Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

Proposal

No description
by

on 23 February 2015

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of Proposal

To be continued
kdiem dkjoew nkld doem mleod keik keosl
idkue kdikx leosle keidk dkeid ledkekp kskem
kesle djleudmek kdiem dkjoew nkld
doem mleod keik keosl kded

Subtitle Text
kdiem dkjoew nkld doem mleod
keik keoslid

PROPOSAL
kdiem dkjoew nkld doem mleod
keik

METHOD
1.What am going to do it?
2.Why am i going to do it?
3.How am i going to do it?
4.When am i going to do it?

TOPIC
PROGRESS
Proposal
Topic
WHAT
WHY
HOW
WHEN
Personalized tourism e-commerce recommender system based on web data mining
in Kunming
Present a personalized tourism e-commerce recommender system based on web data mining
in Kunming
Communicate with consumers effectively and permanently
Provide information timely and accurately
Lasting relationship mechanism and online trading safely
Designed fully consider tourists various demands
Practical problem
Current e-business platform can’t keep pace with the upgrade of tourism information and rapid development in Kunming
More products bring difficulty for customers of finding fitting products
Sellers get puzzled on how when and what to provide targeted products and personalized service
Rebuild personalized tourism e- business recommendation system
Evaluation


Organize actively the information resources through the analysis of the users’ personality and habits

Creating a personalized information environment
Provide customers with the information services they may need
Cultivate the users’ personalized behavior

Guide their information needs

Promote the diversification

Diversified development of the society
Select more important and more appropriate information resources for them

Provide distinctive services for them based on the respect for individual users
E-commerce personalized service is in line with the demand of individual users, enterprises and the society under modern business, and it has the very high economic value to individuals, enterprises and the society.
Investagate practical problem to define a research problem
Study theoretical background

Investigate relevant research methods
Start to write a proposal to explain the research project and time

Examine ethical issues
Carry out more background reach to refine my research problem
Use suitable methods to collect and analyse data

Start writing the background to research
Carry out more detailed research

Describe how i used these methods during collection and anaylsis
Present a complete model as expected

Report my actions and results


Already started
Loading
Select research question
Select information sources
Choose search term
Apply practical screening criteria
Do the review
Synthesize the results
Tranditional Literature Review
Present a personalized tourism e-commerce recommender system based on web data mining in Kunming
Mostly from journals and published papers
Recommendation system, personalized, data mining, information overload, online tourism system,etc.
Chinese,English,information technology papers,etc.
Apply methodological screening criteria
Adequately described, analysis-questions clearly stated,etc.
Overview
Generalized recommendation techniques
content-based filtering
collaborative filtering
Hybrid recommender
light recommender
1.Competition among merchants is more intensive
2.Commodity information is increasing
3.Information overload causes the loss of e-commerce customers

personalized recommendation system
1.Expansion of personalized service in the internet
2.New field of application and development of personalized service
3.Provide differnet service mode and content to different customers
4.Provide service with higher quality

1.Interactive approach
2.Additional features base on data about general customers
3.Offer similar or corrleated items features to suggest items closely related to those in the shopping cart
4.Give the importance of users evaluation in selection
5.Inference techniques to elicit information about the user's preference to "quietly" acquire them
Enable the customer to specify hard and soft constraints to be satisfied by the solution,which can be criticized and refined in an interractive way
Features based on the items already bought by the customer, also informs other items purchesd by the customers who bought the same items
1.Built on the identification of customers which is similar to the current one
2.According to the suggestion of the items that were selected by such group of customers
3.Customers are acknowledged by the ranking generated by the whole customer as the basis
Filtering recommends items that have properties parallel to those of the products that the customer chosen before
pros and cons
Based on a category of items, either manually or autoatically derived from a description of their features,suitable items can be successfully recommended
1.Individual customer has to be under the supervision for a while
2.Prone to suggest items similar to one another that essentially at the expence of variety
3.Accuracy reduce if too many features need to be considered at the same time
pros and cons
Does not require details but the items
1.Not capable of handling items until they have got ranked by a minimum number of customers
2.Hard to classify customers having similar tastes to those of the current one
3.Comparison of rankes produced is very heavy if the web sites are cisited by millions of users during a short period
Content-based and collaborative filtering are jointly adopted to back high-quality recommendation.Its adopted when in the middle of preference acquistion process and dealing with new product.

Aim is to intergrate much more heterogeneous recommendation techniques with the information derived from the source information and link supplementary types
Developed to solve scalability concerns in heavy-loaded recommender system

Able to compute the similarity among items with the intention to recommend items similar to those the customer liked in the past

Its performance does not rely on the number of users accessing the web site, because this algorithm operates on items only
Development
1.Growing interest in personalization is mainly due to the increasing demand for customer-centric services
2.It can flexibly react to dynamically changing market requiements
3.A new perspective of CRM aims at improving customers' loyalty by managing a relationship building activity
4.It focused on companies getting to recognize, understand and ultimately serve their customers
1.CRM includes the provision of feature,but does not limit the offered services
2.Support the shortage and analysis of customers information
3.Support the design of new products matching market trends
4.Substantial reduction in overhead cost
5.Retailers keep in contact with customers by means of multiple channels (E-mail, web sites)
Customer lifetime value
Improving the quality of a recommendation to fulfill customers' need is important in fiercely competitive environments
From the perspective of niche marketing, all customers are not equal,even if they purchase identical products or sevices
Market segmentation is therefore necessary, firms are increasingly recognizing the importance of the CLV
Generally,
recency, frequency and monetary
(RFM) methods have been applied to cluster customers for niche marketing
Recommendation base on CLV
Although VARIOUS recommender system have been proposed
FEW have addressed the customer lifetime value to a e-commence
1.Combine group decision-making and data mining techniques
2.Apply analytic hierarchy process to determine the weights of RFM
3.Implement clustering techniques to group customers according to RFM value
5.Employ an association rule mining approach to provide product recommendation
CLV analysis and RFM evaluation
Market segmentation
Association rule mining
Association rule based recommendation
Customer lifetime valuse
:
A prediction of the net profit attributed to the entire future relationship with a customer
Identify profitable customers and develop strategies to target customers

Recency
: period since the last purchase
Lower value --> higher probability of making a repeat purchase
Frequency
: number of purchases made within a certain period
Higher frequency --> greater loyalty
Monetary
: the money spent during a certain period
Higher value -->company should focus more on that customer
1.Clustering seeks to maximize variance among group while minimizing variance within group

2.Use K-means method to group customers with similar CLV according to weighted RFM

3.K-means clusting is used to partition a set of data into group

4.Select m initial cluster centers and then iteratively refining them
1.Identifies associations among a set of product items frequently purchased together

2.Attempts to find association rules that satisfied minimum support and minimum confidence requirements

3.Indicates how frequently that rule applies to the data
Higher support --> stronger correlation between items

4.Used to find association rules by discovering frequent item sets
1.A customer transaction created to record all the products previouly purchased

2.Mining algorithm is applied to find all the recommendation rules that satisfy the given minimum support and minimum confidence constraints

3.The top-N products is recommended to a customer u
CLV is useful when applid
to e-commerce
1. Useful in helping the firms make strategic as well as tactical decisions

2. Help quantify the relationship of the firm with its customers

3. Subsequently allow the firm to make more informed decisions in a
structured framework

4. Help a firm to know who its profitable customers are

5. Customer profitability provides a metric for the allocation of marketing
resources to consumers and market segments
AHP
+
Clustering
+
Association rule-based methods
1. It clusters customers into segments according to
their lifetime value expressed in terms of weighted
RFM

2. Applying AHP to determine the relative importance
of RFM variables proved important

3. Clustering customers into different groups

4. Use association rule mining to provide reommendation
1. Improves the quality of recommendation

2. Helps decision-makers identify market segments
more clearly

3. Develop more effective strategies

4. Yield recommendations of higher quality

5. More effective for more loyal customers

Shortcomings
and
Opportnities

*Few sites are attempting to extract implicit negative ratings from purchase data
*Many recommender system algorithm perform better with both negative and positive rating, so the negative data can be valuable
*Only use a small subnet of the available information about the customer
*No system effectively uses all this data simultaneously for real-time recommendations
*Maybe individual recommender systems running on each type of data produce independent recommendations
*Recommender systems are currently used as cirtual salespeople, rather than as marketing tools
*Target each individual customer differnetly, making it difficult to produce the reports for marketing
*Need future work
*Current recommender systems are mainly "buy-side"
*Designed to work on behalf of the customer in deciding what products they should purchase
*But modern marketing requires to maximaize utility and the value to the business at the same time
*Recommender system could produce an indication of the price sensitivity of the customer for a given product
*So that E-commerce site could offer each product at the price that maximise the
lifetime value of the customer
to the site
*"sell-side" recommender systems could allow businesses to decide which clients to make special offers towards
Benefits to propose a recommender system based on CLV
1. Company will know how much they can afford to
spend in order to get the first sale, means
companies can take the short term risks necessary
to achieve long-term gain

2.Marketing managers can develop different offers
for different customers, based on the estimated
contribution of such customers in the future

3.Stock exchange analysts can use the company's
customer value to evaluate its worth to potential
investors

4.Customer value can serve as the basis for decisions
regarding new campaigns,allocation of retension

5.Customer value can be used to prioritize incoming
HTTP requests

6.IT resource allocation decisions can be based on
customer value
1. Good predictors for the probability of purchase in the next period

2. Can be created based on historical customer data
Amazon.com(book section)
Customer who bought
1. Structured with an information page for each book
2. Giving details of the text and purchase information
3. Two separate recommendation lists
a, Recommends books frequently purchased by
customers
b, Recommends authors whose books are frequently
purchased by customers
Eyes
1. Allows customers to be notified via email of new
items added to the Amazon.com catalog
2. Customers enter requests based upon author, title,
subject,ISBN,or publication date information
3. Customers can use both simple and more complex
Boolean-based criteria for notification queries
4. Requests can be directly entered from any search
result screen
5. Creates a persistent request based on the search

Delivers
1. A variation on the Eyes feature
2. Customers select checkboxes to choose from a list
of specific categories/genres
3. Send periodically email announcements to notify
subscribers of their latest recommendations
Book matcher
1. Allows customers to give direct feedback about books
2. Rate books on a 5-point scale
3. Non-rated texts are presented which correlate with
the user's indicated tastes
4. Feedback to these recommendations is provided
where customers can indicate a rating

Customer comments
1. Allows customers to receive text recommendations
baed on the opinions of customers
2. Located on the information page for each book is a
list of 1-5 star ratings
3. Provides the written comments and reviews
submitted
4. Customers have the opinion of incorporating these
recommendations into their purchase decision
CDNOW.com
Album advisor
1. In single album mode customer locate
the information page for a given album
2. System recommends 10 other albums
related to the album in question
3. In miltiple artist mode customers enter up
to three artists, the system recommends
10 albums related to the artists in question
My CDNOW
1. Enables customers to set up their own music store based
on albums and artists they like
2. Customers indicate which albums they own and which
artists are their favorites
3. Purchaes from CDNOW are entered automatically into
the "own it" list
4. System will predict 6 albums the customer might like
based eon what is already owned
5. Feedback opinion is avaiable by customers providing a
"onw it","move to wish","not for me" comment for any of
the albums in this prediction list
ebay.com
The feedback profile
1. Allow both buyers and sellers to contribute
to feedback profiles of other customers with
whom they have done business
2. Consist of a satisfaction rating as well as a
specific comment about the other customer
3. Consist of a table of the number of each
rating in the past as well as overall summary
4. Customers can browse the individual ratings
and comments for the sellers
Levis.com
Style finder
1. Allow customers to receive recommendations on
articles of Levi's clothing
2. Customers indicate their gender, then view three
categories "music","looks" and "fun", giving a rate
on a 7-point scale ranging
3. Once the minimum number of ratings are entered
customers may select "get recommendations"
4. Customers may provide feedback by using the
"tell us what you think feature"
5. Feedback may change one or all of the items
recommended
Moviefinder.com
Match maker
1. Allow customers to locate movies with a similar
"mood,theme,genre or cast"
2. Customers click on the "Match Maker" icon and
are provided with the list of recommended moives
3. Also get the links to other films by the original
film's director and key actor

We predict
1. Recommends movies based on customers' preciously
indicated interests
2. Customers enter a rating on a 5-point scale
3. Information page for non-rated movies contains a
personalized textual prediction
4. Customers can use "Powerfinder" to search for top
picks based on syntacitc criteria such as Genre,directors
5. Choose to have these sorted by their personalized
prediction or by the all customer average
Reel.com
Movie Matches
1. Provides recommendations on the information page for
each movie
2. Consist of "close match" and/or "creative match"
3. Each set consist of up to a dozen hyperlinks to the
information pages for each of these "matched" fims
4. The hyperlinks are annotated with one sentence
descriptions of how the new movie is similar to the
original movie in question
Movie map
1. Recommends movies based on syntactic feature
2. Customers enter queries based on movie types,
viewing format and/or prices
3. Request results be constrained to "sleepers" or "best
of this type"
4. the recommendations are editor's recommendations
for movies that fit the specified criteria
Non-Personalized Recommendatoins
1. Based on what other customers have said about
the product the products on average
2. Independent of the customers, each customer gets
the same recommendations
3. Automatic, require little customer effort to
generate the recommendation
4.Ephemeral, does not recongnize the customer from
one session to the next
5. Common in physical store, can be set up on a
display that is viewed without change by customers
1. The
average ratings display
by Amazon.com and
Moviefinder.com
2. Completely independent of the particular customer
targeted by the system
3.
Feedback profile
in eBay, give feedback on each
other, rather than on product
4. The average and individual feedback is available
for consideration by buyers and sellers
Attribute-Based Recommendations
1. Based on syntactic preperties of the product
2. Manual, customer must direstly request the
recommendation by entering desired syntacitic
product properties
3. Depend on whether the E-commerce site
remembers the attribute preference for a customer
1. The
Movie Map
by Reel.com entirely based on the
category of movie the customers select
2. Manual, customers must navigate to a category to
obtain a recommendation
3. Ephemeral, does not remember a customer's
interest from one visit to the next
4.
Delivers
by Amazon.com is manual, since
customers must explicitly sign up and provide a
set of interest categories
Item-to-Item Correlation
1. Based on a small set of product the customers
have expressed interest in
2. Usually ephemeral, do not need to know history
3. Only base on the product customers select
1. The
Customer Who Bought
by Amazon.com and

Movie Matches
by Reel.com,
Match Maker
by
Moviefinder.com
2. Suggest other products a customer might be
interested in
3. Based on a single other product that customer has
expressed interest in
4. Automatic and ephemeral, require neither action
form nor identification of the customer
People-to-People Correlation
1. Based on the correlation between that customer
and other cutomers
2. Close to Automatic, recommendations are
generated automatically by the system
3. Have to learn over time from customers
4. Manual, Based on customers' rating on products
5. Presistent, learn about patterns of agreement
between users requires substantial data
6. Could be ephemeral, if user session are long
enough
1. The
Book Matcher
by Amazon.com and
We Predict

by Moviefinder.com and
Style Finder
by Levi's.com
2. Persistent, based on users' explicitly rating on
products
3. Not fully automatic, ratings are entered only to get
the recommendation
User inputs
1. Recommendation technologies
requires some form of input upon
which to base the recommendations

2. Inputs can be provided by the
customers and sellers
Purchase data
:
procuts customers has purchased
(Amazon.com,My CDNOW)

Likert
:
what customers say they think of a product,
numeric or textual, must be ordered
(eBay.com, Levi's.com)

Text
:
written comments intended for other customers
to read, not interpreted by the computer system
(Amazon.com)

Editor's choice
:
selections within a category made by
hunman editors
(Reel.com,Moviefinder.com)
Literature
Review
Full transcript