CN108776931A - Financial client based on RFM and Canopy is worth loyalty divided method - Google Patents
Financial client based on RFM and Canopy is worth loyalty divided method Download PDFInfo
- Publication number
- CN108776931A CN108776931A CN201810328423.4A CN201810328423A CN108776931A CN 108776931 A CN108776931 A CN 108776931A CN 201810328423 A CN201810328423 A CN 201810328423A CN 108776931 A CN108776931 A CN 108776931A
- Authority
- CN
- China
- Prior art keywords
- client
- value
- loyalty
- canopy
- mixed type
- 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
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000009826 distribution Methods 0.000 claims abstract description 13
- 238000004458 analytical method Methods 0.000 claims abstract description 12
- 238000011156 evaluation Methods 0.000 claims description 21
- 230000011218 segmentation Effects 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 11
- 239000000820 nonprescription drug Substances 0.000 claims description 6
- 210000001520 comb Anatomy 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 description 5
- 230000004069 differentiation Effects 0.000 description 4
- 230000007306 turnover Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 206010003694 Atrophy Diseases 0.000 description 1
- 230000037444 atrophy Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000036461 convulsion Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- 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
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
-
- 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
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/06—Asset management; Financial planning or analysis
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Technology Law (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Operations Research (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses the divided methods that the financial client based on RFM and Canopy is worth loyalty, and the raw data table of loyalty is worth based on financial client, carry out initial data from several dimensions and detect.On the one hand the three-level index system in terms of client's absolute value, relative worth, stock exchange preference, assets distribution is combed.On the other hand, the analysis that customers buying behavior is carried out using RFM, situation of opening an account, nearly product buying behavior in 1 year and the fund outflow for combing client flow into situation.Then, whole input is formed after merging feature, using canopy algorithms, respectively obtains the cluster result of the value models cluster result of mixed type client and the value models of trade type client.Finally, standards of grading are worth according to loyalty, segment subdivision group of each of group and trade type client to each of mixed type client carries out value scoring, loyalty scoring respectively.
Description
Technical field
The present invention relates to financial technology fields, and more specifically, it relates to the financial client values based on RFM and Canopy
Loyalty divided method.
Background technology
In face of increasingly keen competition and open market competition environment, securities broker company should hard objectives segment market,
Formulate following differentiation developing direction.
It carries out differentiation strategy needs to find differences a little from the various aspects of client, finds out target market, it is therefore desirable to right
The case where various aspects of client, is analyzed, and is carried out to client by client's essential attribute, history buying behavior, risk partiality etc.
Subdivision, sorts out each client according to customer segmentation result, realizes personalized service, increases customer satisfaction degree, in turn
Targetedly to target group's customer service, large and complete client development and service are avoided, makes full use of the advantage of oneself to be
Customer service.
Invention content
In view of this, this application provides the divided method that the financial client based on RFM and Canopy is worth loyalty, with
It is horizontal to the fine-grained management of client to improve finance, differentiation financial service is preferably provided.
To achieve the above object, the present invention provides the following technical solutions:
The divided method of financial client value loyalty based on RFM and Canopy, including:
It is worth the raw data table of loyalty based on financial client, carries out initial data and detects and comb client's absolute valency
Three-level index system in terms of value, relative worth, stock exchange preference, assets distribution,
The analysis that customers buying behavior is carried out using RFM, combs the situation of opening an account of client, nearly product buying behavior in 1 year,
And fund outflow flows into situation,
Whole input is formed after merging feature, using canopy algorithms, the value models for respectively obtaining mixed type client are poly-
The cluster result of class result and the value models of trade type client;
Standards of grading are worth according to loyalty, each of group and trade type client are segmented to each of mixed type client
Subdivision group carries out value scoring, loyalty scoring respectively;
Above-mentioned transacting customer:Contribution rate non-zero, only stock exchange, nearly 1 year without purchase OTC products and public offering fund;
Mix client:Contribution rate non-zero, existing stock exchange buy OTC products or public offering fund in nearly 2 years again.
In a preferred embodiment of the invention, pass through average daily assets distribution, liveness, profit, region, age, property
One or more dimensions in not are carried out initial data and are detected.
In a preferred embodiment of the invention, described to be using RFM progress customers buying behavior analyses:
The time that each client's the last time buys product is calculated, and is ranked up according to sequence from the near to the distant, by it
Numerical values recited carry out decile, and according to etc. divide numerical value classified calculating to go out R;
The frequency of each client's purchase is calculated, and is ranked up according to sequence from high to low, is carried out by its numerical values recited
Decile, and according to etc. divide numerical value classified calculating to go out F;
It calculates each client and always buys the amount of money, and be ranked up according to sequence from high to low, carried out by its numerical values recited
Decile, and according to etc. divide numerical value classified calculating to go out M.
In a preferred embodiment of the invention, the value models cluster for obtaining mixed type client and trade type visitor
The cluster of the value models at family realizes that process is:
For the input of above-mentioned mixed type Customer clustering, the input of trade type Customer clustering, it is utilized respectively Canopy algorithms
It is clustered;
Obtain the cluster result of the value models cluster result of mixed type client and the value models of trade type client.
In a preferred embodiment of the invention, described to include according to loyalty value scoring process:
Comb the relevant assessment indicator system of value and the relevant assessment indicator system of loyalty of mixed type client;
The relevant assessment indicator system of value and the relevant evaluation index body of loyalty of trade type client mixed type client
System;
Determine the weight of the Indexes of Value Assessment of mixed type client and the weight of loyalty evaluation index;
Determine the weight of the Indexes of Value Assessment of trade type client and the weight of loyalty evaluation index;
The value score of mixed type customer segmentation group, loyalty score and trade type customer segmentation group is calculated
It is worth score, loyalty score.
In a preferred embodiment of the invention, the Canopy algorithms include:
The given point set S of traversal, is arranged two threshold values:T1, T2 and T1>T2;
A point is selected, calculates it at a distance from other centers Canpoy;
That Canopy is added in the point if distance is less than T1, the point will not become some if distance is less than T2
The center of Canopy;
Whole process is repeated, until S is sky.
In a preferred embodiment of the invention, initial random to select a point as the centers Canpoy, one
Canopy means a point of group, and the point for including in the same Canopy is apart from smaller, it is meant that these put more similar;
Conversely, if different points is distributed in different Canopy, gap is larger between meaning these points.
After adopting the above technical scheme, the invention has the advantages that:
Client is analyzed respectively according to difference the characteristics of mixed type and trade type, improves different clients group in difference
The tactful fining degree of financial product sale.
Meanwhile in embodiment disclosed by the embodiments of the present invention, the clustering algorithm based on Canopy is taken, it is effective to reduce
The range at initial classes centers, can greatly improve the speed of cluster.
Finally, in inventive embodiments disclosed embodiment, by applying the value and loyalty evaluation criterion of expertise,
Contribution level of the different type client for finance is more appreciably showed, thus effectively supports it special to customer care emphasis
The assurance of sign.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of invention for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
A kind of subdivision implementation flow chart for financial client value loyalty based on RFM and Canopy that Fig. 1 is;
Fig. 2 be it is a kind of utilize RFM carry out customers buying behavior analysis implementation flow chart;
The cluster of a kind of value models cluster for mixed type client that Fig. 3 is and the value models of trade type client realizes stream
Cheng Tu;
Fig. 4 be it is a kind of according to loyalty be worth scoring implementation flow chart;
A kind of Canopy algorithms implementation flow chart that Fig. 5 is;
Term " first ", " second ", " third " " the 4th " in specification and claims and above-mentioned attached drawing etc. (if
In the presence of) it is for distinguishing similar part, without being used to describe specific sequence or precedence.It should be appreciated that using in this way
Data can be interchanged in the appropriate case, so that embodiments herein described herein can be in addition to illustrating herein
Sequence in addition is implemented.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Finance in the present invention is suitable for multiple industries such as bank, insurance, stock trader.In the following, the present invention is using stock trader as in fact
It applies example and carries out expansion explanation.
In the embodiment of the present invention, mainly respectively to trade type client and mixed type client from two side of customer value and loyalty
Face synthesis carries out customer segmentation;
In the embodiment of the present invention, customer account data ends 2016.4.13, and duration of opening an account [0,30], data area covers
Client trading, asset data;Customer quantity is related to certain financial 4034389 client;The related data used in analytic process is
Customer transactional data, asset data.
Whether whether finance is contributed and be that trade type client is divided into following three classes according to existing customer:
1, silence client:Contribution rate (wash one's hands continue to pay dues+cease difference) is zero, and such client's contribution degree is zero, therefore is not considered
(1569580 people);
2, transacting customer:Contribution rate non-zero, only stock exchange, nearly 1 year without purchase OTC products and public offering fund;
(2319048 people);
3, client is mixed:Contribution rate non-zero, existing stock exchange buy OTC products or public offering fund in nearly 2 years again.
(145761 people).
Referring to Fig. 1, Fig. 1 is the subdivision implementation process that a kind of financial client provided in an embodiment of the present invention is worth loyalty
Scheme, may include:
Step S11:It is worth the raw data table of loyalty based on financial client, carries out initial data from several dimensions and visits
It looks into.
This step of illustrated in greater detail.The several dimensions mentioned in example of the present invention mainly have:The average daily assets distributions of A, B live
Jerk (stock exchange), C profits, the regions D, E ages, F genders etc..
Specific data exploration content is as shown in following several tables:
The average daily assets distribution situations of A (the average daily assets accounting highest of wholesale of mixed type client)
Total assets | Silence type | Trade type | Mixed type |
0-10W | 99.51% | 78.24% | 36.26% |
10-50W | 0.31% | 16.46% | 37.15% |
50-100W | 0.08% | 2.91% | 12.17% |
100W | 0.09% | 2.39% | 14.43% |
B liveness (stock exchange situation)
Always enliven number | Silence type | Trade type | Mixed type |
0 | 100.00% | 31.49% | 14.46% |
0-1 | 0.00% | 3.18% | 6.38% |
1-5 | 0.00% | 12.87% | 17.18% |
5-50 | 0.00% | 40.12% | 50.21% |
50+ | 0.00% | 12.35% | 11.78% |
C profit situations
Profit situation | Silence type | Trade type | Mixed type |
Less than 0 | 0.63% | 51.84% | 57.65% |
0 | 91.85% | 0.00% | 0.00% |
More than 0 | 7.52% | 48.16% | 42.35% |
D Regional Distribution situations
Such as upper table, some provinces and cities' silence client's accountings are relatively low, and wherein Shanghai, Sichuan are most apparent;Some provinces and cities' silences
Client's accounting is relatively high.
E age distribution situations
F Sex distribution situations
Gender | Silence type | Trade type | Mixed type |
Man | 38.94% | 56.71% | 4.35% |
Female | 37.98% | 56.46% | 5.56% |
Step S12:Comb the three-level index in terms of client's absolute value, relative worth, stock exchange preference, assets distribution
System.
This step of illustrated in greater detail.The absolute value mentioned in example of the present invention, relative worth, stock exchange preference,
Assets distribution system is specific as follows:
A absolute values.The absolute value of client is mainly reflected in client in the contribution degree of sales department, the contribution degree of client
The service charge that namely sales department collects.Using client contribution commission income as criteria for classification, then it can be found that it is traditional according to
There is suitable portion in the client that the mode classification of customer capital amount is unfavorable for is objective, completely reflection client contributes situation capital quantity big
The client's commission divided, contribution are very low.Obviously, merely with clients fund amount number distinguish client, obviously lack science
Foundation.This period commission volume of client's contribution degree=client
B relative worths.The relative worth of client is mainly reflected on client trading liveness.Service charge contribution can conduct
The criterion of client's absolute value, but since the asset size of client is very big apart, so also to investigate the opposite valence of client
Value, that is, client potential value.Relative worth is embodied with the trading capacity of the brisk trade degree of client and client, client
Active degree can be defined from asset turnover, the trading capacity of client is then investment yield index.If the receipts of client
The very low even negative value of beneficial rate, it is meant that his transaction does not bring income, then his trading volume will atrophy.It is only positive
Earning rate can just ensure that the transaction of client is grown, so two indices are combined the relative worth that can preferably judge client.
The calculation formula of the two indexs is as follows:Asset turnover=this period total volume/this period average daily assets, investment yield
=(end of term assets-originally assets-fund net inflow)/beginning assets.
C client's stock exchange preference.Customer priorities index includes stock index, transaction count, average transaction value, stock
Number reflects the single investment of customer priorities or diversification investments, and transaction count reflects client's stock exchange frequency, average to merchandise
Price reflects customer priorities low price stocks or stock at a high price.
D customer capitals are distributed.Its index includes total assets, position ratio, stock position ratio, reflects the assets point of client
Cloth situation.
Step S13:The analysis that customers buying behavior is carried out using RFM combs situation of opening an account, the purchase of nearly 1 year product of client
It buys behavior and fund outflow flows into situation.
This step of illustrated in greater detail.The RFM basic thoughts mentioned in example of the present invention are by 3 important client's rows
For index, i.e., nearest time buying R (Recency), purchase frequency F (Frequency) and total purchase amount of money M
(MonetaryValue) judge the value of client.
Mentioned in example of the present invention RFM analysis based on the assumption that:
1) possibility for having the client of buying behavior to buy again recently is higher than recently the client of not buying behavior.
2) the high client of the purchase frequency client low compared with purchase frequency more likely buys the product of enterprise again.
3) possibility that always the purchase higher client of the amount of money buys again is higher and is the higher client of value.
Step S14:Whole input is formed after merging feature, using canopy algorithms, respectively obtains the valence of mixed type client
It is worth the cluster result of Model tying result and the value models of trade type client.
When doing Canopy clusterings, selection is three-level index, as directly inputting, need not determine the power of index
Weight.Wherein for firsts and seconds index only to determine three-level index, tactful thinking is to obtain step by step.
When doing loyalty value scoring, it is true according to financial business expertise to need the weight of setting target, weight
It is fixed to arrive.
This step of illustrated in greater detail.Feature after the merging mentioned in example of the present invention is as shown in the table:+
The cluster mentioned in example of the present invention is exactly according to some specific criteria (such as distance criterion) a data set point
It is cut into different classes or cluster so that the similitude of the data object in the same cluster is as big as possible, while not in the same cluster
Data object otherness it is also as large as possible.Of a sort data are brought together as far as possible after clustering, different data
It detaches as possible.
Step S15:Standards of grading are worth according to loyalty, group and trade type visitor are segmented to each of mixed type client
Each of family subdivision group carries out value scoring, loyalty scoring respectively.
This step of illustrated in greater detail.The mixed type value scoring mentioned in example of the present invention and loyalty Score index are deposited
In difference.The scoring of trade type value and loyalty Score index have differences.Referring to Fig. 2, Fig. 2 be it is a kind of using RFM into
Row customers buying behavior analyzes implementation flow chart, may include:
Step S011:The time that each client's the last time buys product is calculated, and is carried out according to sequence from the near to the distant
Sequence.By its numerical values recited carry out decile, and according to etc. divide numerical value classified calculating to go out R.
This step of illustrated in greater detail.The time that each client's the last time buys product is inquired, and according to from the near to the distant
Sequence be ranked up, entire customers are divided into 5 deciles after sequence, the time buying client label nearest from current time
For " 5 ", the time buying client farthest from current time is labeled as " 1 ".
Step S012:The frequency of each client's purchase is calculated, and is ranked up according to sequence from high to low.By its numerical value
Size carry out decile, and according to etc. divide numerical value classified calculating to go out F.
This step of illustrated in greater detail.The F calculations mentioned in example of the present invention are according to the method similar to R, to all
The sequence of the frequency that client buys according to it from high to low is ranked up, and entire customers are divided into 5 deciles after sequence, is bought
The high client of frequency be labeled as " 5 ", the client that the frequency of purchase is low is labeled as " 1 ".
Step S013:It calculates each client and always buys the amount of money, and be ranked up according to sequence from high to low.By its numerical value
Size carry out decile, and according to etc. divide numerical value classified calculating to go out M.
This step of illustrated in greater detail.The M calculations mentioned in example of the present invention are according to the method similar to R, to all
Client is ranked up according to its sequence of total purchase amount of money from high to low, and entire customers are divided into 5 deciles after sequence, total to purchase
It buys the high client of the amount of money and is labeled as " 5 ", it is total to buy the low client of the amount of money labeled as " 1 ".
It, can be important into row index according to the concrete numerical value of these three indexs and other indexs in the analytic process of cluster
The judgement of property.
Referring to Fig. 3, the value models of the value models cluster and trade type client for a kind of mixed type client that Fig. 3 is
Implementation flow chart is clustered, may include:
Step S031:Comb the relevant remittance of value relevant the collecting index system and trade type client of mixed type client
Overall performane system.
This step of illustrated in greater detail.The relevant collecting index body of value of the mixed type client mentioned in example of the present invention
System see the table below:
The relevant collecting index system of value of the trade type client mentioned in example of the present invention see the table below:
Index | Index meaning |
Asset_ratio | Assets position ratio |
Avg_stkratio | Stock position ratio |
Avgeff_stkratio | The effective position ratio of stock |
avgprice | Average transaction value |
contri | Contribution degree |
opentime | It opens an account duration |
Outflow_ratio | Fund flows out ratio |
Rate_return | Investment yield |
Stk_num | Stock number |
Stk_trade_num | Transaction count |
totalavgasset | Average daily assets |
totalvelocity | Turnover rate |
Step S032:For the input of above-mentioned mixed type Customer clustering, the input of trade type Customer clustering, it is utilized respectively
Canopy algorithms are clustered.
Step S033:Obtain the cluster of the value models cluster result of mixed type client and the value models of trade type client
As a result.
This step of illustrated in greater detail.The value models cluster result of the mixed type client mentioned in example of the present invention is as follows
Shown in table:
The numerical value is an example data set of customer segmentation;Numerical value in table therein can be according to cluster input pointer
Variation, the variation of clustering method occur different as a result, concrete numerical value result difference, can have differences the deciphering of data.Table
The subsequent word of lattice is the explanation to the meaning of example data set.
Go out the following from the value models cluster centre point analysis of the above mixed type client:
2,6,12 average daily stock position ratios of cluster are relatively low, this three classes contribution degree is relatively low, belongs to product type client;
It is higher to cluster 1,9,10 average daily stock position ratios, cash class product proportion maximum is bought in product, much greatly
In other a few class products;Wherein, orange group's profitability is relatively low;
It is relatively uniform (stock, cash class finance product, public offering fund etc.) to cluster 4 patterns of investment, profitability is best;
Average daily stock position ratio>=60% group's purchase equity class finance product possibility is smaller;Bought power
Group's contribution degree of beneficial class finance product is also maximum;
1,10 product preferences of cluster are cash class product, are the cluster groups of cash class product relative to other product preferences
For body, the investment yield of this two class is negative, it is contemplated that this two types of populations increase throw care for subscribed services, with promotion he
Equity investment ability.
The value models cluster result of the trade type client mentioned in example of the present invention is as shown in the table:
Go out the following from the value models cluster centre point analysis of the above trade type client:
It is relatively low to cluster 1,10 average daily stock position ratios, belongs to silence type client;
It is higher to cluster 4,12 average daily stock position ratios, up to 80%;Wherein, orange group's profitability is relatively low;
It is best to cluster 5 return on investment abilities;
Referring to Fig. 4, Fig. 4 be it is a kind of according to loyalty be worth scoring implementation flow chart, may include:
Step S41:Comb the relevant assessment indicator system of value and the relevant evaluation index of loyalty of mixed type client
System.
This step of illustrated in greater detail.Mentioned in example of the present invention mixed type customer value evaluation index (
Specific evaluation index and segment logic are determined according to business expertise, wherein the index as user is fallen
On section, then the numerical value of user's index is subsequent specific score.
As the original totalavgasset=90w of party A-subscriber then will between section 800,000 to 1,000,000
Totalavgasset updates are set to 64.) such as following table:
The mixed type customer loyalty evaluation index such as following table mentioned in example of the present invention:
Step S42:The relevant assessment indicator system of value and the relevant assessment indicator system of loyalty of trade type client.
The trade type customer value evaluation index such as following table mentioned in example of the present invention:
The trade type customer loyalty evaluation index such as following table mentioned in example of the present invention:
Wherein pro_RFM (tri- numbers of R, F, M are spliced to form, such as R=2, F=5, M=1, then pro_RFM
=251)
Step S43:Determine the weight of the Indexes of Value Assessment of mixed type client and the weight of loyalty evaluation index.
This step of illustrated in greater detail.The mixed type customer value evaluation index and weight mentioned in example of the present invention are such as
Shown in lower:;
The mixed type customer value evaluation index and weight mentioned in example of the present invention are as follows:
Value=0.6*totalavgasset (average daily assets) score value+0.4*contri (contribution degree) score value;
The mixed type type customer loyalty evaluation index and weight mentioned in example of the present invention are as follows:Loyalty=
0.4*maxdategap (maximum consumption interval) score value+0.3*opentime (duration of opening an account) score value+0.3*pro_num (purchases
Product quantity) score value;
Weight in formula is according to the empirically determined of business expert.
Step S44:Determine the weight of the Indexes of Value Assessment of trade type client and the weight of loyalty evaluation index.
This step of illustrated in greater detail.The trade type customer value evaluation index and weight mentioned in example of the present invention are as follows
It is shown:Value=0.3*totalavgasset score value+0.3*contri score values+0.2*rate_return (investment yield) points
Value+0.2*velocity (turnover rate) score value;
The trade type type customer loyalty evaluation index and weight mentioned in example of the present invention are as follows:Loyalty=
0.3*opentime score values+0.3*outflow_ratio (fund outflow ratio) score value+0.4*pro_RFM (OTC& public offering bases
Golden class product RFM values) score value;
Step S45:The value score, loyalty score and trade type client that mixed type customer segmentation group is calculated are thin
Divide value score, the loyalty score of group.
This step of illustrated in greater detail.The value score of the mixed type customer segmentation group mentioned in example of the present invention is as follows
Table:
The loyalty score such as following table of the mixed type customer segmentation group mentioned in example of the present invention:
The value score such as following table of the mixed type customer segmentation group mentioned in example of the present invention:
The loyalty score such as following table of the mixed type customer segmentation group mentioned in example of the present invention:
Referring to Fig. 5, the wherein purpose of Canopy is to go out different groups to subscriber segmentation, evaluation is in Canopy
On the basis of the result of cluster, the score value of scoring is further refined.The rear benefit refined first rough in this way is exactly both can be in difference
Subdivision group between compare score value, can also see the difference of user's score value under same subdivision group.
A kind of Canopy algorithms implementation flow chart that Fig. 5 is, step 5 are the explanations to the clustering method in step S032.
Wherein T1, T2 are determined according to experimental method, usually provide several groups of T1, T2.Original T1 is judged according to the effect of cluster
It is suitable when with T2 being what numerical value.May include:
Step S51:The given point set S of traversal, is arranged two threshold values:T1, T2 and T1>T2;
This step of illustrated in greater detail.The point set S mentioned in example of the present invention is list, T1 and the T2 of a vectorization
Value can be determined with cross check;
Step S52:A point is selected, it and the other centers Canpoy (centers Canpoy are calculated
It is initial random to select a point as the centers Canpoy, it is then arbitrary again to select other remaining points, calculate separately this
The distance of a little points and the initial centers Canopy.In conjunction with T1, T2, according to distance, it is new to judge whether these left points form
The centers Canopy) distance;
This step of illustrated in greater detail.The point mentioned in example of the present invention is any one point in list, and it is (former to calculate point P
Beginning data include N number of point, if it is the centers Canopy to specify some point, one is arbitrarily chosen in remaining N-1 point is
P points, P points can recycle to be selected successively.) with the distance between all Canopy (if there is currently no Canopy, point P
As a Canopy).
Step S53:That Canopy is added in the point if distance is less than T1, not less than the point if T2 if distance
The center of some Canopy can be become;
One Canopy means a point of group, and the point for including in the same Canopy is apart from smaller, it is meant that this
A little points are more similar;Conversely, if different points is distributed in different Canopy, gap is larger between meaning these points.
This step of illustrated in greater detail.It is mentioned in example of the present invention, it, then will point P if fruit dot P and some Canopy distance are within T1
It is added to this Canopy;If fruit dot P is once at a distance from some Canopy within T2, then point P is needed to delete from list
It removes, this step is that think that point P has been reached with this Canopy at this time close, therefore it cannot again be in other Canopy
The heart;
Step S54:Whole process is repeated, until S is sky.
This step of illustrated in greater detail.It is mentioned in example of the present invention, repeats step S52, S53, until list is empty ties
Beam.
Main problem of the model mentioned in example of the present invention encountered in building process:
1, the customers partition feature of cluster result and differentiation unobvious.
2, customer quantity difference of all categories in cluster result is excessive.
For the above problem, the solution taken in example of the present invention:
1, model parameter, such as cluster number are adjusted.The center point feature of all categories from new operation post analysis.
2, it attempts to replace the index into model so that each index has preferable discrimination.
3, it selectes a certain index to be divided, be clustered from newly for each division.
4, model evaluation.Accuracy, the applicability of Result are assessed using various technical indicators, verification model performance is
It is no to meet expection.
Cluster result appraisal procedure and standard:First, each group size of observation:Pass through the size of each customer group, analysis
The distribution situation of cluster.The difference for clustering each group should not be too large or disproportionate.If there is such situation, then cope with too small
Population analysis center point feature and carry out stepping statistics, observe population characteristic, filtering should give for exceptional value, then from newly into
Row cluster.Second is that the average value for each index that observation each clusters:If each finger target value is all clearly distinguished, can be with industry
It is explained in business, then shows that Clustering Effect is preferable.
Each embodiment is described by the way of progressive in this specification, the highlights of each of the examples are with other
The difference of embodiment, just to refer each other for identical similar portion between each embodiment.For device disclosed in embodiment
For, since it is corresponded to the methods disclosed in the examples, so description is fairly simple, related place is said referring to method part
It is bright.
The step of method described in conjunction with the examples disclosed in this document or algorithm, can directly be held with hardware, processor
The combination of capable software module or the two is implemented.Software module can be placed in random access memory (RAM), memory, read-only deposit
Reservoir (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technology
In any other form of storage medium well known in field.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention.
Various modifications to these embodiments will be apparent to those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the invention
It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one
The widest range caused.
Claims (7)
1. the divided method of the financial client value loyalty based on RFM and Canopy, which is characterized in that including:
It is worth the raw data table of loyalty based on financial client, carries out initial data and detects and comb client's absolute value, phase
To the three-level index system in terms of value, stock exchange preference, assets distribution,
The analysis that customers buying behavior is carried out using RFM, combs the situation of opening an account of client, nearly product buying behavior in 1 year, and
Fund outflow flows into situation,
Whole input is formed after merging feature, using canopy algorithms, respectively obtains the value models cluster knot of mixed type client
The cluster result of the value models of fruit and trade type client;
Standards of grading are worth according to loyalty, group and each subdivision of trade type client are segmented to each of mixed type client
Group carries out value scoring, loyalty scoring respectively;
Above-mentioned transacting customer:Contribution rate non-zero, only stock exchange, nearly 1 year without purchase OTC products and public offering fund;
Mix client:Contribution rate non-zero, existing stock exchange buy OTC products or public offering fund in nearly 2 years again.
2. according to the method described in claim 1, it is characterized in that, passing through average daily assets distribution, liveness, profit, region, year
One or more dimensions in age, gender are carried out initial data and are detected.
3. according to the method described in claim 1, it is characterized in that, described be using RFM progress customers buying behavior analyses:
The time that each client's the last time buys product is calculated, and is ranked up according to sequence from the near to the distant, by its numerical value
Size carry out decile, and according to etc. divide numerical value classified calculating to go out R;
The frequency of each client's purchase is calculated, and is ranked up according to sequence from high to low, decile is carried out by its numerical values recited,
And according to etc. divide numerical value classified calculating to go out F;
It calculates each client and always buys the amount of money, and be ranked up according to sequence from high to low, decile is carried out by its numerical values recited,
And according to etc. divide numerical value classified calculating to go out M2.
4. according to the method described in claim 1, it is characterized in that, the value models cluster for obtaining mixed type client and friendship
The cluster of the value models of easy type client realizes that process is:
For the input of above-mentioned mixed type Customer clustering, the input of trade type Customer clustering, it is utilized respectively the progress of Canopy algorithms
Cluster;
Obtain the cluster result of the value models cluster result of mixed type client and the value models of trade type client.
5. according to the method described in claim 1, it is characterized in that, described include according to loyalty value scoring process:
Comb the relevant assessment indicator system of value and the relevant assessment indicator system of loyalty of mixed type client;
The relevant assessment indicator system of value and the relevant assessment indicator system of loyalty of trade type client mixed type client;
Determine the weight of the Indexes of Value Assessment of mixed type client and the weight of loyalty evaluation index;
Determine the weight of the Indexes of Value Assessment of trade type client and the weight of loyalty evaluation index;
The value score of mixed type customer segmentation group, the value of loyalty score and trade type customer segmentation group is calculated
Score, loyalty score.
6. according to the method described in claim 3, it is characterized in that, the Canopy algorithms include:
The given point set S of traversal, is arranged two threshold values:T1, T2 and T1>T2;
A point is selected, calculates it at a distance from other centers Canpoy;
That Canopy is added in the point if distance is less than T1, the point will not become some if distance is less than T2
The center of Canopy;
Whole process is repeated, until S is sky.
7. according to the method described in claim 6, it is characterized in that, initial random select a point as the centers Canpoy, one
A Canopy means a point of group, and the point for including in the same Canopy is apart from smaller, it is meant that these points compare phase
Seemingly;Conversely, if different points is distributed in different Canopy, gap is larger between meaning these points.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810328423.4A CN108776931A (en) | 2018-04-13 | 2018-04-13 | Financial client based on RFM and Canopy is worth loyalty divided method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810328423.4A CN108776931A (en) | 2018-04-13 | 2018-04-13 | Financial client based on RFM and Canopy is worth loyalty divided method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108776931A true CN108776931A (en) | 2018-11-09 |
Family
ID=64033716
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810328423.4A Pending CN108776931A (en) | 2018-04-13 | 2018-04-13 | Financial client based on RFM and Canopy is worth loyalty divided method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108776931A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110147965A (en) * | 2019-05-27 | 2019-08-20 | 安徽省鼎众金融信息咨询服务有限公司 | A kind of customer analysis system based on big data |
CN111091282A (en) * | 2019-12-10 | 2020-05-01 | 焦点科技股份有限公司 | Customer loyalty segmentation method based on user behavior data |
CN111242573A (en) * | 2020-01-08 | 2020-06-05 | 江苏智谋科技有限公司 | Customer relationship management system based on big data and knowledge management |
CN112529628A (en) * | 2020-12-16 | 2021-03-19 | 平安科技(深圳)有限公司 | Client label generation method and device, computer equipment and storage medium |
CN113298442A (en) * | 2021-06-28 | 2021-08-24 | 中国平安人寿保险股份有限公司 | Cluster analysis method, device and equipment for customer data and storage medium |
CN113407827A (en) * | 2021-06-11 | 2021-09-17 | 广州三七极创网络科技有限公司 | Information recommendation method, device, equipment and medium based on user value classification |
CN113743826A (en) * | 2021-09-18 | 2021-12-03 | 中国银行股份有限公司 | Customer classification method and device |
WO2022095864A1 (en) * | 2020-11-05 | 2022-05-12 | 西安邮电大学 | E-commerce platform customer segmentation method based on weighted rfm model |
-
2018
- 2018-04-13 CN CN201810328423.4A patent/CN108776931A/en active Pending
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110147965A (en) * | 2019-05-27 | 2019-08-20 | 安徽省鼎众金融信息咨询服务有限公司 | A kind of customer analysis system based on big data |
CN111091282A (en) * | 2019-12-10 | 2020-05-01 | 焦点科技股份有限公司 | Customer loyalty segmentation method based on user behavior data |
CN111091282B (en) * | 2019-12-10 | 2022-07-22 | 焦点科技股份有限公司 | Customer loyalty segmentation method based on user behavior data |
CN111242573A (en) * | 2020-01-08 | 2020-06-05 | 江苏智谋科技有限公司 | Customer relationship management system based on big data and knowledge management |
WO2022095864A1 (en) * | 2020-11-05 | 2022-05-12 | 西安邮电大学 | E-commerce platform customer segmentation method based on weighted rfm model |
CN112529628A (en) * | 2020-12-16 | 2021-03-19 | 平安科技(深圳)有限公司 | Client label generation method and device, computer equipment and storage medium |
CN112529628B (en) * | 2020-12-16 | 2024-04-09 | 平安科技(深圳)有限公司 | Client label generation method and device, computer equipment and storage medium |
CN113407827A (en) * | 2021-06-11 | 2021-09-17 | 广州三七极创网络科技有限公司 | Information recommendation method, device, equipment and medium based on user value classification |
CN113298442A (en) * | 2021-06-28 | 2021-08-24 | 中国平安人寿保险股份有限公司 | Cluster analysis method, device and equipment for customer data and storage medium |
CN113743826A (en) * | 2021-09-18 | 2021-12-03 | 中国银行股份有限公司 | Customer classification method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108776931A (en) | Financial client based on RFM and Canopy is worth loyalty divided method | |
Gunnarsson et al. | Household financial strategies in Sweden: An exploratory study | |
Chen | Classifying credit ratings for Asian banks using integrating feature selection and the CPDA-based rough sets approach | |
Nguyen et al. | Promotion of fintech application for the modernization of banking-finance system in Vietnam | |
US20080215470A1 (en) | Methods and apparatus for use in association with payment card accounts | |
US20070050288A1 (en) | System and method for integrating risk and marketing objectives for making credit offers | |
JP2003526146A (en) | Method and system for reducing risk by obtaining evaluation values | |
CN109858947A (en) | Retail user value analysis system and method | |
Naveenan | Warning Signals-A Tool to Control NPA in Banks | |
Yorke | The definition of market segments for banking services | |
Vong et al. | Mobility technology solutions can reduce interest rates of microfinance loans | |
Srivastava et al. | Data mining for credit card risk analysis: a Review | |
Zernova et al. | The factors determining credit policy of commercial bank | |
KR20020085499A (en) | Financial Assets Management System and Financial Assets Management Method | |
Clarke | Using finance for competitive advantage | |
Dash et al. | Customer attrition analytics in banking | |
KR20090001456A (en) | Investment advice offer method and investment advice offer server | |
Clarkson et al. | Competitive Strategies in Banking Services-Implications of Segmentation Analysis of Consumers by Age Groups | |
Ibrahım et al. | Impact of investors sentiment on IPO performance: Evidence from NASDAQ and NYSE | |
Chambers et al. | US futures exchanges as nonprofit entities | |
Shah et al. | Proposing Investments Based on Fundamental Analysis | |
Zekić-Sušac et al. | Selecting neural network architecture for investment profitability predictions | |
Zhao et al. | Application of fractal dimension of fractional Brownian motion to supply chain financing and operational comprehensive decision-making | |
Albanis et al. | Five classification algorithms to predict high performance stocks | |
Dassatti et al. | Loan evergreening through banks’ lenses evidence from credit product-level data |
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 |
Application publication date: 20181109 |