor, are trained, and obtain the prediction model of two different scales；
3) joint classification fallout predictor and regressive predictor are trainable linear combination, carry out second stage training, obtain
Joint forecast model；
4) joint forecast model is used, user next month is predicted according to the attributive character of mobile subscriber and consumer behavior
Flow consumption value.
In step 1), before being pre-processed, million grades of data visualizations are carried out, quickly to identify data characteristic and inspection
Rope exceptional value, method for visualizing the following steps are included:
1.1) carrying out Hash to all feature fields of mobile subscriber divides bucket sectionization to operate；
1.2) any two feature field is taken, one is X-axis, another is Y-axis, fastens drawing data in cartesian coordinate
Point；
1.3) the identical data point of characteristic value will not cover mutually, but by it is point-by-point it is compact arranged in the form of be drawn.
In step 1), pretreatment operation include from extracted in data customer flow consumption field as label after, it is right
It carries out a point bucket sectionization processing, to form the value class duality of flow consumption field, is adapted to classification fallout predictor and returns
Return the associated prediction of fallout predictor.
In step 2), the classification fallout predictor is neural network DNN classification fallout predictor, and construction and training process are such as
Under:
2.1.1 input layer, output layer, hidden layer neural unit) are constructed；Each one layer of input layer, output layer, the list of input layer
First number is corresponding with data dimension, and output layer unit number is corresponding with classification number；Hidden layer is any several layers；
2.1.2) neural network is from input layer to hidden layer, then arrives output layer, successively promotes, and it is single to calculate every layer of nerve
Member, formula are as follows:
In formula, a is neural unit activation value, and z is the input of neural unit activation primitive, and g is activation primitive, and l is nerve
The serial number of network layer, the neural unit serial number that k is l+1 layers, slFor l layers of neural unit number, Θ(l)For l layers of parameter
Matrix；Indicate the activation value of l+1 k-th of neural unit of layer,Indicate k-th of neural unit activation letter of l+1 layer
Several inputs；
2.1.3) using cross entropy as cost function, the training of neural network is carried out, cost function is as follows:
Cost function is the sum of two；In first item, m is total sample number, and K is output layer neural unit sum, and i is sample
Serial number, k are neural unit serial number,For the value of k-th of neural unit in the true value of i-th of sample, x(i)For i-th of sample
Feature vector, (hθ(x(i)))kFor the value of k-th of neural unit in the predicted value of i-th of sample；Section 2 is regularization term,
Wherein λ is regularization coefficient, and L is the neural network number of plies,For l layers parameter matrix j-th of parameter of the i-th row；
2.1.4) optimizer of training process uses adam adaptive optimization device, saves "current" model after the completion of training.
In step 2), the regressive predictor is decision tree regressive predictor, and construction and training process are as follows:
2.2.1 the CART regression tree that feature is selected according to Gini coefficient) is constructed, the tree is every time to the value of some feature
Bipartite shape is carried out into binary tree；
2.2.2 constraint condition) is added to tree, the bifurcated smallest sample number for limiting depth capacity, setting intermediate node, and
The smallest sample number of each leaf node；
2.2.3 tree-model) is assessed by cost function of mean square error, and saves the smallest model of error；Mean square error meter
It is as follows to calculate formula:
In formula, MSE is mean square error, and m is total sample number, y(i)For the true value of i-th of sample, x(i)For i-th of sample
Feature vector, hθ(x(i)) be i-th of sample predicted value.
In step 3), linear combination is carried out to the classification fallout predictor and regressive predictor that learn, and instructed again
Practice, obtain joint forecast model, detailed process is as follows:
3.1) activation value for taking neural network classification fallout predictor the last layer, the uninterrupted section class after being followed successively by point bucket
It is not worth corresponding confidence level；
3.2) the uninterrupted predicted value of decision tree regressive predictor is taken；
3.3) value of all categories of classification fallout predictor is carried out with corresponding confidence level by element multiplication, and carries out local linear group
It closes, then under the action of impact factor, classification fallout predictor and regressive predictor carry out global linear combination, and specific formula is such as
Under:
In formula, hθ(x) indicate that the hypothesis function of joint forecast model, Vector_classes are uninterrupted section classification
It is worth vector, Vector_confidence is the corresponding confidence level vector of uninterrupted section class label, wTGeneral ability is predicted for classification
The weight vectors of portion's linear combination,For regressive predictor real value output as a result, α and β be respectively classify fallout predictor and
The impact factor of regressive predictor；
3.4) weight vectors of impact factor and classification fallout predictor are trained simultaneously, using mean square error as cost letter
Number saves "current" model after the completion of training using adam adaptive optimization device.
In step 4), the flow consumption of mobile subscriber's next month is predicted using joint forecast model, compared with independent
Accuracy and robustness are had more using classification fallout predictor or regressive predictor, detailed process is as follows:
4.1) pretreatment operation is carried out to new mobile user data, pretreatment operation includes that user is extracted from data
After flow consumes field as label, a point bucket sectionization processing is carried out to it, to form the value class two of flow consumption field
As property, it is adapted to the associated prediction of classification fallout predictor and regressive predictor；
4.2) using pretreated data as input, joint forecast model is run；
4.3) model exports predicted value.
Compared with prior art, the present invention have the following advantages that with the utility model has the advantages that
1, the present invention is united regressive predictor and classification fallout predictor, in the prediction applied to same label, than
Single model has higher accuracy and robustness.
2, the present invention is trained by dual-stage with classification the combining of fallout predictor in regressive predictor, improves modelling effect
And stability.
3, the present invention realizes the two-dimensional visualization of million grades of data points, to observation data characteristic, retrieval anomalies value, explains
Model etc. is very helpful.
4, the present invention, which carries out flow business marketing to telecom operators, directive function, can provide customization for mobile subscriber
Change flow package to recommend, and then improves the flow business revenue of operator and the usage experience of user.
Detailed description of the invention
Fig. 1 is the method for the present invention logical flow chart.
Specific embodiment
The present invention is further explained in the light of specific embodiments.
As shown in Figure 1, Large-scale Mobile customer traffic provided by the present embodiment consumes intelligent Forecasting, including following
Step:
1) mobile subscriber attribute feature and consumer behavior data are collected, are visualized, and carry out pretreatment operation；
1.1) visualization process is as follows:
1.1.1) carrying out Hash to all feature fields of mobile subscriber divides bucket sectionization to operate；
1.1.2 any two feature field) is taken, one is X-axis, another is Y-axis, fastens drafting number in cartesian coordinate
Strong point；
1.1.3) the identical data point of characteristic value will not cover mutually, but by it is point-by-point it is compact arranged in the form of be drawn.
1.2) visualization result is combined, pretreatment operation is carried out, process is as follows:
1.2.1 it) deletes there are the field of exceptional value, reduces noise present in data set；
1.2.2 the field of 99% or more NaN value accounting) is deleted；
1.2.3) feature field with high radix is counted, branch mailbox is carried out according to the frequency of occurrence of field value,
The feature distribution of the high radix of script is mapped in the new feature distribution of a low radix, and retains the height in former distribution as far as possible
Frequency part；
It 1.2.4 is) field of time, the time ruler for selecting variance moderate according to the floating variation range of field value to type
Degree, is converted into time offset, and be standardized；
1.2.5 one-hot coding) is carried out to classification field；
1.2.6) to there are the field of missing values progress mean values to fill up；
1.2.7 it) extracts flow and consumes field, carry out a point bucket sectionization processing.
2) structural classification fallout predictor and regressive predictor, are trained, and obtain the prediction model of two different scales；
2.1) neural network classification fallout predictor construction, training, preservation process are as follows:
2.1.1 input layer, output layer, hidden layer neural unit) are constructed.Each one layer of input layer, output layer, the list of input layer
First number is corresponding with data dimension, and output layer unit number is corresponding with classification number.Hidden layer is two layers, and unit number is respectively 32,64
It is a.
2.1.2) neural network is from input layer to hidden layer, then arrives output layer, successively promotes, and it is single to calculate every layer of nerve
Member.Formula is as follows:
In formula, a is neural unit activation value, and z is the input of neural unit activation primitive, and g is activation primitive, and l is nerve
The serial number of network layer, the neural unit serial number that k is l+1 layers, slFor l layers of neural unit number, Θ(l)For l layers of parameter
Matrix.Indicate the activation value of l+1 k-th of neural unit of layer,Indicate k-th of neural unit activation letter of l+1 layer
Several inputs.
2.1.3) using cross entropy as cost function, the training of neural network is carried out.Cost function is as follows:
Cost function is the sum of two.In first item, m is total sample number, and K is output layer neural unit sum, and i is sample
Serial number, k are neural unit serial number,For the value of k-th of neural unit in the true value of i-th of sample, x(i)For i-th of sample
Feature vector, (hθ(x(i)))kFor the value of k-th of neural unit in the predicted value of i-th of sample.Section 2 is regularization term,
Wherein λ is regularization coefficient, and L is the neural network number of plies,For l layers parameter matrix j-th of parameter of the i-th row.
2.1.4) optimizer of training process uses adam adaptive optimization device, and initial learning rate is set as 0.0001.Instruction
Experienced batch size and maximum number of iterations are respectively 200,200.
2.1.5) in each training process, batch is converted from input layer input network by hidden layer data one by one,
Output layer obtains prediction result, and predicted value and true value compare, and according to cost function calculation training loss and gradient, updates network
Parameter.When training loss improves less than 0.0001 or network training complete 200 times, in training process after iteration twice in succession
Only.After the completion of training, "current" model is saved.
2.2) decision tree regressive predictor construction, training, preservation process are as follows:
2.2.1 the CART regression tree that feature is selected according to Gini coefficient) is constructed, the tree is every time to the value of some feature
Bipartite shape is carried out into binary tree.
2.2.2 constraint condition, depth capacity 10) are added to tree, the intermediate node of tree should at least there are 1000 sample sides
Bifurcated can be carried out, while each leaf node also should at least there are 50 samples.
2.2.3 tree-model) is assessed by cost function of mean square error, and saves the smallest model of error.Mean square error meter
It is as follows to calculate formula:
In formula, MSE is mean square error, and m is total sample number, y(i)For the true value of i-th of sample, x(i)For i-th of sample
Feature vector, hθ(x(i)) be i-th of sample predicted value.
3) joint classification fallout predictor and regressive predictor are trainable linear combination, carry out second stage training, obtain
Joint forecast model, detailed process is as follows:
3.1) activation value for taking neural network classification fallout predictor the last layer, the uninterrupted section class after being followed successively by point bucket
It is not worth corresponding confidence level.
3.2) the uninterrupted predicted value of decision tree regressive predictor is taken.
3.3) value of all categories of classification fallout predictor is carried out with corresponding confidence level by element multiplication, and carries out local linear group
It closes.Then under the action of impact factor, division fallout predictor and regressive predictor carry out global linear combination.Specific formula is such as
Under:
In formula, hθ(x) indicate that the hypothesis function of joint forecast model, Vector_classes are uninterrupted section classification
It is worth vector, Vector_confidence is the corresponding confidence level vector of uninterrupted section class label, wTGeneral ability is predicted for classification
The weight vectors of portion's linear combination,For regressive predictor real value output as a result, α and β be respectively classify fallout predictor and
The impact factor of regressive predictor.
3.4) weight vectors of impact factor and classification fallout predictor are trained simultaneously, using mean square error as cost letter
Number saves "current" model after the completion of training using adam adaptive optimization device.
4) joint forecast model is used, user next month is predicted according to the attributive character of mobile subscriber and consumer behavior
Flow consumption value, detailed process is as follows:
4.1) new mobile user data is carried out and identical pretreatment operation in step 1.2)；
4.2) using pretreated data as input, joint forecast model is run
4.3) model exports predicted value.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (7)
1. a kind of Large-scale Mobile customer traffic consumes intelligent Forecasting, which comprises the following steps:
1) mobile subscriber attribute feature and consumer behavior data, visualization are collected, and is pre-processed；
2) structural classification fallout predictor and regressive predictor, are trained, and obtain the prediction model of two different scales；
3) joint classification fallout predictor and regressive predictor are trainable linear combination, carry out second stage training, are combined
Prediction model；
4) joint forecast model is used, the flow of user next month is predicted according to the attributive character of mobile subscriber and consumer behavior
Consumption value.
2. a kind of Large-scale Mobile customer traffic according to claim 1 consumes intelligent Forecasting, it is characterised in that:
In step 1), before being pre-processed, million grades of data visualizations are carried out, quickly to identify data characteristic and retrieval anomalies value,
Method for visualizing the following steps are included:
1.1) carrying out Hash to all feature fields of mobile subscriber divides bucket sectionization to operate；
1.2) any two feature field is taken, one is X-axis, another is Y-axis, fastens drawing data point in cartesian coordinate；
1.3) the identical data point of characteristic value will not cover mutually, but by it is point-by-point it is compact arranged in the form of be drawn.
3. a kind of Large-scale Mobile customer traffic according to claim 1 consumes intelligent Forecasting, it is characterised in that:
In step 1), pretreatment operation includes carrying out a point bucket to it after extracting customer flow consumption field in data as label
Sectionization processing is adapted to classification fallout predictor and regressive predictor to form the value class duality of flow consumption field
Associated prediction.
4. a kind of Large-scale Mobile customer traffic according to claim 1 consumes intelligent Forecasting, it is characterised in that:
In step 2), the classification fallout predictor is neural network DNN classification fallout predictor, and construction and training process are as follows:
2.1.1 input layer, output layer, hidden layer neural unit) are constructed；Each one layer of input layer, output layer, the unit number of input layer
Corresponding with data dimension, output layer unit number is corresponding with classification number；Hidden layer is any several layers；
2.1.2) neural network is from input layer to hidden layer, then arrives output layer, successively promotes, and calculates every layer of neural unit, public
Formula is as follows:
In formula, a is neural unit activation value, and z is the input of neural unit activation primitive, and g is activation primitive, and l is neural network
The serial number of layer, the neural unit serial number that k is l+1 layers, slFor l layers of neural unit number, Θ(l)For l layers of parameter square
Battle array；Indicate the activation value of l+1 k-th of neural unit of layer,Indicate k-th of neural unit activation primitive of l+1 layer
Input；
2.1.3) using cross entropy as cost function, the training of neural network is carried out, cost function is as follows:
Cost function is the sum of two；In first item, m is total sample number, and K is output layer neural unit sum, and i is sample sequence
Number, k is neural unit serial number,For the value of k-th of neural unit in the true value of i-th of sample, x(i)For i-th sample
Feature vector, (hθ(x(i)))kFor the value of k-th of neural unit in the predicted value of i-th of sample；Section 2 is regularization term,
Middle λ is regularization coefficient, and L is the neural network number of plies,For l layers parameter matrix j-th of parameter of the i-th row；
2.1.4) optimizer of training process uses adam adaptive optimization device, saves "current" model after the completion of training.
5. a kind of Large-scale Mobile customer traffic according to claim 1 consumes intelligent Forecasting, it is characterised in that:
In step 2), the regressive predictor is decision tree regressive predictor, and construction and training process are as follows:
2.2.1 the CART regression tree that feature is selected according to Gini coefficient) is constructed, which every time carries out the value of some feature
Bipartite shape is at binary tree；
2.2.2 constraint condition) is added to tree, the bifurcated smallest sample number for limiting depth capacity, setting intermediate node, and it is each
The smallest sample number of leaf node；
2.2.3 tree-model) is assessed by cost function of mean square error, and saves the smallest model of error；Mean square error calculates public
Formula is as follows:
In formula, MSE is mean square error, and m is total sample number, y(i)For the true value of i-th of sample, x(i)For the spy of i-th of sample
Levy vector, hθ(x(i)) be i-th of sample predicted value.
6. a kind of Large-scale Mobile customer traffic according to claim 1 consumes intelligent Forecasting, it is characterised in that:
In step 3), linear combination is carried out to the classification fallout predictor and regressive predictor that learn, and be trained again, combined
Prediction model, detailed process is as follows:
3.1) activation value for taking neural network classification fallout predictor the last layer, the uninterrupted section class label after being followed successively by point bucket
Corresponding confidence level；
3.2) the uninterrupted predicted value of decision tree regressive predictor is taken；
3.3) value of all categories of classification fallout predictor is carried out with corresponding confidence level by element multiplication, and carries out local linear combination,
Then under the action of impact factor, classification fallout predictor and regressive predictor carry out global linear combination, specific formula is as follows:
In formula, hθ(x) indicate joint forecast model hypothesis function, Vector_classes be uninterrupted section class label to
Amount, Vector_confidence are the corresponding confidence level vector of uninterrupted section class label, wTFor fallout predictor local line of classifying
Property combination weight vectors,For real value output fallout predictor and recurrence as a result, α and β respectively classifies of regressive predictor
The impact factor of fallout predictor；
3.4) weight vectors of impact factor and classification fallout predictor are trained simultaneously, using mean square error as cost function, are made
With adam adaptive optimization device, "current" model is saved after the completion of training.
7. a kind of Large-scale Mobile customer traffic according to claim 1 consumes intelligent Forecasting, it is characterised in that:
In step 4), the flow consumption of mobile subscriber's next month is predicted using joint forecast model, it is pre- compared with classification is used alone
Accuracy and robustness will be had more by surveying device or regressive predictor, and detailed process is as follows:
4.1) pretreatment operation is carried out to new mobile user data, pretreatment operation includes that customer flow is extracted from data
After field is consumed as label, a point bucket sectionization processing is carried out to it, to form the value class duality of flow consumption field,
It is adapted to the associated prediction of classification fallout predictor and regressive predictor；
4.2) using pretreated data as input, joint forecast model is run；
4.3) model exports predicted value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910006654.8A CN109787821B (en) | 2019-01-04 | 2019-01-04 | Intelligent prediction method for large-scale mobile client traffic consumption |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910006654.8A CN109787821B (en) | 2019-01-04 | 2019-01-04 | Intelligent prediction method for large-scale mobile client traffic consumption |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109787821A true CN109787821A (en) | 2019-05-21 |
| CN109787821B CN109787821B (en) | 2020-06-19 |
Family
ID=66499919
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910006654.8A Expired - Fee Related CN109787821B (en) | 2019-01-04 | 2019-01-04 | Intelligent prediction method for large-scale mobile client traffic consumption |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109787821B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112330368A (en) * | 2020-11-16 | 2021-02-05 | 腾讯科技（深圳）有限公司 | Data processing method, system, storage medium and terminal equipment |
| CN112926845A (en) * | 2021-02-18 | 2021-06-08 | 上海翰声信息技术有限公司 | Big data based outbound method, electronic device and computer readable storage medium |
| CN113869750A (en) * | 2021-09-30 | 2021-12-31 | 中国计量大学 | Automatic elevator maintenance enterprise rating system based on big data |
| CN117057852A (en) * | 2023-10-09 | 2023-11-14 | 北京光尘环保科技股份有限公司 | Internet marketing system and method based on artificial intelligence technology |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102982373A (en) * | 2012-12-31 | 2013-03-20 | 山东大学 | OIN (Optimal Input Normalization) neural network training method for mixed SVM (Support Vector Machine) regression algorithm |
| US20150356576A1 (en) * | 2011-05-27 | 2015-12-10 | Ashutosh Malaviya | Computerized systems, processes, and user interfaces for targeted marketing associated with a population of real-estate assets |
| CN105302831A (en) * | 2014-07-18 | 2016-02-03 | 上海星红桉数据科技有限公司 | High-speed calculation analysis method based on mass user behavior data |
| CN106096623A (en) * | 2016-05-25 | 2016-11-09 | 中山大学 | A kind of crime identifies and Forecasting Methodology |
| US20160358290A1 (en) * | 2012-04-20 | 2016-12-08 | Humana Inc. | Health severity score predictive model |
| CN107972662A (en) * | 2017-10-16 | 2018-05-01 | 华南理工大学 | To anti-collision warning method before a kind of vehicle based on deep learning |
| WO2018123051A1 (en) * | 2016-12-28 | 2018-07-05 | 株式会社日立製作所 | Information processing system and method |
| CN108388954A (en) * | 2018-01-05 | 2018-08-10 | 上海电力学院 | A kind of cascade hydropower robust Optimization Scheduling based on random security domain |
-
2019
- 2019-01-04 CN CN201910006654.8A patent/CN109787821B/en not_active Expired - Fee Related
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150356576A1 (en) * | 2011-05-27 | 2015-12-10 | Ashutosh Malaviya | Computerized systems, processes, and user interfaces for targeted marketing associated with a population of real-estate assets |
| US20160358290A1 (en) * | 2012-04-20 | 2016-12-08 | Humana Inc. | Health severity score predictive model |
| CN102982373A (en) * | 2012-12-31 | 2013-03-20 | 山东大学 | OIN (Optimal Input Normalization) neural network training method for mixed SVM (Support Vector Machine) regression algorithm |
| CN105302831A (en) * | 2014-07-18 | 2016-02-03 | 上海星红桉数据科技有限公司 | High-speed calculation analysis method based on mass user behavior data |
| CN106096623A (en) * | 2016-05-25 | 2016-11-09 | 中山大学 | A kind of crime identifies and Forecasting Methodology |
| WO2018123051A1 (en) * | 2016-12-28 | 2018-07-05 | 株式会社日立製作所 | Information processing system and method |
| CN107972662A (en) * | 2017-10-16 | 2018-05-01 | 华南理工大学 | To anti-collision warning method before a kind of vehicle based on deep learning |
| CN108388954A (en) * | 2018-01-05 | 2018-08-10 | 上海电力学院 | A kind of cascade hydropower robust Optimization Scheduling based on random security domain |
Non-Patent Citations (1)
| Title |
|---|
| 杨彬: "一种基于RFM模型数据挖掘处理双阶段客户关联分类方法", 《统计与决策》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112330368A (en) * | 2020-11-16 | 2021-02-05 | 腾讯科技（深圳）有限公司 | Data processing method, system, storage medium and terminal equipment |
| CN112926845A (en) * | 2021-02-18 | 2021-06-08 | 上海翰声信息技术有限公司 | Big data based outbound method, electronic device and computer readable storage medium |
| CN113869750A (en) * | 2021-09-30 | 2021-12-31 | 中国计量大学 | Automatic elevator maintenance enterprise rating system based on big data |
| CN117057852A (en) * | 2023-10-09 | 2023-11-14 | 北京光尘环保科技股份有限公司 | Internet marketing system and method based on artificial intelligence technology |
| CN117057852B (en) * | 2023-10-09 | 2024-01-26 | 头流(杭州)网络科技有限公司 | Internet marketing system and method based on artificial intelligence technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109787821B (en) | 2020-06-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109523018B (en) | Image classification method based on deep migration learning | |
| Bi et al. | Daily tourism volume forecasting for tourist attractions | |
| Li et al. | Prediction for tourism flow based on LSTM neural network | |
| CN109787821A (en) | A kind of Large-scale Mobile customer traffic consumption intelligent Forecasting | |
| CN110363282B (en) | Network node label active learning method and system based on graph convolution network | |
| CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
| CN108009674A (en) | Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks | |
| CN103761254B (en) | Method for matching and recommending service themes in various fields | |
| CN113762595B (en) | Traffic time prediction model training method, traffic time prediction method and equipment | |
| Chen et al. | Groundwater level prediction using SOM-RBFN multisite model | |
| CN111127246A (en) | Intelligent prediction method for transmission line engineering cost | |
| CN111860989B (en) | LSTM neural network short-time traffic flow prediction method based on ant colony optimization | |
| CN112925908A (en) | Attention-based text classification method and system for graph Attention network | |
| CN108090788A (en) | Ad conversion rates predictor method based on temporal information integrated model | |
| CN109614495A (en) | A kind of associated companies method for digging of combination knowledge mapping and text information | |
| CN106970986A (en) | Urban waterlogging influence degree method for digging and system based on deep learning | |
| CN111178585A (en) | Fault reporting amount prediction method based on multi-algorithm model fusion | |
| Kayhomayoon et al. | A new approach for regional groundwater level simulation: clustering, simulation, and optimization | |
| CN110990718A (en) | Social network model building module of company image improving system | |
| Huang et al. | Research on urban modern architectural art based on artificial intelligence and GIS image recognition system | |
| CN108416524A (en) | Estate planning based on a figure general framework refines deciphering method | |
| CN115310532A (en) | Basin multipoint prediction early warning method based on space-time association mixed deep learning model | |
| CN113326919A (en) | Traffic travel mode selection prediction method based on computational graph | |
| CN115080795A (en) | Multi-charging-station cooperative load prediction method and device | |
| CN113868537B (en) | Recommendation method based on multi-behavior session graph fusion |
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: 20200619 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |