CN107229948A - A kind of method for reducing customer churn on line based on customer problem forecast model - Google Patents
A kind of method for reducing customer churn on line based on customer problem forecast model Download PDFInfo
- Publication number
- CN107229948A CN107229948A CN201710355159.9A CN201710355159A CN107229948A CN 107229948 A CN107229948 A CN 107229948A CN 201710355159 A CN201710355159 A CN 201710355159A CN 107229948 A CN107229948 A CN 107229948A
- Authority
- CN
- China
- Prior art keywords
- mrow
- user
- data
- customer
- forecast model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- G06Q30/01—Customer relationship services
-
- 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
- G06Q40/02—Banking, e.g. interest calculation or account maintenance
Abstract
The present invention provides a kind of method for reducing customer churn on line based on customer problem forecast model, is related to technical field of data processing, including:S1:User enters the Internet bank page;S2:Whether judge with user with strong rule, such as hit strong rule, then go strong logic rules, then guesses the problem of client may meet TOPN;Do not hit strong rule such as, then into problem forecast model, the problem of conjecture client may meet TOPN;S3:In the problem of detecting TOPN whether alarm in need the problem of, if so, then initiating correspondence to client retrieves action, then carry out data and bury a little;If the problem of not needing alarm, directly carry out data and bury a little.The present invention calculates client the problem of may meet during real-time operation by problem forecast model, and pass through system real-time data collection, computing is carried out in real time, when client does not also propose problem, the true intention to user can just be guessed, the purpose for keeping client is reached, is conducive to lifting Consumer's Experience.
Description
Technical field
The present invention relates in technical field of data processing, more particularly to a kind of reduction line based on customer problem forecast model
The method of customer churn.
Background technology
With the development of bank's Internet service, the turnover rate of client is increasingly taken a fancy on line, to realize staying for user
Deposit, it is necessary to analyze the Drain Causes of client, the problem of finding user place can just make improvements.Found by As-Is analysis,
If guided according only to artificial experience by page official documents and correspondence to client, after user has been introduced into bank interface, do not have
Kit can go out what client may be met in operation according to the data real-time estimate such as user behavior, user profile
The model of problem, certain customers do not have found its problems faced in time in operation, and user, which leaves the page, can directly contribute visitor
The loss at family, such as be up to 97%, User logs in page business mistake rate is 94% in user's registration page turnover rate.Churn rate
Height, and client retrieves cost height, can only count customer revenue under line at present, be retrieved afterwards, can only also propose ask in client
The problem of just can know that client after topic, retrieves that cost is huge and produce little effect, and it is very passive to deal with.And calculated by model
Go out client to be transported the problem of during real-time operation may meet, and by system real-time data collection, and in real time
Calculate, correspondence is initiated to front and back end in real time and retrieves strategy, being capable of effectively customer retaining, and lift Consumer's Experience.
Being explained as follows for some technical terms used is needed herein:
CNN convolutional neural networks:CNN is a kind of multilayer neural network, based on artificial neural network, in artificial neural network
Before, feature extraction is carried out with wave filter, using convolution kernel as feature extractor, automatic training characteristics withdrawal device, that is volume
Product core and threshold parameter these be required for by network go study.
Cosine similarity model:Cosine similarity, also known as cosine similarity.By calculating two vectorial included angle cosines
Value assesses their similarity.
Problem forecast model:According to customer action, browse many data extraction features in track, historical data etc., training power
Weight, client is the problem of may meet for prediction, after be transmitted to rear and carry out corresponding measure.
The content of the invention
The purpose of the present invention is as follows:Solving existing Internet bank does not have kit can be according to user behavior, use
Family information data real-time estimate goes out the model for the problem of client may be met in operation, also not according to this model
Reduce the method for customer churn on line, cause customer churn and retrieve cost to produce little effect greatly, deal with very passive ask
Topic.The present invention provides a kind of method for reducing customer churn on line based on customer problem forecast model.
Technical scheme is as follows:
A kind of method for reducing customer churn on line based on customer problem forecast model, comprises the following steps:
S1:User enters the Internet bank page;
S2:Whether judge with user with strong rule, such as hit strong rule, then go strong logic rules, then guesses client
The problem of may meeting TOPN;Do not hit strong rule such as, then into problem forecast model, the problem of conjecture client may meet
TOPN;Strong rule is determined that such as blacklist client, white list user, VI P users etc., rule are such as summarized by business personnel
Property explain can say:Strong rule is determined by business personnel, on model, belongs to the strong factor that must be judged.
S3:Detect TOPN the problem of in whether alarm in need the problem of, if so, then to client initiate correspondence retrieve dynamic
Make, then carry out data and bury a little;If the problem of not needing alarm, directly carry out data and bury a little;Burying point data will be by machine
The mode of study is fed back in problem forecast model, carries out the processing of problem forecast model tuning;
S4:The sample of error be will determine that as negative sample, judicious sample passes through sample data as positive sample
The mode of machine learning feeds back a problem forecast model and carries out tuning processing;
S5:Terminate.
Specifically, the strong rule model is obtained by expert model and artificial experience.
Specifically, it is in S2, the step of problem forecast model:
S21:Sample data is gathered;
S22:Sample data is divided into positive sample data and negative sample data, the problem of user runs into actual environment and
The problem of bank matches is unanimously positive sample, and the problem of the problem of user runs into actual environment and bank match is inconsistent then
For negative sample;
S23:By CNN convolutional neural networks algorithms, the validity feature in positive sample data and negative sample data, life are extracted
In positive sample characteristic of division user characteristics, referred to as validity feature;Find out and be lost in letter entrained in user's sample data
Cease data and behavioural characteristic;
S24:Weight training, finds out correspondence validity feature classified weight;
S25:Input output model file;The result drawn according to weight training, is ranked up further according to prediction probability, takes forward
Customer problem the TOPN that predicts the outcome of problem may be met as user;After input output model file, model file is in reality
Model application in can produce new positive sample and negative sample again, to the new side for passing through machine learning in positive sample and negative sample
Formula is fed back in weight training.
Specifically, in S24, the method for weight training is svm classifier Forecasting Methodology, is comprised the following steps that:
S241:Behavior, operation history data are extracted according to user's browsing time dimension, user's operation dimension;Behavior, operation
The historical datas such as historical data is clicked on including page PV, user, user fills in, user profile;
S242:Cleaning behavior, operation history data, extract converting characteristic;Specifically, user's browsing time dimension is divided into
Multiple intervals, the data of each each interval interior dimension are divided into multiple intervals;This dimension of such as PV is in the page
Following characteristics can be formed under category1;
S243:By converting characteristic vectorization;The feature and the index of feature extracted according to step 2, each user is used
Characteristic vector is expressed as:F=(f1, f2, f3.., fn)
Wherein fi values are 0 or 1, and vectorial dimension represents the index of feature;
Specifically, in S24, weight training is realized by cosine similarity algorithm, is concretely comprised the following steps:
S241:Sample data is extracted, user behavior feature is calculated based on statistics, user's operating characteristics preference calculates public
Formula is as follows:
S242:User's each feature preferences combination is expressed as feature preferences vector:
v1=(p1, p2, p3..., pn)
S243:It is vector by customer problem character representation:
v2=(1,0,1 ..., 0)
The characteristic vector dimension values value of customer problem is 0 or 1;
S244:Calculate cosine similarity:
Wherein simi values are bigger represents more similar;
S245:Perform sequence;It is ranked up according to simi values, takes the problem of most like a part of user judges user,
And the weight sequencing that follows up.
Further, in S2, sample data includes the essential information, user's operation behavior, user service history rail of user
Mark, user credit information, page log data.
After such scheme, the beneficial effects of the present invention are:
(1) client is calculated the problem of may meet during real-time operation by problem forecast model, and led to
System real-time data collection is crossed, and carries out computing in real time, so as to when client does not also propose problem, can just guess to the true of user
Sincere figure, reaches the purpose for keeping client, is conducive to lifting Consumer's Experience.
(2) by the positive and negative sample data of accumulation, analysis badcase is trained and self study to model in time, to not
Same user, different periods dynamically change the efficiency factor of each user, and each Factor Weight, are allowed to adapt to current newest in real time
Situation.
Brief description of the drawings
Fig. 1 is overall flow figure of the invention;
Fig. 2 is the flow chart of this problem forecast model;
Fig. 3 is the flow chart of prior art.
Embodiment
All features disclosed in this specification, can be with any in addition to mutually exclusive feature and/or step
Mode is combined.
The present invention is elaborated below in conjunction with the accompanying drawings.
A kind of method for reducing customer churn on line based on customer problem forecast model, comprises the following steps:
S1:User enters the Internet bank page;
S2:Whether judge with user with strong rule, such as hit strong rule, then go strong logic rules, then guesses client
The problem of may meeting TOPN;Do not hit strong rule such as, then into problem forecast model, the problem of conjecture client may meet
TOPN;As strong rule is:When user is black list user, front end displaying black list user interface is pointed out in strong rule triggering, and
Strategy is retrieved in termination, now, such as user click on " modification of individual center-phone number " and when, it is possible that former phone number connects
It can not receive the problems such as short message, existing phone number do not receive short message;The strong rule model passes through expert model and artificial experience
Obtain.
S3:Detect TOP N the problem of in whether alarm in need the problem of, if so, then to client initiate correspondence retrieve
Act, then carry out data and bury a little;If the problem of not needing alarm, directly carry out data and bury a little;Burying point data will be by machine
The mode of device study is fed back in problem forecast model, carries out the processing of problem forecast model tuning;
S4:The sample of error be will determine that as negative sample, judicious sample passes through sample data as positive sample
The mode of machine learning feeds back a problem forecast model and carries out tuning processing;
S5:Terminate.
It is in S2, the step of problem forecast model:
S21:Sample data is gathered;
S22:Sample data is divided into positive sample data and negative sample data, the problem of user runs into actual environment and
The problem of bank matches is unanimously positive sample, and the problem of the problem of user runs into actual environment and bank match is inconsistent then
For negative sample;
S23:By CNN convolutional neural networks algorithms, the validity feature in positive sample data and negative sample data, meter are extracted
Details as Follows for calculation:
Correspondence validity feature can be extracted from sample data by calculating.X can enter Mobile state with the change of sample size
Adjustment.Find out and be lost in information data and behavioural characteristic entrained in user's sample data;
S24:Weight training, finds out correspondence validity feature classified weight;
S25:Input output model file;The result drawn according to weight training, is ranked up further according to prediction probability, takes forward
Customer problem the TOPN that predicts the outcome of problem may be met as user.
On the other hand, in S24, when the method for weight training is svm classifier Forecasting Methodology, comprise the following steps that:
S241:Behavior, operation history data are extracted according to user's browsing time dimension, user's operation dimension;Behavior, operation
The historical datas such as historical data is clicked on including page PV, user, user fills in, user profile;
S242:Cleaning behavior, operation history data, extract converting characteristic;Specifically, user's browsing time dimension is divided into
Multiple intervals, the data of each each interval interior dimension are divided into multiple intervals;This dimension of such as PV is in the page
Following characteristics can be formed under category1;
S243:By converting characteristic vectorization;The feature and the index of feature extracted according to step 2, each user is used
Characteristic vector is expressed as:F=(f1, f2, f3.., fn)
Wherein fi values are 0 or 1, and vectorial dimension represents the index of feature.
5. a kind of side for reducing customer churn on line based on customer problem forecast model according to claim 3
Method, it is characterised in that in S24, weight training is realized by cosine similarity algorithm, is concretely comprised the following steps:
S241:Sample data is extracted, user behavior feature is calculated based on statistics, user's operating characteristics preference calculates public
Formula is as follows:
S242:User's each feature preferences combination is expressed as feature preferences vector:
v1=(p1, p2, p3... pn)
S243:It is vector by customer problem character representation:
v2=(1,0,1 ..., 0)
The characteristic vector dimension values value of customer problem is 0 or 1;
S244:Calculate cosine similarity:
Wherein simi values are bigger represents more similar;
S245:Perform sequence;It is ranked up according to simi values, takes the problem of most like a part of user judges user,
And the weight sequencing that follows up.
In S2, sample data includes:
The essential information of user, such as " station address ", " address name ", " user identity card ", " user's registration time ",
" user account grade " etc..
User's operation behavior, such as " browser interface ", " interface residence time ", " browsing track ", " input time ", " defeated
Enter content ", " click track ", " click behavior " etc..
User service historical track, such as " number of times of being sent a telegram here in two hours ", " whether user is by self-service channel in 72 hours
Put question to ", " user whether sought help customer service " etc..
User credit information, such as " credit scoring of user's sesame ", " the white knight's score data of user ", " user I manage it letter
With score data ", " user whether black list user ", " user whether white list user " etc..
Page log data, such as " page PV ", " page UV ", " page report an error number of times ".
Other data, such as " what day is it today ", " Social Public Feelings information ", " what's date today ".
The flow of prior art is:(1) user initially enters;(2) user browses webpage;(3) user operates;(4) operate
During click option decide whether enter next link, if user enter next link, continue repeated pages browse operation,
Know that business handling finishes end;If user does not enter next link, flow is terminated after this link stays for some time.Can
See, user is entered after the page, the problem of bank may not meet to any customer issue is identified and taken phase
The measure for the reply answered, obtain visitor it is with high costs in the case of, user enter the page be costs of loss consumption greatly, certain customers
Originally it can convert and be lost in because real-time effective measures are not taken.
Claims (6)
1. a kind of method for reducing customer churn on line based on customer problem forecast model, it is characterised in that including following step
Suddenly:
S1:User enters the Internet bank page;
S2:Whether judge with user with strong rule, such as hit strong rule, then go strong logic rules, then guess that client may
The problem of meeting TOPN;Do not hit strong rule such as, then into problem forecast model, the problem of conjecture client may meet
TOPN;Strong rule is determined by business personnel, on model, belongs to the strong factor that must be judged;
S3:Detect TOPN the problem of in whether alarm in need the problem of, if so, then to client initiate correspondence retrieve action,
Data are carried out again to bury a little;If the problem of not needing alarm, directly carry out data and bury a little;Burying point data will be by machine learning
Mode feed back in problem forecast model, carry out the processing of problem forecast model tuning;
S4:The sample of error be will determine that as negative sample, sample data is passed through machine by judicious sample as positive sample
The mode of study feeds back a problem forecast model and carries out tuning processing;
S5:Terminate.
2. a kind of method for reducing customer churn on line based on customer problem forecast model according to claim 1, its
It is characterised by, the strong rule model is obtained by expert model and artificial experience.
3. a kind of method for reducing customer churn on line based on customer problem forecast model according to claim 1, its
It is characterised by, is in S2, the step of problem forecast model:
S21:Sample data is gathered;
S22:Sample data is divided into positive sample data and negative sample data, the problem of user runs into actual environment and bank
The problem of matching is unanimously positive sample, and inconsistent the problem of the problem of user runs into actual environment and bank match is then negative
Sample;The problem of the problem of bank matches occurs for the possibility that bank's customer service early stage is sorted out according to artificial experience;
S23:By CNN convolutional neural networks algorithms, the validity feature in positive sample data and negative sample data is extracted;Find out stream
Entrained information data and behavioural characteristic in the sample data of appraxia family;
S24:Weight training, finds out the corresponding validity feature classified weight in positive sample data and negative sample data;
S25:Input output model file;The result drawn according to weight training, is ranked up further according to prediction probability, takes forward use
Family problem may meet the TOPN that predicts the outcome of problem as user;
After input output model file, model file can produce new positive sample and negative sample again in actual model application, right
New is fed back in weight training in positive sample and negative sample by way of machine learning.
4. a kind of method for reducing customer churn on line based on customer problem forecast model according to claim 3, its
It is characterised by, in S24, the method for weight training is svm classifier Forecasting Methodology, is comprised the following steps that:
S241:Behavior, operation history data are extracted according to user's browsing time dimension, user's operation dimension;Behavior, operation history
Data are including page PV, user click on, user fills in, user profile historical data;
S242:Cleaning behavior, operation history data, extract converting characteristic;Specifically, user's browsing time dimension is divided into multiple
Interval, the data of each each interval interior dimension are divided into multiple intervals;
S243:By converting characteristic vectorization;The feature and the index of feature extracted according to step 2, by each user's feature
Vector representation is:F=(f1, f2, f3.., fn)
Wherein fi values are 0 or 1, and vectorial dimension represents the index of feature.
5. a kind of method for reducing customer churn on line based on customer problem forecast model according to claim 3, its
It is characterised by, in S24, weight training is realized by cosine similarity algorithm, is concretely comprised the following steps:
S241:Sample data is extracted, user behavior feature is calculated based on statistics, user's operating characteristics preference, calculation formula is such as
Under:
<mrow>
<mi>p</mi>
<mi>r</mi>
<mi>e</mi>
<mi>f</mi>
<mi>e</mi>
<mi>r</mi>
<mi>e</mi>
<mi>n</mi>
<mi>c</mi>
<mi>e</mi>
<mrow>
<mo>(</mo>
<mi>a</mi>
<mi>t</mi>
<mi>t</mi>
<mi>r</mi>
<mi>V</mi>
<mi>a</mi>
<mi>l</mi>
<mi>u</mi>
<mi>e</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mi>s</mi>
<mi>u</mi>
<mi>m</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>a</mi>
<mi>t</mi>
<mi>t</mi>
<mi>r</mi>
<mi>V</mi>
<mi>a</mi>
<mi>l</mi>
<mi>e</mi>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>s</mi>
<mi>u</mi>
<mi>m</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>a</mi>
<mi>t</mi>
<mi>t</mi>
<mi>r</mi>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
S242:User's each feature preferences combination is expressed as feature preferences vector:
v1=(p1, p2, p3..., pn)
S243:It is vector by customer problem character representation:
v2=(1,0,1 ..., 0)
The characteristic vector dimension values value of customer problem is 0 or 1;
S244:Calculate cosine similarity:
<mrow>
<mi>s</mi>
<mi>i</mi>
<mi>m</mi>
<mi>i</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>v</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msub>
<mi>v</mi>
<mn>1</mn>
</msub>
<mrow>
<mo>&lsqb;</mo>
<mi>i</mi>
<mo>&rsqb;</mo>
</mrow>
<mo>*</mo>
<msub>
<mi>v</mi>
<mn>2</mn>
</msub>
<mrow>
<mo>&lsqb;</mo>
<mi>i</mi>
<mo>&rsqb;</mo>
</mrow>
</mrow>
<mrow>
<msqrt>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>v</mi>
<mn>1</mn>
</msub>
<mrow>
<mo>&lsqb;</mo>
<mi>i</mi>
<mo>&rsqb;</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
<mo>*</mo>
<msqrt>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>v</mi>
<mn>2</mn>
</msub>
<mrow>
<mo>&lsqb;</mo>
<mi>i</mi>
<mo>&rsqb;</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
</mfrac>
</mrow>
Wherein simi values are bigger represents more similar, and simi values are weighted value;
S245:Perform sequence;It is ranked up according to simi values, the problem of taking most like a part of user judgement user, and with
Enter weight sequencing.
6. a kind of method for reducing customer churn on line based on customer problem forecast model according to claim 2, its
It is characterised by, in S2, essential information of the sample data including user, user's operation behavior, user service historical track, Yong Huxin
With information, page log data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710355159.9A CN107229948A (en) | 2017-05-19 | 2017-05-19 | A kind of method for reducing customer churn on line based on customer problem forecast model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710355159.9A CN107229948A (en) | 2017-05-19 | 2017-05-19 | A kind of method for reducing customer churn on line based on customer problem forecast model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107229948A true CN107229948A (en) | 2017-10-03 |
Family
ID=59933351
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710355159.9A Pending CN107229948A (en) | 2017-05-19 | 2017-05-19 | A kind of method for reducing customer churn on line based on customer problem forecast model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107229948A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107832581A (en) * | 2017-12-15 | 2018-03-23 | 百度在线网络技术(北京)有限公司 | Trend prediction method and device |
CN108509322A (en) * | 2018-01-16 | 2018-09-07 | 平安科技(深圳)有限公司 | Avoid the method excessively paid a return visit, electronic device and computer readable storage medium |
CN109657723A (en) * | 2018-12-20 | 2019-04-19 | 四川新网银行股份有限公司 | A method of enhancing higher-dimension category feature ability to express |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090006371A1 (en) * | 2007-06-29 | 2009-01-01 | Fuji Xerox Co., Ltd. | System and method for recommending information resources to user based on history of user's online activity |
CN102779193A (en) * | 2012-07-16 | 2012-11-14 | 哈尔滨工业大学 | Self-adaptive personalized information retrieval system and method |
CN102968416A (en) * | 2011-09-01 | 2013-03-13 | 佳能株式会社 | Device and method for identifying implemented recommendation based on user intent |
US9275119B2 (en) * | 2010-04-19 | 2016-03-01 | Facebook, Inc. | Sharing search queries on online social network |
CN106131203A (en) * | 2016-07-21 | 2016-11-16 | 四川易想电子商务有限公司 | A kind of automatic customer service system for electronic commerce |
-
2017
- 2017-05-19 CN CN201710355159.9A patent/CN107229948A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090006371A1 (en) * | 2007-06-29 | 2009-01-01 | Fuji Xerox Co., Ltd. | System and method for recommending information resources to user based on history of user's online activity |
US9275119B2 (en) * | 2010-04-19 | 2016-03-01 | Facebook, Inc. | Sharing search queries on online social network |
CN102968416A (en) * | 2011-09-01 | 2013-03-13 | 佳能株式会社 | Device and method for identifying implemented recommendation based on user intent |
CN102779193A (en) * | 2012-07-16 | 2012-11-14 | 哈尔滨工业大学 | Self-adaptive personalized information retrieval system and method |
CN106131203A (en) * | 2016-07-21 | 2016-11-16 | 四川易想电子商务有限公司 | A kind of automatic customer service system for electronic commerce |
Non-Patent Citations (2)
Title |
---|
吴增源: "《电子商务创业管理》", 30 September 2015 * |
李云飞: "服务失误及服务补救在网上银行信息安全产品服务中的应用研究", 《中国优秀硕士学位论文全文数据库 经济与管理科学辑》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107832581A (en) * | 2017-12-15 | 2018-03-23 | 百度在线网络技术(北京)有限公司 | Trend prediction method and device |
CN108509322A (en) * | 2018-01-16 | 2018-09-07 | 平安科技(深圳)有限公司 | Avoid the method excessively paid a return visit, electronic device and computer readable storage medium |
WO2019140738A1 (en) * | 2018-01-16 | 2019-07-25 | 平安科技(深圳)有限公司 | Method for avoiding excess return visits, and electronic apparatus and computer-readable storage medium |
CN108509322B (en) * | 2018-01-16 | 2020-05-12 | 平安科技(深圳)有限公司 | Method for avoiding excessive return visit, electronic device and computer readable storage medium |
CN109657723A (en) * | 2018-12-20 | 2019-04-19 | 四川新网银行股份有限公司 | A method of enhancing higher-dimension category feature ability to express |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101216825B (en) | Indexing key words extraction/ prediction method | |
CN103324745B (en) | Text garbage recognition methods and system based on Bayesian model | |
CN103077172B (en) | A kind of method and apparatus for excavating cheating user | |
CN103176982B (en) | The method and system that a kind of e-book is recommended | |
CN105095187A (en) | Search intention identification method and device | |
CN103793503A (en) | Opinion mining and classification method based on web texts | |
CN102622373B (en) | Statistic text classification system and statistic text classification method based on term frequency-inverse document frequency (TF*IDF) algorithm | |
CN103955547B (en) | Method and system for searching forum hot-posts | |
CN105138558B (en) | The real time individual information collecting method of content is accessed based on user | |
CN102170447A (en) | Method for detecting phishing webpage based on nearest neighbour and similarity measurement | |
CN103854064B (en) | Event occurrence risk prediction and early warning method targeted to specific zone | |
CN104391860A (en) | Content type detection method and device | |
CN109918560A (en) | A kind of answering method and device based on search engine | |
CN106547871A (en) | Method and apparatus is recalled based on the Search Results of neutral net | |
CN101763431A (en) | PL clustering method based on massive network public sentiment information | |
CN102033883A (en) | Method, device and system for improving data transmission speed of website | |
CN101820366A (en) | Pre-fetching-based phishing web page detection method | |
CN106355446B (en) | A kind of advertisement recommender system of network and mobile phone games | |
CN107870957A (en) | A kind of popular microblogging Forecasting Methodology based on information gain and BP neural network | |
CN106294590A (en) | A kind of social networks junk user filter method based on semi-supervised learning | |
CN103617235A (en) | Method and system for network navy account number identification based on particle swarm optimization | |
CN107229948A (en) | A kind of method for reducing customer churn on line based on customer problem forecast model | |
CN107341183A (en) | A kind of Website classification method based on darknet website comprehensive characteristics | |
CN102495860A (en) | Expert recommendation method based on language model | |
CN104636407B (en) | Parameter value training and searching request treating method and apparatus |
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: 20171003 |
|
RJ01 | Rejection of invention patent application after publication |