CN108459997A - High skewness data value probability forecasting method based on deep learning and neural network - Google Patents
High skewness data value probability forecasting method based on deep learning and neural network Download PDFInfo
- Publication number
- CN108459997A CN108459997A CN201810121033.XA CN201810121033A CN108459997A CN 108459997 A CN108459997 A CN 108459997A CN 201810121033 A CN201810121033 A CN 201810121033A CN 108459997 A CN108459997 A CN 108459997A
- Authority
- CN
- China
- Prior art keywords
- data
- value
- value coefficient
- probability
- deep learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Mathematical Analysis (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Mathematical Optimization (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- Algebra (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to the high skewness data value probability forecasting methods based on deep learning and neural network, and this approach includes the following steps：Initial data is analyzed with Feature Engineering by data analysis, using input data of the specific collection algorithm extraction comprising user information and its feature based on custom logic；The value coefficient of prediction client after the input data is obtained by preset data model；The value coefficient is converted to people's intuitive probability data to by mating probabilistic forecasting formula.The present invention can effectively help the platform for possessing mass data to carry out data analysis, high value individual be accurately positioned, to improve operation and marketing efficiency.
Description
Technical field
The present invention relates to big data treatment technologies, and in particular to the high skewness data valence based on deep learning and neural network
It is worth probability forecasting method.
Background technology
With Internet of Things, SAAS service platforms, the fast development in the fields such as internet, a large amount of data are collected, arrange,
Structuring is simultaneously stored in database.However the value that these data are included is not developed and utilized but well.With certain
For the SAAS service platforms (hereinafter simply referred to as platform) of line education services, the trimestral client in only one site includes data just
There are as many as up to ten thousand, the information of multiple dimensions, such as gender, contact method, customer habits etc. are included per data.These multidimensional
The information of degree and client's visiting rate, client can directly help the business of the platform at single rate, the information associations such as customer value,
Especially marketing efficiency.However the Data Analyst of current platform still relies on artificial calculate and based on the simple of Excel tables
Formula calculates.Such analysis method brings problems with：1, data volume is big, manpower analysis long time period, and analyze by
More people complete, and lack unanimously and comparability；2, general statistical method is only applicable to overall condition analysis, can not be directed to individual
Characteristic information is accurately judged, often misses the gold after establishing archives 48 hours；3, data dimension it is more (10 dimensions with
On), artificial and EXCEL analyses are difficult to correctly quantify the influence of different dimensions, count the relevant information of all visiting users merely
It is difficult to find evident regularity；4, data distribution is highly asymmetric (high skewness), using visiting rate as value goal for, be marked
For high value individual ratio less than 4%, therefore carry out analysis using conventional statistics or general machine learning and be difficult to obtain
Satisfactory recall rate (recal l).
Invention content
It is difficult to obtain the skill of satisfactory recall rate when analyzing data it is an object of the present invention to solve the prior art
Art problem.
To achieve the above object, the present invention provides the high skewness data value probability based on deep learning and neural network
Prediction technique, this approach includes the following steps：
Initial data is analyzed with Feature Engineering by data analysis, extraction is defeated comprising user information and its feature
Enter data；The value coefficient of prediction client after the input data is obtained by preset data model；It is pre- by mating probability
Formula is surveyed to convert to people's intuitive probability data the value coefficient to.
Preferably, described that initial data is analyzed with Feature Engineering by data analysis, it is patrolled using based on self-defined
Input data step of the specific collection algorithm extraction comprising user information and its feature collected；Including：It will be on SAAS service platforms
The common numerical value for including binary or quantization that finish message is appreciated that at machine learning.
Preferably, the value coefficient step that prediction client after the input data is obtained by preset data model,
Including：The value coefficient that client is predicted using deep learning and nerual network technique, wherein each deep learning component units needle
It sums after being multiplied by a weight to each input information, the summation is then acted on numerically by an activation primitive.
Preferably, described to be converted to people's intuitive probability number the value coefficient to by mating probabilistic forecasting formula
According to step, including：The value coefficient is converted to people's intuitive probability data to by following mating probabilistic forecasting formula
Wherein, it is probability data, p is the precision that model shows on training dataset, and α is the value coefficient of client, θ
For the threshold value of high-value user group.
Preferably, the θ values are 0.25.
The present invention is based on the multi-layered perception neural networks of data-driven deep learning, and the multidimensional number of degrees are established by Feature Engineering
According to the association between the slight change and dimension of each dimension prejudges, for solving the very uneven feelings of data distribution
Identification decision application under condition.The present invention can also effectively help the platform for possessing mass data to carry out data analysis, accurate fixed
Position high value individual, to improve operation and marketing efficiency.
Description of the drawings
Fig. 1 is the high skewness data value probabilistic forecasting provided in an embodiment of the present invention based on deep learning and neural network
Method flow schematic diagram；
Fig. 2 is deep learning component units logical operation schematic diagram；
Fig. 3 is the sample schematic diagram of multi-layered perception neural networks；
Fig. 4 is using the visiting of each individual consumer of the invention obtained or at single rate.
Specific implementation mode
After embodiments of the present invention are described in detail by way of example below in conjunction with attached drawing, of the invention its
His features, characteristics, and advantages will be more obvious.
Fig. 1 is the high skewness data value probabilistic forecasting provided in an embodiment of the present invention based on deep learning and neural network
Method flow schematic diagram.As shown in Figure 1, this approach includes the following steps：
S101 analyzes initial data with Feature Engineering by data analysis, using the spy based on custom logic
Determine input data of the set based algorithm extraction comprising user information and its feature；
Specifically, include user information using the specific collection algorithm extraction based on custom logic in the embodiment of the present invention
And its input data step of feature：
Binary that finish message is appreciated that at machine learning or quantization are included by common on SAAS service platforms
Numerical value.Set of algorithms is made of multiple custom logics, and set of algorithms list and logical specification record include mainly three in table 1
Kind logic：Quantization, one-hot coding and polynary (containing binary) classification.
Case 1 (quantization) raw information is 11 phone numbers, using the quantity of quantization extracted wherein 6 and 8, such as hand
Machine number is 18566279381, then the input generated is 4.
Case 2 (one-hot coding) raw information is parent's identity, including father, mother, grandmother, a variety of situations such as unknown,
Two binary inputs (input 3 and input 4) are generated using one-hot coding logic：If raw information is father/father, input
3 be 1, and input 4 is 0；If raw information is mother/mother, input 3 is 0, and input 4 is 1；If raw information is other
Situation, then it is all 0 to input 3 and input 4.
Case 3 (multivariate classification) raw information is gender, and information is divided into man, female and unknown 3 using multivariate classification logic
Kind situation is then set as 1 if it is man, and to be set as if female, -1. is unknown and other situations are set as 0 and (consider the 1 of male to female ratio:1 point
Cloth, and unknown situation does not interfere with the mathematic expectaion of male to female ratio and is influenced caused by model)
S102 obtains the value coefficient of prediction client after the input data by preset data model；
The embodiment of the present invention is directed to distribution very non-uniform data (such as visiting rate<4%, sign-on rate<
0.8%), general machine learning method can not find corresponding characteristic rule from the information input of more than 10 dimensions.Compared to
Statistical model, AI models are more suitable for the scene that input feature vector does not have evident regularity or difference, it is therefore desirable to apply deep learning
Neural network recognization high-value user group.The Schema information of preset data model of the embodiment of the present invention is as follows：
Multi-layered perception neural networks, except input (14 units) with output (2 units, respectively for visiting probability at
Work(probability) outside layer, including double hidden layers, one layer is 10 neurons, and one is 5 neurons.Model liberalization parameter matrix
For 206 weight/deviations.
Realization platform is that Python has invoked tensorflow, numpy modules using Anacoda3 (64bit) environment.
Hidden layer activation primitive is Tanh, it is contemplated that exports the probability results between being 0 to 1, output layer uses s
The activation primitive of igmoid (0 to 1).
It amounts to and obtains 10000 customer informations, data are used with 7:3(7000:3000) ratio cut partition is at training dataset
And test data set.
Training method conducts coaching method after being standard Adam, cost function is prediction variance yields, is followed for 10000 cycle of training
Ring, study digit rate are 0.01.
The performance for weighing model needs the binary situation for being 0/1 in view of output target, connects between being 0 to 1 due to output
Continuous numerical value, therefore the dualization for being exported model by a threshold value when weighing prediction accuracy, and confusion matrix
Quantify performance results：
Parameter for weighing model performance is usually predictablity rate accuracy：
Rate of precision precision when being predicted as 1：
With recall rate recall：
The case where Accuracy is working standard, but not applicable data are high skewness, it is comprehensive to use precision,
Recall is more suitable.Defining the threshold θ of high-value user group can flexibly select, and high threshold often represents tightened up screening,
Therefore precision precisions meeting higher, but therefore many high conversion users are also screened out, therefore recall rate recall meetings
Decline.Therefore threshold value is selected as an equalization point, and threshold θ is 0.25 in embodiments of the present invention.
The value coefficient that client is predicted using deep learning and nerual network technique, wherein each deep learning component units
It sums after being multiplied by a weight for each input information, the summation numerically (such as Fig. 2 is then acted on by an activation primitive
Shown in Fig. 3).Fig. 3 is the sample for the multi-layered perception neural networks being made of many virtual neuron links, and Bias is effect
Deviation factor on hidden layer.
Deep learning (deeplearning) is the branch of machine learning, be one kind attempt use comprising labyrinth or by
Multiple process layers that multiple nonlinear transformation is constituted carry out data the algorithm of higher level of abstraction.Deep learning is to data carry out table
Levy the algorithm of study.Observation (such as piece image) can use a plurality of ways to indicate, such as each pixel intensity value to
Amount, or be more abstractively expressed as a series of sides, specific shape region etc..And it is easier using certain specific representation methods
The learning tasks from example.
Artificial neural network is a kind of common deep learning model, is a kind of structure and work(of mimic biology neural network
The mathematical model or computation model of energy, for carrying out estimation or approximation to function.Neural network is joined by a large amount of artificial neuron
Knot is calculated.In most cases artificial neural network can change internal structure on the basis of external information, be it is a kind of from
Adaptive system.Typical neural network has following three parts：
Structure (Architecture) refers to the topological relation of variable and they in network.For example, the change in neural network
Amount can be the excitation value (activities of the neurons) of the weight (weights) and neuron of neuron connection.
There are one the dynamics of short-term time scale to advise for excitation function (Activity Rule) major part neural network model tool
Then, define how neuron changes the excitation value of oneself according to the activity of other neurons.
Learning rules (Learning Rule) learning rules specify how the weight in network is adjusted with time stepping method
It is whole.This is generally seen as a kind of dynamics rule of long time scale.
S103 is converted to people's intuitive probability data the value coefficient to by mating probabilistic forecasting formula.
Specifically, it is converted to people's intuitive probability number the value coefficient to by following mating probabilistic forecasting formula
According to
Wherein, it is probability data, p is the precision that model shows on training dataset, and α is the value coefficient of client, θ
For the threshold value of high-value user group.Preferably, the θ values are 0.25.
The model obtained by training data can predict the value coefficient α (0-1) of client, it is generally the case that value system
Number higher (close to the 1) probability that then user is high-value user is higher.Obtaining can be by threshold from a large amount of after value coefficient
High-value user group is positioned in the list of user, or user is sorted from high to low by sort method, so that flat
The business people of platform can more accurately position potential customers, improve operation and marketing efficiency comprehensively.Although however value coefficient
It is directly proportional to the probability of high value customer, but coefficient does not have intuitive for the people used, therefore the present invention provides
Probabilistic forecasting formula with set of model acts on the value coefficient of model output and obtains more intuitive probability data η and make for platform
User refers to.
The model of the embodiment of the present invention, which is applied, draws a portrait in user in analysis, can be accurate according to the various dimensions information that user leaves
The really value coefficient of analysis different user, to provide accurate OA operation analysis/sales data for customer service and operation personnel and support,
The conversion ratio of operational efficiency and sale is improved comprehensively.Three innovative points of the embodiment of the present invention are：
1, the system for carrying out the big data analysis of high speed for online saas platforms and accurately identifying high-value user；
2, various dimensions input information can be analyzed and handle the deep learning neural network model of high partial velocities data；
3, user's value coefficient that machine directly exports can be converted to and is applied a formula matching for intuitive probability data.
The embodiment of the present invention be used in certain education SAAS service platforms user data on, Accurate Prediction its include user
Visiting rate and signing probability.The original visiting rate calculated according to the canonical statistics model of control group：3.97%, original Cheng Dan
Rate：0.73%, and deep learning neural network model positioning high value group visiting rate through the invention>20%, at single rate~
20%, improvement effect is notable.In addition, the present invention can precisely predict the visiting of each individual consumer or at single rate (such as Fig. 4 institutes
Show)：
By the sequence to value coefficient, can high value be effectively filtered out in user information from original include of magnanimity
The high-value user group of Stock discrimination that is more enriched with of user conversion ratio (visiting rate or at single rate) be significantly higher than it is average
It is horizontal.Such as original visiting rate is 3.97%, the visiting rate of the high value group of identification is more than 20%：
Test result shows, the visiting rate of the high-value user group based on the identification of deep learning neural network model and at
Single rate is significantly higher than initial data, can help to optimize sale effect for largely registering the output sequence after user information is analyzed
Rate.
It should be noted that above-described embodiment only is used for illustrating the structure and its working effect of the present invention, and it is not used as
It limits the scope of the invention.One of ordinary skilled in the art is right without prejudice to thinking of the present invention and structure
The adjustment or optimization that above-described embodiment carries out, should regard as the claims in the present invention and be covered.
Claims (5)
1. the high skewness data value probability forecasting method based on deep learning and neural network, which is characterized in that including following
Step：
Initial data is analyzed with Feature Engineering by data analysis, input number of the extraction comprising user information and its feature
According to；
The value coefficient of prediction client after the input data is obtained by preset data model；
The value coefficient is converted to people's intuitive probability data to by mating probabilistic forecasting formula.
2. according to the method described in claim 1, it is characterized in that, it is described by data analysis and Feature Engineering to initial data
It is analyzed, is walked using input data of the specific collection algorithm extraction comprising user information and its feature based on custom logic
Suddenly；Including：
By the numerical value for including binary or quantization that finish message is appreciated that at machine learning common on SAAS service platforms.
3. according to the method described in claim 1, it is characterized in that, described obtain the input data by preset data model
The value coefficient step of client is predicted afterwards, including：
Using the value coefficient of self-defined framework and the multi-layered perception neural networks technological prediction client of parameter, value coefficient is most
Zhongdao visit rate or conversion ratio of signing a bill.
4. according to the method described in claim 1, it is characterized in that, it is described by mating probabilistic forecasting formula by the value
Coefficient is converted into the intuitive probability data step of people, including：
The value coefficient is converted to people's intuitive probability data to by following mating probabilistic forecasting formula
Wherein, it is probability data, p is the precision that model shows on training dataset, and α is the value coefficient of client, and θ is height
It is worth the threshold value of user group.
5. according to the method described in claim 4, it is characterized in that, the θ values are 0.25.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810121033.XA CN108459997A (en) | 2018-02-07 | 2018-02-07 | High skewness data value probability forecasting method based on deep learning and neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810121033.XA CN108459997A (en) | 2018-02-07 | 2018-02-07 | High skewness data value probability forecasting method based on deep learning and neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108459997A true CN108459997A (en) | 2018-08-28 |
Family
ID=63239625
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810121033.XA Pending CN108459997A (en) | 2018-02-07 | 2018-02-07 | High skewness data value probability forecasting method based on deep learning and neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108459997A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109784979A (en) * | 2018-12-19 | 2019-05-21 | 重庆邮电大学 | A kind of supply chain needing forecasting method of big data driving |
| CN109976908A (en) * | 2019-03-15 | 2019-07-05 | 北京工业大学 | A kind of server cluster dynamic retractility method based on RNN time series forecasting |
| CN111191133A (en) * | 2019-12-31 | 2020-05-22 | 口口相传(北京)网络技术有限公司 | Service search processing method, device and equipment |
| CN113052383A (en) * | 2021-03-29 | 2021-06-29 | 上海酷量信息技术有限公司 | Income prediction method and device based on machine learning |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6625317B1 (en) * | 1995-09-12 | 2003-09-23 | Art Gaffin | Visual imaging system and method |
| CN101482888A (en) * | 2009-02-23 | 2009-07-15 | 阿里巴巴集团控股有限公司 | Website caller value computing method and system |
| CN103310286A (en) * | 2013-06-25 | 2013-09-18 | 浙江大学 | Product order prediction method and device with time series characteristics |
| CN103578057A (en) * | 2012-08-10 | 2014-02-12 | 北京奥齐都市网络科技有限公司 | Real estate value estimation method based on artificial neural network statistic model |
| CN105160498A (en) * | 2015-10-21 | 2015-12-16 | 北京普猎创新网络科技有限公司 | Personal value calculation method based on big data |
| CN105574741A (en) * | 2015-12-22 | 2016-05-11 | 北京奇虎科技有限公司 | Flow value assessment method, device, application method and system |
| CN106447489A (en) * | 2016-09-12 | 2017-02-22 | 中山大学 | Partially stacking blend based user credit assessment model |
| CN107274228A (en) * | 2017-06-23 | 2017-10-20 | 北京好豆网络科技有限公司 | The evaluation method and device of cuisines |
| CN107609700A (en) * | 2017-09-08 | 2018-01-19 | 欧若纳信息科技(苏州)有限公司 | A kind of customer value model optimization method based on machine learning |
-
2018
- 2018-02-07 CN CN201810121033.XA patent/CN108459997A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6625317B1 (en) * | 1995-09-12 | 2003-09-23 | Art Gaffin | Visual imaging system and method |
| CN101482888A (en) * | 2009-02-23 | 2009-07-15 | 阿里巴巴集团控股有限公司 | Website caller value computing method and system |
| CN103578057A (en) * | 2012-08-10 | 2014-02-12 | 北京奥齐都市网络科技有限公司 | Real estate value estimation method based on artificial neural network statistic model |
| CN103310286A (en) * | 2013-06-25 | 2013-09-18 | 浙江大学 | Product order prediction method and device with time series characteristics |
| CN105160498A (en) * | 2015-10-21 | 2015-12-16 | 北京普猎创新网络科技有限公司 | Personal value calculation method based on big data |
| CN105574741A (en) * | 2015-12-22 | 2016-05-11 | 北京奇虎科技有限公司 | Flow value assessment method, device, application method and system |
| CN106447489A (en) * | 2016-09-12 | 2017-02-22 | 中山大学 | Partially stacking blend based user credit assessment model |
| CN107274228A (en) * | 2017-06-23 | 2017-10-20 | 北京好豆网络科技有限公司 | The evaluation method and device of cuisines |
| CN107609700A (en) * | 2017-09-08 | 2018-01-19 | 欧若纳信息科技(苏州)有限公司 | A kind of customer value model optimization method based on machine learning |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109784979A (en) * | 2018-12-19 | 2019-05-21 | 重庆邮电大学 | A kind of supply chain needing forecasting method of big data driving |
| CN109976908A (en) * | 2019-03-15 | 2019-07-05 | 北京工业大学 | A kind of server cluster dynamic retractility method based on RNN time series forecasting |
| CN111191133A (en) * | 2019-12-31 | 2020-05-22 | 口口相传(北京)网络技术有限公司 | Service search processing method, device and equipment |
| CN111191133B (en) * | 2019-12-31 | 2023-12-01 | 口口相传(北京)网络技术有限公司 | Service search processing method, device and equipment |
| CN113052383A (en) * | 2021-03-29 | 2021-06-29 | 上海酷量信息技术有限公司 | Income prediction method and device based on machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Azadeh et al. | Improved estimation of electricity demand function by integration of fuzzy system and data mining approach | |
| CN109118013A (en) | A kind of management data prediction technique, readable storage medium storing program for executing and forecasting system neural network based | |
| Maknickienė et al. | Application of neural network for forecasting of exchange rates and forex trading | |
| CN108459997A (en) | High skewness data value probability forecasting method based on deep learning and neural network | |
| CN103714261B (en) | Intelligent auxiliary medical treatment decision supporting method of two-stage mixed model | |
| Kanavos et al. | Deep learning models for forecasting aviation demand time series | |
| Azadeh et al. | An adaptive network based fuzzy inference system–genetic algorithm clustering ensemble algorithm for performance assessment and improvement of conventional power plants | |
| CN112700324A (en) | User loan default prediction method based on combination of Catboost and restricted Boltzmann machine | |
| Ippolito et al. | Tax Crime Prediction with Machine Learning: A Case Study in the Municipality of São Paulo. | |
| Straton et al. | Big social data analytics for public health: Predicting facebook post performance using artificial neural networks and deep learning | |
| Wanke et al. | Revisiting camels rating system and the performance of Asean banks: a comprehensive mcdm/z-numbers approach | |
| Çakıt et al. | Comparative analysis of machine learning algorithms for predicting standard time in a manufacturing environment | |
| Nagaraju et al. | Methodologies used for customer churn detection in customer relationship management | |
| JP5401885B2 (en) | Model construction method, construction system, and construction program | |
| Zosimov et al. | Calculation the Measure of Expert Opinions Consistency Based on Social Profile Using Inductive Algorithms | |
| Wu et al. | Customer churn prediction for commercial banks using customer-value-weighted machine learning models | |
| Loaiza et al. | Modeling demand for air cargo in the Colombian context | |
| Couto et al. | Water quality modeling using artificial intelligence-based tools | |
| Viloria et al. | Predictive model for detecting customer’s purchasing behavior using data mining | |
| KR101918359B1 (en) | A Data Processing System Using AI | |
| El Mouna et al. | A Comparative Study of Urban House Price Prediction using Machine Learning Algorithms | |
| Shahoud et al. | Incorporating unsupervised deep learning into meta learning for energy time series forecasting | |
| Gaikwad et al. | Unveiling Market Dynamics through Machine Learning: Strategic Insights and Analysis | |
| Haghighi et al. | Customer Loyalty evaluation and prediction based on decision tree and artificial neural network: Case of Ofogh koorosh stores in tehran | |
| Dixit et al. | Investigating determinants of information technology investments by indian firms |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180828 |