CN110363439A - A kind of credit-graded approach based on consumer demographics' portrait - Google Patents

A kind of credit-graded approach based on consumer demographics' portrait Download PDF

Info

Publication number
CN110363439A
CN110363439A CN201910653033.9A CN201910653033A CN110363439A CN 110363439 A CN110363439 A CN 110363439A CN 201910653033 A CN201910653033 A CN 201910653033A CN 110363439 A CN110363439 A CN 110363439A
Authority
CN
China
Prior art keywords
month
user
credit
class
consumption
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
Application number
CN201910653033.9A
Other languages
Chinese (zh)
Inventor
安程治
李锐
于治楼
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong Inspur Artificial Intelligence Research Institute Co Ltd
Original Assignee
Shandong Inspur Artificial Intelligence Research Institute Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shandong Inspur Artificial Intelligence Research Institute Co Ltd filed Critical Shandong Inspur Artificial Intelligence Research Institute Co Ltd
Priority to CN201910653033.9A priority Critical patent/CN110363439A/en
Publication of CN110363439A publication Critical patent/CN110363439A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q10/00Administration; Management
    • G06Q10/06Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063Operations research, analysis or management
    • G06Q10/0639Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393Score-carding, benchmarking or key performance indicator [KPI] analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q30/00Commerce
    • G06Q30/06Buying, selling or leasing transactions
    • G06Q30/0601Electronic shopping [e-shopping]
    • G06Q30/0609Buyer or seller confidence or verification

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Educational Administration (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Finance (AREA)
  • Theoretical Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Physics & Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The present invention discloses a kind of credit-graded approach based on consumer demographics' portrait, is related to credit evaluation technical field.It is not perfect for existing grading system, think to judge the strong defect of subjectivity, the consumption data acquired using technical solution based on common carrier, known credit scoring corresponding with acquisition consumption data, classify first to the consumption data of acquisition, then consumption classification belonging to the consumption data and consumption data acquired by word2vec model treatment, portray consumer demographics' portrait, learn the consumption data of affiliated identity set using light GBM as base learner, consumption classification belonging to consumption data, the output information of word2vec model, credit scoring model is constructed with training, finally, the consumption data of affiliated identity set is analyzed using the credit scoring model that building is completed, consumption classification belonging to consumption data, the output information of word2vec model, that is, exportable Credit scoring realizes the purpose that automatic scoring is carried out using consumption data.

Description

A kind of credit-graded approach based on consumer demographics' portrait
Technical field
The present invention relates to credit evaluation technical field, specifically a kind of credit scoring based on consumer demographics' portrait Method.
Background technique
With the deep propulsion of the establishment of social credit system, the rapid development of social credibility standard working-out, relevant standard phase After publication, including the acquisition of credit services standard, credit data and service standard, credit repairing standard, city credit standard, industry Multi-level standards system including credit standard etc. is urgently put into effect, and social credibility standards system is expected to quickly propel.
Credit scoring refers to the credit history data according to client, using certain credit scoring model, obtains different etc. The credit score of grade.According to the credit score of client, the credit grade that providers of credit can analyze consumer is consumed with this to give The different degrees of discount of person and preferential.During credit scoring, most critical be exactly credit scoring model building.It is used to The model for generating credit scoring is too numerous to enumerate, each model has its unique rule.
The rule of development of the credit score of a usual people is: basis point obtained, as time goes by, has bonus point event, Occur with deduction event.Along with these events, credit score gradually goes up or declines.The basic principle of credit scoring model is It determines the factor for influencing Default Probability, then gives weight, calculate its credit score.
Traditional credit scoring is mainly measured with the dimension of the minority such as client's consuming capacity, it is difficult to comprehensive, objective, timely Reaction client credit.How to carry out intelligent scoring to client based on big data abundant is current problem.This patent from Business is set out, and based on the analysis to data, more powerful features is sufficiently excavated using word2vec, and construct reasonable Model realizes the intelligent scoring to client.
Based on this, a kind of credit-graded approach based on consumer demographics' portrait is researched and developed in design, by constructing reasonable letter With Rating Model, common carrier is facilitated to carry out intellectual analysis to the communication service of consumer, so to the credit of consumer into Row intelligent scoring.
Summary of the invention
The present invention for traditional credit scoring model because measured according to the dimension of the minority such as client's consuming capacity into And be difficult to it is comprehensive, objective, timely react customers' credit the shortcomings that, provide it is a kind of based on consumer demographics portrait credit scoring Method facilitates common carrier to score the credit of consumer by constructing reasonable credit scoring model.
A kind of credit-graded approach based on consumer demographics' portrait of the invention, solves the skill that above-mentioned technical problem uses Art scheme is as follows:
A kind of credit-graded approach based on consumer demographics' portrait, the credit-graded approach are acquired based on common carrier Consumption data, the consumption data of acquisition belong to same user unique identification be phone number;
The credit-graded approach the realization process includes:
According to unique identification, the consumption data for belonging to different user is stored in different set;
According to the feature of consumption data, the consumption data for being subordinated to the same set is divided into different consumption classifications;
Consumption classification belonging to the consumption data for being subordinated to the same set, consumption data is inputted in natural language processing Word2vec model, portray consumer demographics portrait;
Known credit scoring corresponding with acquisition consumption data is obtained, known credit scoring is based on, utilizes light GBM As base learner, study be subordinated to consumption classification belonging to the consumption data of the same set, consumption data and The output information of word2vec model constructs credit scoring model with training;
The credit scoring model of building receives the consumption data for being subordinated to the same set, different consumption classifications are included When the output information of consumption data and word2vec model, direct export credit is scored.
Optionally, involved consumption data includes 29 kinds of consumption information, specifically:
It is subscriber-coded;Whether user's system of real name passes through verification;Age of user;
Whether university student client;Whether blacklist client;Whether the unhealthy client of 4G;User's length of surfing the Net (moon);
User's the last time paid the fees away from modern duration (moon);It pays the fees user's the last time payment amount of money (member);
The nearly 6 monthly average consumption values (member) of user;Subscriber's account this month total cost (member);
User's this month account balance (member);Paying the fees, user currently whether pay the fees by arrearage;
User telephone fee susceptibility;This month call relationship cycle number;The people whether often to go shopping;
Nearly three months monthly market frequency of occurrence;Whether this month strolled market;
Whether this month arrived chartered shop;Whether this month watches movie;
It is of that month that whether sight spot is gone sight-seeing;It is of that month that whether stadiums are consumed;
Of that month online shopping class is using number;Of that month logistics express delivery class is using number;
Of that month finance and money management class is using total degree;Of that month video playback class is using number;
Of that month aircraft class is using number;Of that month train class is using number;
Of that month travel information class is using number.
Optionally, according to the feature of consumption data, 29 kinds of consumption information for being subordinated to the same set are divided into four Consume classification:
A) user go on a journey class, include: subscriber-coded, user's system of real name whether pass through verification, age of user, whether university student Client, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, of that month aircraft class are using number, of that month train Class is using number, of that month travel information class using number;
B) customer consumption class includes: subscriber-coded, user's system of real name whether pass through verification, age of user, whether university student Client, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, user's the last time pay the fees and use away from modern duration, payment Family the last time payment amount of money, the nearly 6 monthly average consumption values of user, subscriber's account this month total cost, user's this month account balance, Pay the fees user currently whether arrearage payment, user telephone fee susceptibility, of that month call relationship cycle number;
C) user's life kind includes: subscriber-coded, user's system of real name whether pass through verification, age of user, whether university student Client, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, the people whether often to go shopping, nearly three months it is monthly Whether market frequency of occurrence, this month strolled market, whether this month arrived whether chartered shop, this month watch movie, is of that month Whether sight spot is gone sight-seeing, of that month whether stadiums are consumed, of that month online shopping class makes using number, of that month logistics express delivery class application With number, of that month finance and money management class using total degree, of that month video playback class using number, of that month aircraft class application Access times, of that month train class are using number, of that month travel information class using number;
D) user's intersection information class includes: whether subscriber-coded, user's system of real name passes through verification, age of user, whether big Student client, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, of that month online shopping class are using number, this month Logistics express delivery class is using number, of that month finance and money management class using total degree, of that month video playback class using secondary Number, of that month aircraft class are using number, of that month train class using number, of that month travel information class using number.
Optionally, before portraying consumer demographics' portrait, the Skip- of word2vec model in natural language processing is utilized Consumption classification belonging to the consumption data for being subordinated to the same set, consumption data is projected to k dimension by Gram mode Vector space constitute sentence, excavate different user consumption habit and contract rule characterized user positioning;
In projection process, each word is indicated with a K dimensional vector, distance of the word similar in relationship in K dimension space It is same close, mathematic(al) representation such as:
P (s)=p (w1, w2..., wn)=Π p (wn|context)
Wherein behalf current sentence, w1Road wnFor the word for forming the sentence, Context is context p (wn|context) To there is this w in given contextnProbability.
Optionally, the concrete operations for constructing credit scoring model include:
1) high level critical point and low value critical point are set, relatively more known credit scoring, high level critical point, low value critical point three Person's size;
2) MSE loss function and MAE loss function are introduced;
3) known credit scoring be greater than high level critical point or be less than low value critical point when, using light GBM as Base learner, study be subordinated to the consumption data of the same set, the consumption data that different consumption classification is included and The output information of word2vec model introduces MSE loss function and constructs MSE credit scoring model;
4) when known credit scoring is between high level critical point and low value critical point, using light GBM as base Learner, study are subordinated to consumption classification and word2vec model belonging to the consumption data of the same set, consumption data Output information, introduce MAE loss function construct MSE credit scoring model;
5) MSE credit scoring model and MSE credit scoring model are constructed as credit scoring model, credit scoring mould The output information of type analysis word2vec model show that credit scoring, the credit scoring are MSE credit scoring model and MSE The average value of credit scoring model output scoring.
Specifically, involved high level critical point is 650 points, the low value critical point is 525 points.
Optionally, during constructing credit scoring model, it is based on known credit scoring, in MSE credit scoring model When output scoring and the deviation value of known credit scoring with MSE credit scoring model are more than ± 5%, in light GBM base The range and step-length that self-setting parameter is found in device are practised, is then carried out using the GridSearchCV parameter adjustment method in python The grid search of automation obtains optimal parameter, MSE credit scoring model and MSE credit scoring model.
Specifically, involved parameter includes learning rate learning_rate, leaf number num_leaves, depth capacity max_depth。
A kind of credit-graded approach based on consumer demographics' portrait of the invention, what is had compared with prior art is beneficial Effect is:
It is consumption data that credit-graded approach of the invention is acquired based on common carrier, corresponding with acquisition consumption data Known credit scoring, by word2vec model portray consumer demographics portrait, be based on known credit scoring, utilize light GBM as the study of base learner be subordinated to consumption classification belonging to the consumption data of the same set, consumption data and The output information of word2vec model constructs credit scoring model with training, finally, the credit scoring model completed using building Analysis is subordinated to the output of consumption classification and word2vec model belonging to the consumption data of the same set, consumption data Information, can export credit scoring, solve that existing grading system is not perfect, thinks to judge the strong defect of subjectivity, realize and utilize The purpose of consumption data progress automatic scoring.
Detailed description of the invention
Attached drawing 1 is the flow diagram of present invention building credit scoring model;
Attached drawing 2 is the flow diagram that credit scoring model of the invention scores.
Specific embodiment
The technical issues of to make technical solution of the present invention, solving and technical effect are more clearly understood, below in conjunction with tool Body embodiment carries out clear, complete description to technical solution of the present invention, it is clear that described embodiment is only this hair Bright a part of the embodiment, instead of all the embodiments.
Embodiment one:
In conjunction with attached drawing 1,2, the present embodiment proposes a kind of credit-graded approach based on consumer demographics' portrait, which comments The consumption data for dividing method to acquire based on common carrier, the unique identification that the consumption data of acquisition belongs to same user is mobile phone Number.
The credit-graded approach the realization process includes:
According to unique identification, the consumption data for belonging to different user is stored in different set;
According to the feature of consumption data, the consumption data for being subordinated to the same set is divided into different consumption classifications;
Consumption classification belonging to the consumption data for being subordinated to the same set, consumption data is inputted in natural language processing Word2vec model, portray consumer demographics portrait;
Known credit scoring corresponding with acquisition consumption data is obtained, known credit scoring is based on, utilizes light GBM As base learner, study be subordinated to consumption classification belonging to the consumption data of the same set, consumption data and The output information of word2vec model constructs credit scoring model with training;
The credit scoring model of building receives the consumption data for being subordinated to the same set, different consumption classifications are included When the output information of consumption data and word2vec model, direct export credit is scored.
In the present embodiment, involved consumption data includes 29 kinds of consumption information, specifically:
It is subscriber-coded;Whether user's system of real name passes through verification;Age of user;
Whether university student client;Whether blacklist client;Whether the unhealthy client of 4G;User's length of surfing the Net (moon);
User's the last time paid the fees away from modern duration (moon);It pays the fees user's the last time payment amount of money (member);
The nearly 6 monthly average consumption values (member) of user;Subscriber's account this month total cost (member);
User's this month account balance (member);Paying the fees, user currently whether pay the fees by arrearage;
User telephone fee susceptibility;This month call relationship cycle number;The people whether often to go shopping;
Nearly three months monthly market frequency of occurrence;Whether this month strolled market;
Whether this month arrived chartered shop;Whether this month watches movie;
It is of that month that whether sight spot is gone sight-seeing;It is of that month that whether stadiums are consumed;
Of that month online shopping class is using number;Of that month logistics express delivery class is using number;
Of that month finance and money management class is using total degree;Of that month video playback class is using number;
Of that month aircraft class is using number;Of that month train class is using number;
Of that month travel information class is using number.
In the present embodiment, according to the feature of consumption data, 29 kinds of consumption information for being subordinated to the same set are divided For four consumption classifications:
A) user go on a journey class, include: subscriber-coded, user's system of real name whether pass through verification, age of user, whether university student Client, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, of that month aircraft class are using number, of that month train Class is using number, of that month travel information class using number;
B) customer consumption class includes: subscriber-coded, user's system of real name whether pass through verification, age of user, whether university student Client, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, user's the last time pay the fees and use away from modern duration, payment Family the last time payment amount of money, the nearly 6 monthly average consumption values of user, subscriber's account this month total cost, user's this month account balance, Pay the fees user currently whether arrearage payment, user telephone fee susceptibility, of that month call relationship cycle number;
C) user's life kind includes: subscriber-coded, user's system of real name whether pass through verification, age of user, whether university student Client, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, the people whether often to go shopping, nearly three months it is monthly Whether market frequency of occurrence, this month strolled market, whether this month arrived whether chartered shop, this month watch movie, is of that month Whether sight spot is gone sight-seeing, of that month whether stadiums are consumed, of that month online shopping class makes using number, of that month logistics express delivery class application With number, of that month finance and money management class using total degree, of that month video playback class using number, of that month aircraft class application Access times, of that month train class are using number, of that month travel information class using number;
D) user's intersection information class includes: whether subscriber-coded, user's system of real name passes through verification, age of user, whether big Student client, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, of that month online shopping class are using number, this month Logistics express delivery class is using number, of that month finance and money management class using total degree, of that month video playback class using secondary Number, of that month aircraft class are using number, of that month train class using number, of that month travel information class using number.
In the present embodiment, before portraying consumer demographics' portrait, word2vec model in natural language processing is utilized Skip-Gram mode is projected using consumption classification belonging to the consumption data for being subordinated to the same set, consumption data as word The vector space tieed up to k constitutes sentence, excavates user's positioning that different user consumption habit and contract rule are characterized;
In projection process, each word is indicated with a K dimensional vector, distance of the word similar in relationship in K dimension space It is same close, mathematic(al) representation such as:
P (s)=p (w1, w2..., wn)=Π p (wn|context)
Wherein behalf current sentence, w1Road wnFor the word for forming the sentence, Context is context p (wn|context) To there is this w in given contextnProbability.
In the present embodiment, the concrete operations for constructing credit scoring model include:
1) high level critical point and low value critical point are set, involved high level critical point is 650 points, involved low value critical point It is 525 points, relatively more known credit scoring, high level critical point, low value critical point three's size;
2) MSE loss function and MAE loss function are introduced;
3) known credit scoring be greater than high level critical point or be less than low value critical point when, using light GBM as Base learner, study are subordinated to consumption classification and word2vec mould belonging to the consumption data of the same set, consumption data The output information of type introduces MSE loss function and constructs MSE credit scoring model;
4) when known credit scoring is between high level critical point and low value critical point, using light GBM as base Learner, study are subordinated to consumption classification and word2vec model belonging to the consumption data of the same set, consumption data Output information, introduce MAE loss function construct MSE credit scoring model;
5) MSE credit scoring model and MSE credit scoring model are constructed as credit scoring model, credit scoring mould The output information of type analysis word2vec model show that credit scoring, the credit scoring are MSE credit scoring model and MSE The average value of credit scoring model output scoring.
In the present embodiment, during constructing credit scoring model, it is based on known credit scoring, is commented in MSE credit When the output scoring of sub-model and MSE credit scoring model and the deviation value of known credit scoring are more than ± 5%:
The range and step-length that self-setting parameter is found in light GBM base learner, involved parameter include study Rate learning_rate, leaf number num_leaves, depth capacity max_depth;
The grid search automated using the GridSearchCV parameter adjustment method in python, obtain optimal parameter, MSE credit scoring model and MSE credit scoring model.
It is consumption data that the credit-graded approach of the present embodiment is acquired based on common carrier, opposite with acquisition consumption data The known credit scoring answered portrays consumer demographics' portrait by word2vec model, is based on known credit scoring, utilizes Light GBM as the study of base learner be subordinated to consumption classification belonging to the consumption data of the same set, consumption data, with And the output information of word2vec model, credit scoring model is constructed with training, finally, the credit scoring mould completed using building Type analysis is subordinated to the defeated of consumption classification belonging to the consumption data of the same set, consumption data and word2vec model Information can export credit scoring out.
In summary, using a kind of credit-graded approach based on consumer demographics' portrait of the invention, existing comment is solved Point system is not perfect, thinks to judge the strong defect of subjectivity, realizes the purpose using consumption data progress automatic scoring.
Use above specific case elaborates the principle of the present invention and embodiment, these embodiments are It is used to help understand core of the invention technology contents.Based on above-mentioned specific embodiment of the invention, the technology of the art Without departing from the principle of the present invention, any improvement and modification to made by the present invention should all fall into the present invention to personnel Scope of patent protection.

Claims (8)

1. a kind of credit-graded approach based on consumer demographics' portrait, which is characterized in that the credit-graded approach is based on communication The consumption data of operator's acquisition, the unique identification that the consumption data belongs to same user is phone number;
The credit-graded approach the realization process includes:
According to unique identification, the consumption data for belonging to different user is stored in different set;
According to the feature of consumption data, the consumption data for being subordinated to the same set is divided into different consumption classifications;
Consumption classification belonging to the consumption data for being subordinated to the same set, consumption data is inputted in natural language processing Word2vec model portrays consumer demographics' portrait;
Obtain known credit scoring corresponding with acquisition consumption data, based on known credit scoring, using light GBM as Base learner, study are subordinated to consumption classification and word2vec mould belonging to the consumption data of the same set, consumption data The output information of type constructs credit scoring model with training;
The consumption that the credit scoring model of building receives the consumption data for being subordinated to the same set, different consumption classifications are included When data and the output information of word2vec model, direct export credit is scored.
2. a kind of credit-graded approach based on consumer demographics' portrait according to claim 1, which is characterized in that described Consumption data includes 29 kinds of consumption information, specifically:
It is subscriber-coded;Whether user's system of real name passes through verification;Age of user;
Whether university student client;Whether blacklist client;Whether the unhealthy client of 4G;User's length of surfing the Net;
User's the last time pays the fees away from modern duration;Payment user's the last time payment amount of money;
The nearly 6 monthly average consumption values of user;Subscriber's account this month total cost;
User's this month account balance;Paying the fees, user currently whether pay the fees by arrearage;
User telephone fee susceptibility;This month call relationship cycle number;The people whether often to go shopping;
Nearly three months monthly market frequency of occurrence;Whether this month strolled market;
Whether this month arrived chartered shop;Whether this month watches movie;
It is of that month that whether sight spot is gone sight-seeing;It is of that month that whether stadiums are consumed;
Of that month online shopping class is using number;Of that month logistics express delivery class is using number;
Of that month finance and money management class is using total degree;Of that month video playback class is using number;
Of that month aircraft class is using number;Of that month train class is using number;
Of that month travel information class is using number.
3. a kind of credit-graded approach based on consumer demographics' portrait according to claim 2, which is characterized in that according to 29 kinds of consumption information for being subordinated to the same set are divided into four consumption classifications by the feature of consumption data:
A) user go on a journey class, include: subscriber-coded, user's system of real name whether pass through verification, age of user, whether university unfamiliar guest Family, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, of that month aircraft class are using number, of that month train class Using number, of that month travel information class using number;
B) customer consumption class includes: subscriber-coded, user's system of real name whether pass through verification, age of user, whether university unfamiliar guest Family, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, user's the last time pay the fees away from modern duration, payment user The last payment amount of money, subscriber's account this month total cost, user's this month account balance, is paid at the nearly 6 monthly average consumption values of user Expense family currently whether arrearage payment, user telephone fee susceptibility, of that month call relationship cycle number;
C) user's life kind includes: subscriber-coded, user's system of real name whether pass through verification, age of user, whether university unfamiliar guest Family, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, the people whether often to go shopping, nearly three months monthly quotient Frequency of occurrence, it is of that month whether strolled market, it is of that month whether arrived chartered shop, whether this month watches movie, this month is The visit of no sight spot, it is of that month whether stadiums consumption, of that month online shopping class using number, of that month logistics express delivery class using Number, of that month finance and money management class make using total degree, of that month video playback class using number, of that month aircraft class application With number, of that month train class using number, of that month travel information class using number;
D) user's intersection information class includes: subscriber-coded, user's system of real name whether pass through verification, age of user, whether university student Client, whether blacklist client, whether the unhealthy client of 4G, user's length of surfing the Net, of that month online shopping class are using number, of that month logistics Express delivery class using number, of that month finance and money management class using total degree, of that month video playback class using number, when Month aircraft class is using number, of that month train class using number, of that month travel information class using number.
4. a kind of credit-graded approach based on consumer demographics' portrait according to claim 1, which is characterized in that portray Before consumer demographics' portrait, using the Skip-Gram mode of word2vec model in natural language processing, it will be subordinated to same Consumption classification belonging to the consumption data of a set, consumption data constitutes sentence as the vector space that word projects to k dimension, digs User's positioning that pick different user consumption habit and contract rule are characterized;
In projection process, each word indicates that word similar in relationship is same in the distance of K dimension space with a K dimensional vector It is close, mathematic(al) representation such as:
P (s)=p (w1, w2..., wn)=Π p (wk|context)
Wherein behalf current sentence, w1Road wnFor the word for forming the sentence, Context is context p (wn| context) be There is this w when given contextnProbability.
5. a kind of credit-graded approach based on consumer demographics' portrait according to claim 1, which is characterized in that building The concrete operations of credit scoring model include:
1) high level critical point and low value critical point are set, relatively more known credit scoring, high level critical point, low value critical point three are big It is small;
2) MSE loss function and MAE loss function are introduced;
3) when known credit scoring is greater than high level critical point or is less than low value critical point, using light GBM as base Practise device, study is subordinated to consumption classification belonging to the consumption data of the same set, consumption data and word2vec model Output information introduces MSE loss function and constructs MSE credit scoring model;
4) when known credit scoring is between high level critical point and low value critical point, learnt using light GBM as base Device, study are subordinated to the defeated of consumption classification belonging to the consumption data of the same set, consumption data and word2vec model Information out introduces MAE loss function and constructs MSE credit scoring model;
5) MSE credit scoring model and MSE credit scoring model are constructed as credit scoring model, credit scoring model point The output information for analysing word2vec model, show that credit scoring, the credit scoring are MSE credit scoring model and MSE credit The average value of Rating Model output scoring.
6. a kind of credit-graded approach based on consumer demographics' portrait according to claim 5, which is characterized in that described High level critical point is 650 points, and the low value critical point is 525 points.
7. a kind of credit-graded approach based on consumer demographics' portrait stated according to claim 5, which is characterized in that constructing During credit scoring model, it is based on known credit scoring, in the defeated of MSE credit scoring model and MSE credit scoring model When the deviation value of scoring and known credit scoring is more than ± 5% out, self-setting parameter is found in light GBM base learner Range and step-length, the grid search then automated using the GridSearchCV parameter adjustment method in python obtained Optimal parameter, MSE credit scoring model and MSE credit scoring model.
8. a kind of credit-graded approach based on consumer demographics' portrait according to claim 6, which is characterized in that described Parameter includes learning rate learning_rate, leaf number num_leaves, depth capacity max_depth.
CN201910653033.9A 2019-07-19 2019-07-19 A kind of credit-graded approach based on consumer demographics' portrait Pending CN110363439A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910653033.9A CN110363439A (en) 2019-07-19 2019-07-19 A kind of credit-graded approach based on consumer demographics' portrait

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910653033.9A CN110363439A (en) 2019-07-19 2019-07-19 A kind of credit-graded approach based on consumer demographics' portrait

Publications (1)

Publication Number Publication Date
CN110363439A true CN110363439A (en) 2019-10-22

Family

ID=68220371

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910653033.9A Pending CN110363439A (en) 2019-07-19 2019-07-19 A kind of credit-graded approach based on consumer demographics' portrait

Country Status (1)

Country Link
CN (1) CN110363439A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110852796A (en) * 2019-10-31 2020-02-28 支付宝(杭州)信息技术有限公司 Position positioning method, device, medium and apparatus
CN110868409A (en) * 2019-11-08 2020-03-06 中国科学院信息工程研究所 Passive operating system identification method and system based on TCP/IP protocol stack fingerprint
CN111489055A (en) * 2020-03-16 2020-08-04 中国铁道科学研究院集团有限公司电子计算技术研究所 Passenger data processing method and device, storage medium and computer equipment
CN111931717A (en) * 2020-09-22 2020-11-13 平安科技(深圳)有限公司 Semantic and image recognition-based electrocardiogram information extraction method and device
CN112084242A (en) * 2020-09-02 2020-12-15 深圳市铭数信息有限公司 Consumption information display method, device, terminal and medium
CN111861174B (en) * 2020-07-09 2021-04-13 北京睿知图远科技有限公司 Credit assessment method for user portrait

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105243566A (en) * 2015-10-28 2016-01-13 联动优势科技有限公司 Method and apparatus for evaluating credit of users through different mobile phone number information from operators
CN107832306A (en) * 2017-11-28 2018-03-23 武汉大学 A kind of similar entities method for digging based on Doc2vec
CN108256993A (en) * 2017-12-29 2018-07-06 浪潮天元通信信息系统有限公司 A kind of credit score appraisal procedure and credit score Evaluation Platform
CN108921686A (en) * 2018-06-19 2018-11-30 阿里巴巴集团控股有限公司 A kind of credit-graded approach and device of personal user
CN109934623A (en) * 2019-02-26 2019-06-25 中山大学 Individual economy consuming capacity prediction technique based on user's APP usage behavior
CN110008338A (en) * 2019-03-04 2019-07-12 华南理工大学 A kind of electric business evaluation sentiment analysis method of fusion GAN and transfer learning

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105243566A (en) * 2015-10-28 2016-01-13 联动优势科技有限公司 Method and apparatus for evaluating credit of users through different mobile phone number information from operators
CN107832306A (en) * 2017-11-28 2018-03-23 武汉大学 A kind of similar entities method for digging based on Doc2vec
CN108256993A (en) * 2017-12-29 2018-07-06 浪潮天元通信信息系统有限公司 A kind of credit score appraisal procedure and credit score Evaluation Platform
CN108921686A (en) * 2018-06-19 2018-11-30 阿里巴巴集团控股有限公司 A kind of credit-graded approach and device of personal user
CN109934623A (en) * 2019-02-26 2019-06-25 中山大学 Individual economy consuming capacity prediction technique based on user's APP usage behavior
CN110008338A (en) * 2019-03-04 2019-07-12 华南理工大学 A kind of electric business evaluation sentiment analysis method of fusion GAN and transfer learning

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110852796A (en) * 2019-10-31 2020-02-28 支付宝(杭州)信息技术有限公司 Position positioning method, device, medium and apparatus
CN110852796B (en) * 2019-10-31 2021-03-16 支付宝(杭州)信息技术有限公司 Position positioning method, device, medium and apparatus
CN110868409A (en) * 2019-11-08 2020-03-06 中国科学院信息工程研究所 Passive operating system identification method and system based on TCP/IP protocol stack fingerprint
CN111489055A (en) * 2020-03-16 2020-08-04 中国铁道科学研究院集团有限公司电子计算技术研究所 Passenger data processing method and device, storage medium and computer equipment
CN111861174B (en) * 2020-07-09 2021-04-13 北京睿知图远科技有限公司 Credit assessment method for user portrait
CN112084242A (en) * 2020-09-02 2020-12-15 深圳市铭数信息有限公司 Consumption information display method, device, terminal and medium
CN111931717A (en) * 2020-09-22 2020-11-13 平安科技(深圳)有限公司 Semantic and image recognition-based electrocardiogram information extraction method and device
CN111931717B (en) * 2020-09-22 2021-01-26 平安科技(深圳)有限公司 Semantic and image recognition-based electrocardiogram information extraction method and device

Similar Documents

Publication Publication Date Title
CN110363439A (en) A kind of credit-graded approach based on consumer demographics' portrait
Joyce Strengthening skills: expert review of Australia’s vocational education and training system
Trilokekar et al. IMAGINE: Canada as a Leader in International Education. How Can Canada Benefit from the Australian Experience?.
CN110413746A (en) The method and device of intention assessment is carried out to customer problem
CN110414716A (en) A kind of enterprise based on LightGBM breaks one's promise probability forecasting method and system
CN108256098A (en) A kind of method and device of determining user comment Sentiment orientation
CN110363571A (en) The sensed in advance method and apparatus of trade user
McBurnie et al. Trends and future scenarios in programme and institution mobility across borders
CN108876076A (en) The personal credit methods of marking and device of data based on instruction
Zarra et al. Sustainability in the Age of Platforms
Gugler et al. Smart city strategy of dubai
CN109598607A (en) Method, apparatus and storage medium based on artificial intelligence monitoring self learning model
CN109685300A (en) Enterprise's big data application power index system and its construction method
Srivastava et al. Kickstart or jumpstart? Understanding women entrepreneurs’ crowdfunding performance
Guild The state, infrastructure and economic growth in Jokowi's first term
Bogdanova University business models in crisis management
Flyvbjerg et al. Report for the Edinburgh tram inquiry
Cuong Diversification of investment capital for road construction in Vietnam
Zhao Reshaping the Economic Cooperation Pattern of the Belt and Road Initiative
Steelberg Late Development and the Private Sector: A Perspective on Public-Private Partnerships in Vietnam
Fang et al. Evaluation index system of enterprise digital transformation based on balanced scorecard
Hou Value-for-Money assessment in China's public-private-parternship project present situation, problems and policy suggestions
Thilagaraj CUSTOMER SATISFACTION TOWARDS MOBILE BANKING-A STUDY
Giglio The future of energy transition between critical materials and geopolitics.
Tilak Current trends in private higher education in Asia

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: 20191022