Project Documentation - Njit

aspiringtokΤεχνίτη Νοημοσύνη και Ρομποτική

15 Οκτ 2013 (πριν από 3 χρόνια και 10 μήνες)

96 εμφανίσεις

Project Delivery for IS687


Sightseeing



1




Sightseeing

-

A solution approach for making recommendation for NY tourists




Editors:

Joachim Braun

Jerry Lin

Ibrahim George
-
Sankoh

Arya
Thachenkary

Naitik Shah

Nelson Mercado


Project Delivery for IS687


Sightseeing



Statement of Authorship

I declare that this document and the accompanying code have
been composed by myself, and describe my own work, unless
otherwise acknowledged in the text. It has not been accepted in
any previous application for a degree. All verbati
m extracts
have been distinguished by quotation marks, and all sources of
information have been specifically acknowledged.


____________________________________________________________________




Project Delivery for IS687


Sightseeing



3

Table of Contents

Sightseeing
................................
................................
................................
................................
.................

1

Table of Contents

................................
................................
................................
................................
....

3

Table of Figures

................................
................................
................................
................................
.......

4

1. Introduction

................................
................................
................................
................................
.....

5

Problem description

................................
................................
................................
.....................

5

Solution Approach

................................
................................
................................
........................

5

2. System Design and Prototype

................................
................................
................................
...

7

Home Page

................................
................................
................................
................................
.......

9

Process Description

................................
................................
................................
...................

13

3. Advanced Information Retrieval

................................
................................
..........................

16

Which techniques can be used to gather more user data

................................
...........

16

How are these data collated

................................
................................
................................
...

16

Personality types

................................
................................
................................
........................

17

The generalizing of personality types

................................
................................
................

17

4. Database Design (Ibrahim + Naitik + Joachim)

................................
...............................

19

Cardinality constrains and relationship mapping

................................
.........................

20

Entities and attributes.

................................
................................
................................
.............

20

5. Machine Learning Design (Joachim)

................................
................................
...................

22

Suggestion Model without
user Data

................................
................................
..................

22

Suggestion Model with user Data

................................
................................
.........................

25

Results

................................
................................
................................
................................
............

30

6. Business Benefits

................................
................................
................................
........................

31

Business Model

................................
................................
................................
............................

31

Investment

................................
................................
................................
................................
....

32

Potential

................................
................................
................................
................................
.........

32

Roadmap

................................
................................
................................
................................
........

32

Project Managemen
t

................................
................................
................................
.................

33

Risk Management

................................
................................
................................
.......................

34

SWOT Analysis

................................
................................
................................
............................

35

8. Conclusion

................................
................................
................................
................................
.....

36

Appendix
................................
................................
................................
................................
.............

38

Bibliography
................................
................................
................................
................................
...........

49


Project Delivery for IS687


Sightseeing



Table of
Figures

Figure 1: Decision Tree for categorization of HTTP Header information

......................

23

Figure 2: Structure of the trainingset of the Decision Tree

................................
.................

23

Figure 3: Training set for the Decision Tree

................................
................................
..............

24

Figure 4: The Machine Learning process of the Decision Tree

................................
..........

25

Figure 5: Results of the machine learning

................................
................................
..................

25

Figure 6: User Table

................................
................................
................................
............................

26

Figure 7: Machine Learn Process for peer evaluation

................................
...........................

26

Figure 8: ML_Userclusters table

................................
................................
................................
.....

27

Figure 9

-

Machine Learning process for classification

of User Data using personality


types

................................
................................
................................
................................
......

28

Figure 10

-

Training Set
for Machine Learning


ML_TrainingsSetClassificationPersonaltyType

................................
..................

29

Figure 11
-

Result table of the machine learning process
-

ML_Userclassificat
ion

....

29


Project Delivery for IS687


Sightseeing



5

1
. Introduction

Problem description

Tourists might find it difficult to explore foreign countries without a tour guide. To
see places of their interes
t, to have
things, which suit their disposition,

and to
participate in events which amuse their curiosity cannot be achieved in an alien
place without the help of an expensive guided tour. It is difficult to find places
which satisfy our appetite
and that

suit our budget constraints.

The probability to go
astray in a strange place will be high.
This project address
es

the problems described
above and
helps
to
urists to

visit places on their own with the assistance of a web
app.

Solution Approach














This is primarily focused on tourists to provide the best route to v
isit all of
their places of interest

in that location using GPS technology and machine
learning.



The app will have the list of tourist spots, their descriptions, related
destinations, photos videos and the ratings of that place.

Places of Interest

Real time feedback to the customer


Best Route

Personalized suggestions




Sightseeing App

Project Delivery for IS687


Sightseeing





Suggestions are provided based on the personality type

analyze
d through the
survey
. It provides real time information by suggesting places conforming to
the user’s interests if the user answers the survey questionnaire.



The user can search places, which assent to their budget, or places most
ranked by other users.



While they are travelling from one location to another, suggestions of places
in proximity such as restaurants or events in tune with their individual taste
will be recommended.



Tourists can plan trips. Special offers, packages and Highlights of the
special
events of a particular month in that location will be shown.



The web app also shows souvenirs of a particular location and places to buy
them.



The users who register in the site are asked a series of survey questionnaire.



If the user doesn’t prov
ide any information an alternate method of analyzing
the user by collecting information from their http

readers is deployed.















Project Delivery for IS687


Sightseeing



7

2. System Design

and Prototype


In the following section our system design will be explained in depth, and you will

learn the different devices the user are able to use to access our service, how these
device connect and send data to our servers, and where the data is stored and the
process through which it is put through, before our application delivers results.




















User:

There are no restrictions as to which our public can be. So all users with
the
ability to utilize a personal computer, mobile device, or a tablet will be able and are
welcomed to use Sightseeing.

Server

User

Users

Third Party Vendors

Database

Internet

Computer

Mobile Phone

Machine Learning

Project Delivery for IS687


Sightseeing




Devices:

at the beginning stages our app will support the more popular devices
such as, Devices with IOS, Android, and other devices that can access the internet
will be supported as well.


Gateway:

The internet will be our main gateway to send and receive data to

and
from our users.


Servers:

Our servers will be cloud based, this will help us easily manage traffic
spikes by expanding and downsizing efficiently. In the servers will also be stored
our database, and the different applications execute our application.


Third party vendors: Our Application will deliver results to our registered
companies, which will have a live feed of our different users what they are
accessing, and the suggestions that we are providing to them, allowing the Third
party vendors to bid
on advertisement.


As depicted in the diagram:

Once the user has provided their information, the

data is sent to our servers,

through
the internet

where we will

populate our databases and

process their data

through our
cluster analysis and data mining mod
el.

Once the data has been classified,

The results
are sa
ved to our databases, Information about the user and the final result are
sent to
third party vendors

allowing the opportunity to Bid On advertisement, and or coupons
which will be integrated and sen
t back to the user.


This process does not only happen once; as the user is moving our servers are receiving
their GPS location constantly, and depending where the user is, our third party vendors
will be up
-
to
-
date, and giving them the option to once agai
n bid on advertisement.


Project Delivery for IS687


Sightseeing



9

Home Page







There are two ways of accessing our content and services:

User Log
-
in, or web surfing.


First time users will be greeted with a Sign Up page, which will also serve

as a log in
page.


If User is new to the system they will be prompted to provide the following
information


First time Users will be required to provide:




First Name



Last Name



Date of Birth

(this will help us determine the users age, and classify them and
help us provided a more optimal result)



Location

(Where the person resides at, and/ or where they plan to visit)



Income

(They’re gross income, this information is utilized to give our third
-
party vendors a better picture of who the user is)


Log
-
in fields

Sign Up!

Venue fields

Project Delivery for IS687


Sightseeing









Optional Information:



Recommendation for Friends Emails they would like to recommend



Interest In Music



Preferred food



General Interest


With this initial
Information, we can populate our Databases with these attributes
helping us classify this user, and the more the user utilizes the Sight Seeing
Application, our model will be learning more and more about this User.











New user


Sign Up fields

Project Delivery for IS687


Sightseeing



11

Survey Page:








A
fter the user has filled up the registration details, there is a short survey consisting of
four questions where the user has to select one option. This survey helps the model to
analyze the personality of the user.

This is a one time survey for every user. This survey
appears only during the registration for a new user and the responses are saved in the
database.













User
selects either of
the two options.

Project Delivery for IS687


Sightseeing



Venue Search

and Suggestions

Page:







Web Surfer

For Users who do not wish t
o sign up to our service,

but wi
ll like to use it;
we will
collec
t information from HTTP headers. F
rom the
HTTP

header we can collect data
such as:


1.

What device are they’r
e requesting our service from (iPa
d, iP
hone, Home
Computer)

2.

IP address

3.

Geo location

4.

Language

5.

Cookies

6.

Time


With this Information we can determine broadly the user

s interest with data that
we have put through the machine based on their location and cookies


When a user access our
s
ervice through
their home computer and no wish to log
-
in,
just with the HTTP

Header we can determine
their
geo
-
location and decipher where
this person is connecting from.

Empire State
Building

Times
Square

Suggestions

Project Delivery for IS687


Sightseeing



13

An example would be that an anonymous user logs in from Brazil, from the
HTTP

header the application can determine the users location, the application can
translate the page to they’re local language, and based on this minimal information
our system can determine that
there

is a high probability that this person is
Brazilian, Det
ermining this helps us provide content more relevant to the type of
person that the web surfer is, opposed to just showing everybody the same content.


Process
Description

Once t
he user has opened up the Sight
seeing App from
their i
Phone,
iPad, Android
pho
ne e
t
c
, they will be prompted to the following:


In the first phase it is clear to see that the user has input Nmercado as their user
name and their password, in the next step, they select the different activities that
they would like to participate in an
d
the interests
.



After they’re information has be collected, it will be introduced into our model, and
the data clustering
and analysis
process will be executed.



Nmercado


Password

Login


Food


Outdoor
activity

Select
activities


Hiking


Music


Adventure

Interests

Project Delivery for IS687


Sightseeing




Our Model will then output what would be considered the activities that are most
appealing to the user base on their data.




After rece
i
ving the results, our system will display the different activities / venues,
that our Model, finds that will suit the user the most. The app will give the

user
various options, they c
an select from.


Having selected the activities that the user wants, the app proceeds to calculate and
output the most optimal route for the activities selected.


User Data

::::Data Clustering::::

Activity
output

Based On the user data
and the new info provided
our app will deliver
activities that would
appeal to the users


Gino Pizza


Piccolos's rest.


frank and Joes
Pizza


Mountain
Climbing


Bike Riding


kayaking

Display

Output of
activities


Piccolo's Rest.


Mountain
climbing


Kayaking

Select
activities

Project Delivery for IS687


Sightseeing



15



After the data (activities selected) has been processed , the results are delivered
back to the user, as a

optimal route to the activities the user has selected.


Conclusion

Our app works seamlessly and effortlessly for the user, Our app is design to
continuously learn more and more from the user, based on they’re, selections,
GPS

locations, and they’re intere
st, encouraging the user to use the app more and more,
to help it learn the different tendencies of the user, and optimize the results with
every use of the Sightseeing Application.


Activities
Selected

With GPS Location
and user defined
settings, our model
will pick the best
route for the activities

Generated
route

Route For the three
activities and general
Information (distance
time)

Project Delivery for IS687


Sightseeing



3. Advanced Information Retrieval


Which techniques can be used to
gather more user data



The users who register in the site are asked a
series of
survey questionnaire.
Myers Briggs Type Indicator

is used to analyze personality and thereby the
user’s preferences. The

Myers
-
Briggs Type Indicator

(
MBTI
) assessment is
a

psychometric

questionnaire designed to
measure

psychological

preferences in how people perceive the world and make decisions.



A
decision tree

is used to analyze the answers provided and to explore
about the personality of that user.



Personality types

are

categorized into 4 different clusters.



K
-
means
Clustering

a type of unsupervised learning is used to provide
personalized suggestions to the users.



Suggestions are provided based on the personality type. It provides real time
information by suggesting pla
ces conforming to the user’s interests if the
user answers the survey questionnaire.



If the user doesn’t provide any information an alternate method of analyzing
the user by collecting information from their http

readers is deployed. With
this data the loc
ation and the cookies data can be tracked. Decision tree is
again used to give a personalized suggestion based on the information
retrieved from their http readers.


How are these data collated



Personality
types are denoted with a four letter MBTI(
Myers
-
Br
iggs Type
Indicator

).
CPP Inc., the publisher of the MBTI instrument, calls it "the
world's most widely used personality assessment",

with as many as two
million assessments administered annually. The CPP and other proponents
state that the indicator meet
s or exceeds the reliability of other psychological
instruments

and cite reports of individual behavior.



The questions asked during the survey analyze the
psychological

preferences
. Each question measures the cognitive function of
an

individual. It checks whether an individual is Extrovert/Introvert,
Project Delivery for IS687


Sightseeing



17

Sensual/ Intuitive, Thinker/ feeler and Judger/ Perceiver.



A decision tree is used to determine the answers and
to explore about the
personality of that user.



They are then assigned an MBT
I tag that corresponds to their personality
type.



Personality types are clustered using Clustering, a type of unsupervised
learning according to their preferences.



Clustering is done to 4 categories

traditionalists, experiencers, idealists and
conceptuali
sts.

Personality types

The project use
s

the MBTI to

classify pe
ople into specific groups. The

decision
of

user personality type

is

determin
ed within four questions. The questions will be
answered in a sequential order. The following table illustrates the p
ossible answers
on

each question (0
-
Option1 / 1
-
Option 2
) and the MBTI code. Furthermore the
right attributes a generic type related to the MBTI code:


Extrovert/

Introvert

Sensual/

Intuitive

Thinker/

Feeler

Judger/

Perceiver

MBTI
code

Personality

T
ag

0

0

0

0

ESTJ

The Guardian

0

0

0

1

ESTP

The Adventurer

0

0

1

0

ESFJ

The Custodian

0

0

1

1

ESFP

The Entertainer

0

1

0

0

ENTJ

Field Marshal

0

1

0

1

ENTP

The Visionary

0

1

1

0

ENFJ

The Mentor

0

1

1

1

ENFP

The Inspirer

1

0

0

0

ISTJ

Go
-
Getter

1

0

0

1

ISTP

Craftsman

1

0

1

0

ISFJ

Defender

1

0

1

1

ISFP

An Artist

1

1

0

0

INTJ

The Scientist

1

1

0

1

INTP

Architect

1

1

1

0

INFJ

Confidant

1

1

1

1

INFP

Idealist

Figure
1

-

Personality types and their meanings




The generalizing of
personality types


Project Delivery for IS687


Sightseeing



Surveying

The

common

interests of the corresponding clusters have been determined
by
survey
ing

15 people.


Traditionalists

-
1000,1010,0000,0010

Dominan
t Introverted Sensing Types
ISTJ, ISFJ

historical places, zoo

Dominant Extr
avert
ed Sensing Types
ESTJ,
ESFJ

botanical garden
,

art museums, books


Experience
rs
-

1001,1011,0001,0011

Dominant Introverted
Sensing Types ISTP, ISFP
adventurous
,
zoo

Dominant Extraver
ted Sensing Types ESTP,ESFP

art, nature,
historical museums
,









science and tech


Idealists
-

1110,1111,0110,0111

Domin
ant Introverted Intuitive Types INFJ &INFP

nature, music

Dominan
t Extraverted Intuitive Types ENFJ&ENFP

entertainment,
art


Conceptualists
-

1100,1101,0100,0101

Dominant
Introverted Intuitive Types
INTJ & INTP

historical places

Dominant Extraver
ted Intuitive Types ENTJ &ENTP

science&tech

museums




Traditionalists(SJ)

Experiencers(SP)


Idealists(NF)

Conceptualists(NT)

Analytical

Introverted
Sensing

ISTJ, ISFJ

17%

Introverted
Sensing

ISTP, ISFP

15%

Drivers

Introverted
Intuitive

INFJ, INFP

5%

Introverted Intuitive

INTJ, INTP

1%

Amiable

Extraverted
Sensing

ESTJ, ESFJ

28%

Extraverted
Sensing

ESTP,ESFP

25%

Expressive

Extraverted
intuitive

ENFJ, ENFP

7%

Extraverted intuitive

ENTJ, ENTP

2%


Project Delivery for IS687


Sightseeing



19


4. Database Design (Ibrahim + Naitik + Joachim)

The design of the database for the project site seeing has been modeled to provide
access to the following data;



USER



USER INTEREST



MUSIC



FOOD



VENUE

The second part of the design incorporates data from users not providing personal
information but data is captured through header files giving information about their
Internet protocol address. Data has also been designed for machine learn
ing
training with the following tables:



USER CLUSTERS



VENUE CLUSTERS



STATIC SIGHTS



DECISION TREE


Entity Relationship Model (ERM)
.








1

N















USER

INTEREST

VENUE

HAS

dob

fname

lname

userid

passwor
d

interesti
d

Interest
description

type

location

venueid

description

LIKES

HAS

GIVES

N

1

description

1


N

Project Delivery for IS687


Sightseeing



C
ardinality constrains and relationship mapping

This shows how the entities relate

with each other and also defines how the foreign
keys will be included in the related tables to enable queries from normalised tables.

User and interest

There is a one to many relationships between user and interest, because one user
may have many interes
ts. Therefore the primary key of user is going to be included
in the interest table to reference which specific interest a user may have.

User and venue

There is a one to many relationships between a user and a venue because one user
may want to visit diff
erent venues in one geographic location. Therefore a mapping
has been created to include the primary keys of user into the venue table.

User and likes

The entity LIKES is a weak entity that is defendant on USER. Therefore the primary
key of the user has be
en mapped to give a relation UserLikes.


In the database another table Ranking has been created for machine learning
purposes and also to create a link to venues for the user to access adequate
suggestions about places and make assessment based on experien
ce. This has been
mapped with the venue_id.


Entities and attributes.

The emerging entities from the design have been modeled together with the
respective attributes describing the database physical structure.

Elementary Entities and attributes


USER
(userid, fname, lname, password, DOB)

INTEREST (interestid, interest description, type)

VENUE (venueid, location, description)

LIKES (description)


Final normalized relations.


INTEREST (
interest_id
, interest type, interest description)

USER (
user_id
, fn
ame, lname, password, DOB, email)

VENUE (
venue_id
, venue type, venue name, street, city, state)

RANKING (
ranking_no
, venue type, rank)

USER_INTEREST ( user_id, interest_id)

USER_VENUE ( user_id, venue_id, venue type)

VENUE_RANKING ( venue_id, ranking _
no)

USER_LIKES (venue_id, user_id)


Project Delivery for IS687


Sightseeing



21



Machine learning relations.

The relations described have been modeled as a training data set to give sample
data in the form of tables used to give output to the prototype of the system.


DECISION TREE WITH NO USER D
ATA

( id, operation system, ip address, confidence interval (A), confidence interval (B),
confidence interval (C), confidence interval (D), prediction category)


HTTP HEADER INFORMATION

(Id, operation system, language, ip address)


HTTP TRAINING SET
DECISION TREE (id, operation system, language, ip address,
category)


ML_TRAINING SET CLASSIFICATION PERSONALITY TYPE (user_id, label)


ML_USERCLASSIFICATION (user_id, confidence trad, confidence exp, confidence
idea, confidence con prediction label)


USE
R CLUSTERS (user_id, cluster name)


VENUE CLUSTERS (venue_id, cluster name).


In addition also included for machine learning purposes, is the relation STATIC
SIGHT. This table has been created to expedite learning from data of the user or
without user data

and give specific suggestion of sight locations.


STATIC SITES (sight_id, name, street, state, zip).


Project Delivery for IS687


Sightseeing



5. Machine Learning Design

Basically this section handles three parts: The

Techniques

used

for each Suggestion
Model, How the data will be generalized a
nd how the machine learning process
could be
implemented. In

this area, it is important to differentiate between the
suggestion with existing user data and without user data. As it is explained in
previous chapters the recommender uses different techniques

for each process.

Suggestion Model without user Data

The system allows its users to get recommendations without inserting any user
specific data. In this case the system is using meta
-
data from the users access. These
meta
-
data is gathered from the HTTP H
eader:



Operation System

iPhone, Android, Windows PC, Blackberry OS



IP Address


Geo
-
location



Accepted Language


The following figure illustrates how this decision tree could be set up.




Figure
2
: Decision Tree for categorization of HTTP Header information

Project Delivery for IS687


Sightseeing



23

This Data could be used to classify the user, without having any specific data. In this
project this method is based static business rules. This business rules are
hierarchically organized. T
his leads to the approach, that the user is classified via a
decision tree. After gathering and parsing the data from the HTTP Header the
system analyzes t
he data with following approach in Figure 1.
This tries to analyze
and predict with static business r
ules, what kind of user is accessing the system. The
boundary of this system is limited to the number of data fields of the HTTP Header
and the number of identified patterns.

For creating the decision tree there are two
possibilities:



Static Decision Tree



Dynamic Decision Tree with Training set

The static decision tree is a simple combination of If, Then, Else decisions. In many
cases these decisions are hard coded into the website or software. At the end of each
statement there is a set of business rules,
which solve the event and decide. The
disadvantage of
this approach is, that is not flexible enough to address special
events. Additionally in some cases it is impossible to such a decision tree an
d
business rules in between can
not be addressed.


The other

approach is more dynamic decision tree, which is set up by a training set.
In general the training set is a set of labeled data, which can be used to generalize a
bigger set of data. One of the major challenges is the right number and quality of
datasets
within the training set. In this case the training data could have the
following structure:



Figure
3
: Structure of the trainingset of the Decision Tree

The column id is for identification and adds no value for the process of mac
hine
learning. The
Operation System
, Language, IP
Address

are the core attributes to set
up the decision tree. The Category is the label of the data. In this project the label is a
category from A to D. Behind these Categories are a set of venues, which ar
e the
recommendations for this event. A possible training set could be the following:


Project Delivery for IS687


Sightseeing




Figure
4
: Training set for the Decision Tree


After the structure and training set is set up the machine learning process could be
created.
The author uses an open source tool
, called RapidMiner
1

to set up the
machine learning
process.

Machine learning Process

The Machine learning process
consists of five basic steps. The results are also stored
in the database, that the application can use th
e result without running the process
to access the result.



Read Database


reads the training set from the database



Execute SQL


flush the export table



Set Role


Set id as id and category as label



Decision Tree


The core element, which creates the decis
ion tree



Apply Model


Applies the model to the data se



Write Database
-

Write the results back to the database





1
RapidMineris an open sourcedataminingtoolwritten in Java. The
communitiyeditionisfreetouseandcanbeused in academiccontexts. The
sourcescanbedownloaded

Project Delivery for IS687


Sightseeing



25

The elements can be used and connected in a sequential order. The following
figure
3 illustrated the Machine
Learning process to generate a dec
ision. The result will be
stored in the database


Figure
5
: The Machine Learning process of the Decision Tree


The results are the confidence for every attribute and the recommendation. The
following figure shows the results
of th
e machine learning process
.



Figure
6
: Results of the machine learning

The results can be used to make recommendations and suggestions for specific
combinations.

Suggestion Model with user Data

This stream analyzes user data, which is received through the registration process.
The registration process allows its users to get better suggestion,
which perfectly
fits their demands.
In chapter 3 the author explained, how personality types can be
gener
ated and applied to
persons. These data is stored in the database table User as
a binary code 1 / 0. The following chapter
will describe, which techniques and
processes are used to get
suitable

results
.


Project Delivery for IS687


Sightseeing



P
eer evaluation

The peer evaluation is using the cl
ustering technique to get results from peers. In
the first part the users needs to input their information. This information is used to
create the clusters. For this case the project team decided to use k
-
mean clustering
to cluster people into a certain nu
mber of groups (k). Basically the data from the
survey is used to build up the clusters. The following table illustrates, which data is
used for the clustering:



Figure
7
: User Table

In general the machine learning process reads
the whole table and clusters the
userids into specific clusters based on k
-
means algorithm.


The machine learning for this clustering activity is created with the RapidMiner
open source tool, which allows easily reading the data from the database write the

results into the database. The process consists of five basic steps:



Read Database


Reads the Data set from the database (User Table)



Execute SQL


Flush the Table ML_Userclusters



Set Role


Assign roles to the specific attributes of the Data set



Cluster
ing


Clusters the data
with k
-
means algorithm / max runs 10 max
optimization steps 100



Write
Database



Writes the results to the Table ML_Userclusters for usage


The following figure shows the sequential relationship between these steps:



Figure
8
: Machine Learn Process for peer evaluation


After the clusters are set up, the result is stored in a database table
(ML_Userclusters). This table is a relation between the userid and the cluster name.
The following table
illustrates the relationship between these both:

Project Delivery for IS687


Sightseeing



27



Figure
9
: ML_Userclusters table

The machine learning process fills this table at the end of this process.


The application has not to run the process every time when the user requ
ests a
recommendation. The process should be scheduled as kind of batch job. The project
team realizes, that this has also its downturns, but for the first release of the product
the batch processing would be the most convenient.


The user

gets the recomme
ndations through an SQL statement which accesses the
Table ML_Userclusters and the table Userlikes which is a table, which realizes the
relation between userid and venueid. This table stores, which users liked, which
venue in the past. Out of this informat
ion the system can make recommendations
from peers. For example, if Arya and Joachim are in the same cluster it would be a
suitable suggestion that Arya likes the places, which Joachim already visited,
because
s
he is in the same cluster as Joachim.

The fol
lowing SQL statement realizes
this recommendation behavior.


SELECT

distinct ul.`venueid`, v.name

FROM
ML_userclusters as uc, userlikes as ul , venues as v

W
HERE
uc.clustername = uc.clustername and ul.venueid = v.venueid


This SQL needs the userid as an inp
ut value. In the application this information will
be received through the generated session id. This will be explained in chapter 7 in a
greater detail. The results are the venueid and all related information like
venuename, address, etc. After that proce
ssing the application can display the
recommendation from peers

Static business rules

The second approach is via Business rules, which
basically consists of two parts. The
first part is training the classification to group specific personality types togeth
er to
generalize in a better way. The second step is applying the machine
-
learning model
to the whole user table. The result can be used to make suggestions for each user
using the relationship in the venue table and the classification.


First, the
classification needs a training set, which basically is a relation between all
the possible answers 1/0 and the labeled data
. The following figure illustrates the
training set:

Project Delivery for IS687


Sightseeing



The table is using an abbreviation for the
personality types. The table will make
this relationship


Abbrev
i
ation

Personality Type

TRAD


EXP


IDEA


CON


Table
1

-

Abbrev
i
ation of Person
a
lity types

This abbreviation is also used in the
venue table.











The machine learning process consists of 6 activities,

which are sequential executed.
The process is designed with RapidMiner integrates following tasks:



Read Database



Set Role



Decisio
n Tree



Read Database



Apply Model



Write Database

The following figure will illustrate how this process is organized:


Figure
11

-

Machine Learning process for classification of User Data using personality types

After the previously

explained steps were executed the results will be reflected to
the database. The advantage of this approach is, that the application can access this
information without running the machine learning process again.

The result is
Figure
10

-

Training Set for Machine Learning
ML_Tr
ainingsSetClassificationPersonaltyType

Project Delivery for IS687


Sightseeing



29

stored in the table
ML_Userc
lassification

and consists of a confidence level for each
attribute and the labeled

data:



Figure
12

-

Result table of the machine learning process
-

ML_Userclassification

The most valuable relationship is the prediction(Label) a
nd the userid, because this
is the result of this process and can be used to make suggestions. The project is
using the relationship in the venue table and the gathered data in the
ML_Userclassification table to make suggestion. A useful approach is using
a SQL
Query to get suggestions
for each user. The following SQL gets the recommendations
using the personality type for each userid:


SELECT DISTINCT v.*

FROM venue as v, `User` as u, `ML_Userclassification` as ml

WHERE v.personalitytype =
ml.`prediction(Label)` and u.`User_id` = ml.user_id


The application has to limit the result using the sight_id reference key. This would
reduce the
result set to only the venues,
which are geographically near to the route
of the user.

Combination / Ranki
ng

This section describes how these two basic approaches can be combined to make a
specific

kind of r
anking possible. For example if a venue is rated in both processes it
is in the opinio
n of the project team more valu
able than a single rated
venue. A
poss
ible solution could be a set of SQL statements, which process the results and
merge them into a set of ranked venues.

The suggestions are coming from three
specific recommender models:



Peer evaluation from other users



Results from Business rules

Project Delivery for IS687


Sightseeing





Results
from HTTP Header information

The basic idea is, that the implementation is using every evaluation and finds the
insection between them. The following figure illustrates in a more convient way:
















The idea is that results which are appearing in different results sets are higher
ranked and more valuable than other results. A possible implementation of that
could be a set of SQL statements with a union
at the end of the query. This union
combines al the results. After the union is complete the system should count the
same results and returns distinct values. After this step is comp
leted the system
should respond.


In the last step the results set should

contain the best
-
ranked results but also a
variety of different activities. This is possible through the venue_type field in the
venue table. This filter should be at the last place of the process for example by a set
of order by and group statements with
in the ranking process.

Results

The previously described approaches give a good insight into how suggestions are
made and aggregated. In the
project team’s
opinion there are
still
a lot of
improvements inside the whole
machine learning
process
. Especially
the

real
-
time
processing or the ranking algorithm

could be improved and refined
. Additionally,
the project teams thinking about to combine more different process to get better
recommendations. In the prototype, the team decided to start with a small set of

algorithms to get a clear understanding how machine learning can be useful to reach
the project goals.

Peer evaluation

Results from
Business Rules

Results fr
om HTTP
Header

Best Results

Project Delivery for IS687


Sightseeing



31

6. Business Benefits

This section will elaborate, what the business model behind this project and the
poten
tial risks are. Furthermore this part will de
velop a roadmap how further
releases are structured and scheduled.

Business Model

Sightseeing mobile app will have three versions: Apple version, Android version and
Windows version. The initial release of this app will be free to download. While we
provide free service to our customer, we gain our profits through advertisements,
rankin
g venues and selling data to business partners. Just like Google search, which
use two ranking system: one is based on customer review and the other is based on
how much the
business pays

for the ranking. Our Sightseeing project will also
combine this two
review systems. Companies have to pay us if they want to move up
in the ranking system. If a customer actually went in to a venue the app suggested,
that company will pay us extra fee. In addition, we can also generate revenue
through professional consulti
ng. Through data mining on the data we gather from
the app, we have the ability to provide venue owners with professional advice in
terms of economic environment and marketing strategies.










Company A

Company B

Company C

Our Service

Bet against

each other

Accurate customer /
tourist portfolio

Sensitive Advertisement

data


Benefits

Project Delivery for IS687


Sightseeing



I
nvestment

To start the company, we will have to rent an office, buy office supply, hiring the
mobile app developers and advertise our product. The initial
investment will
roughly around $500,000 since the majority of the investment will focused on hiring
the developers, that include hiring one senior mobile app developer and several
junior developers.

Potential

The first release will mainly focus on New York

City area since it has the most
visitors in the Tri
-
State Area. Plus, there are also a lot of venues in New York City. As
our profit increases, we will expand our service to other popular travel areas in the
United States, such as Boston, Atlantic City an
d Philadelphia. Eventually, we will
cover services to the whole country.


If the initial release goes pretty well, more functions will be added to the sightseeing
app. It not only can suggest best routes and venues that meet the customers’ needs,
but also
allows customers to book flight ticket, reserve restaurant seats, check
bus/subway schedule, do shopping online and etc. With this single app, users don’t
need to carry travel book or pay a lot of money to the tourist guide anymore. Every
travel related in
formation is stored in this single app. As we add more functions to
our mobile app, we will have more business partners.

Roadmap

For the first release, a Sightseeing website will be created. It can suggest the best
route for a user who has several places w
ants to travel, and also provide
recommendations for various venues and events that the user may interest in.
During the second release, more mobile app developers will be hired. They will
work on the first Sightseeing mobile app. In the second release, lo
cation based
products and services will be included. A customer can also leave a review for a
venue he or she visited. That will help us cluster the data and provide more
personalized suggestion for the user. In the third release, we will include weather
a
nd coupon features. Since the weather is one big factor that affects the user’s travel
plan, Sightseeing mobile app will adjust the route and venues according to the
weather. For example, Sightseeing app will suggest indoor activities to the user in a
rain
y day. Coupon feature will not only attract more and more customers to use our
service, but also promote sales for various businesses. In the later release, we plan
to expand our service to other popular travel destinations.

Project Delivery for IS687


Sightseeing



33


Figure
13

-

Roadmap and future releases

Project Management

This Sightseeing project will approximately take three months to complete. At the
end project, a prototype has to be created. For our project, we intend to design a
mobile app for various smart phones.

Due to the lack of technical skillset to design a
mobile app, we decide that we will design a Sightseeing website for our project. It
can suggest the best route according the customers’ information, and also
recommend venues that attract the users. Since
the prototype is not a mobile app, it
won’t suggest location based products and services. The project management model
we used for our Sightseeing project is Spiral Model. Every week we have to
determine our project objectives, alternatives and constraints

in class. During the
each group meeting, we have to evaluate alternatives, identify and resolve risks.
Then we also have to develop our prototype, and verify it. At the end of each group
meeting, we have to plan the next project deliverable for next week.

Each week is
like an iteration. For the first iteration, we have to come up a requirement plan; then
the development plan, integration plan and so on for the next several iterations.


Prototype
Dec 2011

Alpha
Release
Jan 2012

Beta
Release
March
2012

Go
-
live
June
2012

Project Delivery for IS687


Sightseeing




Figure
14

-

Gantt Chart for Project Management

Risk Management

In team identified two risks within the project scope:

1.

Group members lack the technical skills that were required by this project,
such as data mining skill, database skill and website design skill. To
overcome this problem, we divide the project into several parts, such as
documentation, data mining, database design and website design. Everyone
has its own part, if there is a problem, a team member cannot solve. That
team member can bring it to the meet
ing, and ask for help. Or the team
member can search the internet for relevant information.

2.

How to collect user’s information that can help us design a model to suggest
customer their interested venues is the biggest problem we are facing right
now. We nee
d to come up a few good survey questions that can really cluster
the users into different clusters. As the professor mentioned, we can use http
header to identify the type of users’ cell

phones for the unregistered users.
For example, we may suggest
an exp
e
nsive restaurant for a user who
Project Delivery for IS687


Sightseeing



35

use
s
ourapp through iPhone 4s. In the http header, it will contain the basic
smart

phone information.


SWOT Analysis






Project Delivery for IS687


Sightseeing



Strengths:


1.

The model has been successfully deployed using php and mysql.

2.

The Google maps API has been integrated for an enhanced user
experience.

3.

The suggestions are pinned around the route instead of advertisement
pop
-
ups. Such suggestions have no chance of being blocked by a pop
-
up
blocker.

Weaknesses:


1.

The model needs a
better look and feel.

2.

The complete address of the location suggested by the model is not
displayed
.

Only the name of the venue is pinned up into the Google Maps.


Opportunities:


1.

The model has wide opportunities as it can expand beyond New York
City.

2.

The
model has the potential to tap the best of opportunities in tri
-
state
and make the best out of advertising.

3.

The model can have discussions and forums which can help the users to
give reviews and rankings for the places visited by them.


Threat
s:


1.

The surve
y questions are based on a psychological model, the results of
which may vary from person to person. So, the results might not be
accurate for everyone.

2.

The ‘Email’ and ‘Date of Birth’ fields do not have a specific input type. So,
this leverages the user t
o enter any information in those fields.








8. Conclusion

In conclusi
on the project has reached its

project goals. Additionally the prototype
gives the reader a good insight into the capabilities

and the potential

of
the

idea. The
Project Delivery for IS687


Sightseeing



37

machine learning part

takes a central position in the whole system and feeds the
application with information. In the authors view, the machine learning part could
be extended by additional machine learning techniques. One major challenge is to
combine different information so
urces and to find a suitable algorithm to rank or
weight each source individually. This algorithm should also address efficiency
issues, because the application should process information in real
-
time.


The idea behind MB
T
I

personality type finder is
a goo
d starting point to classify
persons
. It gives a good understanding, what is important for making suggestions
and gives
classification possibilities
. In future developments the project team should
expand the idea and integrate more attributes to refine the

classification and
clustering of user data.


The
project team gets a good insight into the end
-
2
-
end process of designing and
implementing a machine learning system.

In the initial phase the project team really
focused on, which attributes are significant

to make suggestions. The team stuck
into thinking about the ultimate solution instead of minimizing the number of
attributes and sta
rt with less attributes. The MB
T
I
helped the project team
significantly in finding a suitable classification scheme.


In sum
mary, the potential of this project is quite high and developed algorithms are
a good starting point

for further developments.
The project team is thinking about
developing this idea more further to get a better understanding of machine learning
processes
and techniques.

Project Delivery for IS687


Sightseeing



Appendix


Table descriptions


1.

USER



FIELD
NAME

DATA
TYPE

LENGTH

VALIDATION
RULES

DESCRIPTION

User_id

Integer

11

Identifier

Unique with an
auto increment
as the database
progresses,

fname

Varchar

225



lname

Varchar

225



password

Varchar


Unique

Stored uniquely
for entry into the
database

Email

Varchar

225

Unique

Specific
identification

DOB

Varchar

12




2.

INTEREST


FIELD NAME

DATA
TYPE

LENGTH

VALIDATION
RULES

DESCRIPTION

Interest_id

Interger

11

Identifier

Uniquely
describes a
particular event
or activity.

Interest_type

Varchar

50



Interest_description

Varchar

50




3.

VENUE


FIELD NAME

DATA TYPE

LENGTH

VALIDATION
RULES

DESCRIPTION

Venue_id

Integer

12

Identifier of a
venue

Shows a distinct
venue

Venue_type

Varchar

50




4.

RANKING


FIELD NAME

DATA TYPE

LENGTH

VALIDATION

RULES

DESCRIPTION

Project Delivery for IS687


Sightseeing



39

Ranking _no

varchar

50



Venue_type

varchar

50



rank

Integer

11




5.

USER_INTEREST


FIELD
NAME

DATA TYPE

LENGTH

VALIDATION

RULES

DESCRIPTION

User_id

Integer

20

Unique

References
user

Interest_id

Integer

20

Unique

References
interest


6.

USER_ VENUE


FIELD
NAME

DATA TYPE

LENGTH

VALIDATION

RULES

DESCRIPTION

User_id

Integer

11

Foreign key

References user

Venue _id

Integer

12

Foreign key

References
venue

Venue type

Varchar

50






7.

VENUE RANKING


FIELD NAME

DATA TYPE

LENGTH

VALIDATION
RULES

DESCRIPTION

Venue_id

Varchar

10

Foreign key

References
venue

Ranking_no

Varchar

10

Foreign key

References
ranking



8.

DECISION TREE WITH NO USER DATA.


FIELD
NAME

DATA TYPE

LENGTH

VALIDATION

RULES

DESCRIPTION

Id

Integer

11



Operation
system

varchar

10



Ip address

varchar

2



Confidence
interval (A)

Double




Confidence
interval (B)

Double




Project Delivery for IS687


Sightseeing



Confidence
interval (C)

Double




Confidence
interval (D)

Double




Confidence
category

varchar

1









9.

HTTP HEADER INFORMATION.


FIELD
NAME

DATA TYPE

LENGTH

VALIDATION

RULES

DESCRIPTION

Id

Integer

11



Operation
system

Varchar

255



Ip address

varchar

255










10.

HTTP TRAINING SET DECISION TREE


FIELD
NAME

DATA TYPE

LENGTH

VALIDATION
RULES

DESCRIPTION

id

Int

11



Operation
system

Varchar

100



language

varchar

100



Ip address

varchar

100



category

varchar

100




11.

USER _CLUSTER


FIELD NAME

DATA TYPE

LENGTH

VALIDATION

RULES

DESCRIPTION

User_id

Integer

11

Foreign key

References user

Cluster_name

Varchar

100




12.


VENUE_CLUSTER


FIELD NAME

DATA TYPE

LENGTH

VALIDATION
RULES

DESCRIPTION

Venue_id

Integer

11

foreign

References
venue

Cluster_name

Varchar

100




Project Delivery for IS687


Sightseeing



41



13.


STATIC SITES


FIELD
NAME

DATA TYPE

LENGTH

VALIDATION
RULES

DESCRIPTION

Sight_id

Integer

11

Primary key

Uniquely
indentify a site

name

Varchar

100



street

Varchar

100



state

Varchar

100



zip

Varchar

10





14.

USERLIKES


FIELD
NAME

DATA
TYPE

LENGTH

VALIDATION
RULES

DESCRIPTION

User_id

Integer

10

Foreign key

References User

Venue_id

Interger

10

Foreign key

References
Venue


15.


USERCLUSTERS


FIELD
NAME

DATA TYPE

LENGTH

VALIDATION
RULES

DESCRIPTION

User_id

Integer

10

Foreign key

References User

Confidence
Trad

double




Confidence
Exp

double




Confidence
Idea

double




Confidence
Con

Double




Prediction
Label

Varchar

2












Project Delivery for IS687


Sightseeing



INTERNAL STORAGE STRUCTURE OF TABLES


Table structure for table interest

Field

Type

Null

Default

interest
_
id

int(11)

No

0

interest_type

varchar(50)

No


interest_description

varchar(50)

Yes

NULL





Dumping data for table interest

interest_id

interest_type

interest_description

1

sky diving

recreation

2

Restaurant

Italian food

3

Restaurant

American food

4

Restaurant

Chinese food

5

Restaurant

Spanish food

6

Site

empire
state building

7

Site

statue of liberty

8

Site

convention Centre

9

Recreation

ice skating

10

Recreation

Lowes movie theatre

11

Recreation

Broadway performing art studio

12

Recreation

guygien museum

13

Shopping

Rockefeller center

14

Shopping

Manhattan Mall






Table structure for table ML_DecisionTreeNoUserData

Field

Type

Null

Default

id

int(11)


Yes


NULL

OpSys

varchar(10)

Yes


NULL

IP Add varchar(9 Yes


NULL

con(A)


double


Yes


NULL

con(B) double

Yes


NULL

con(C)


double


Yes


NULL

con(D)


double


Yes



NULL

pred(Cat) varchar(1) Yes


NULL




Project Delivery for IS687


Sightseeing



43






Dumping data for table ML_DecisionTreeNoUserData

id

OpSys


IP


con(A) con(B) con(C) con(D)

pred(Cat)

1

iPhone

local


0.5

0.5

0

0

A

2

iPhone

local


0.5

0.5

0

0

A

3

iPhone

non
-
local

0

0

0.6

0.4

C

4

Android

local


0.5

0.5

0

0

A

5

Blackberry

non
-
local

0

0

0.6

0.4

C

6

Android

local

0.5

0.5

0

0

A

7

Blackberry

non
-
local

0

0

0.6

0.4

C

8

Windows

non
-
local

0

0

0.6

0.4

C

9

Blackberry

non
-
local

0

0

0.6

0.4

C

10

Windows

non
-
local

0

0

0.6

0.4

C

11

Blackberry

non
-
local

0

0

0.6

0.4

C

12

Windows

local


0.5

0.5

0

0

A

13

iPhone

non
-
local

0

0

0.6

0.4

C

14

Android

non
-
local

0

0

0.6

0.4

C




Table structure for table ML_HTT
PHeaderInformation


Field

Type

Null

Default

id

int(11)

No


OpSys

varchar(255)

Yes

NULL

IP Adressvarchar(255
Yes

NULL


Dumping data for table ML_HTTPHeaderInformation

id

Operation_System

IP Adress

1

iPhonelocal

2

iPhone

local

3

iPhone

non
-
local

4

Android

lo
cal

5

Blackberry

non
-
local

6

Android

local

7

Blackberry

non
-
local

8

Windows

non
-
local

9

Blackberry

non
-
local

10

Windows

non
-
local

11

Blackberry

non
-
local

12

Windows

local

13

iPhone

non
-
local

Project Delivery for IS687


Sightseeing



Table structure for table ML_HTTPTrainingSetDecisionTree

Field

Type

Null

Default

id

int(11)

No


Ops_Sys

varchar(100)Yes

NULL

IP Adress

varchar(100)Yes

NULL

Category

varchar(100)

Yes

NULL


Dumping data for table ML_HTTPTrainingSetDecisionTree

id

Operating
_System

IP Adress

Category

1

iphone

local

A

2

iphonelocal

A

3

bl
ackberry

local

A

4

PC

local

B

5

PC

local

B

6

iphonenon local

C

7

blackberry

non local

C

8

PC

non local

D

9

Linux

non local

D

10

Linuxlocal

B

11

android

non local

C



Table structure for table ML_TrainingsSetClassificationPersonaltyType

Field

Type

Null

Default

id

int(11)

No


Q1

int(11)

Yes

NULL

Q2

int(11)

Yes

NULL

Q3

int(11)

Yes

NULL

Q4

int(11)

Yes

NULL

Label

varchar(100)

Yes

NULL



Dumping data for table ML_TrainingsSetClassificationPersonaltyType


id

Q1

Q2

Q3

Q4

Label

1

1

0

0

0

TRAD

2

1

0

1

0

TRAD

3

0

0

0

0

TRAD

4

0

0

1

0

TRAD

5

1

0

0

1

EXP

6

1

0

1

1

EXP

7

0

0

0

1

EXP

8

0

0

1

1

EXP

9

1

1

1

0

IDEA

10

1

1

1

1

IDEA

11

0

1

1

0

IDEA

Project Delivery for IS687


Sightseeing



45

id

Q1

Q2

Q3

Q4

Label

12

0

1

1

1

IDEA

13

1

1

0

0

CON

14

1

1

0

1

CON

15

0

1

0

0

CON

16

0

1

0

1

CON



Table structure for table
ML_Userclassification

Field

Type

Null

Default

id

int(11)

Yes

NULL

Q1

int(11)

Yes

NULL

Q2

int(11)Yes

NULL

Q3

int(11)

Yes

NULL

Q4

int(11)

Yes

NULL

conf
(
trad
)

double

Yes

NULL

conf
(
exp
)

double

Yes

NULL

conf
(
idea
)

double

Yes

NULL

conf
(
con
)

double

Yes

NULL

pred(
label
)

varchar(4)
Yes

NULL


Dumping data for table ML_Userclassification

Id Q1 Q2 Q3 Q4 conf(trad) conf(exp) conf(idea) conf(con)

pred(Label)

1
1


1

1
1

0

0

1

0

IDEA

2
1


1

1
0

0

0

1

0

IDEA

4
1


1

0
1

0

0

0

1

CON

5
1


1

0
0

0

0

0

1

CON

6
1


0

1
1

0

1

00

EXP

7
1

0

0
1

0

1

0

0

EXP

8
1


0

1
0

1

0

0

0

TRAD

9
1


0

0
0

1

0

0

0

TRAD

10
0


1

1
1

0

0

1

0

IDEA

11
0


1

1
0

0

0

1

0

IDEA

12
0


1

0
1

0

0

0

1

CON

13
0


1

0
0

0

0

0

1

CON

14
0


0

1
1

0

1

0

0

EXP

15
0


0

1
0

1

0

0

0

TRAD

16
0


0

0
1

0

1

0

0

EXP

17
0


0

0
0

1

0

0

0

TRAD

18
0


0

0
0

1

0

0

0

TRAD

22
0


1

1
1

0

0

1

0

IDEA

23
1


0

0
0

1

0

0

0

TRAD





Project Delivery for IS687


Sightseeing



Table structure for table ML_Userclusters

Field

Type

Null

Default

user_id

int(11)

No


clustername

varchar(100)

Yes

NULL


Dumping data for table ML_Userclusters

user_id

clustername

1

A

2

A

3

A

4

B

5

B

6

B

7

B

8

C

9

C

10

C






Table structure for table ML_Venueclusters

Field

Type

Null

Default

venue_id

int(11)No


clustername

varchar(100)

Yes

NULL


Dumping data for table ML_Venueclusters

venue_id

clustername

1

A1

2

A1

3

A2

4

A2

5

A3

6

A4

6

A4

7

A5

8

A5

9

A6



10

A6






Project Delivery for IS687


Sightseeing



47


Table structure for table Ranking

Field

Type

Null

Default

ranking_no

varchar(10)

No


venue_type

varchar(50)

No


rank

int(11)

No




Dumping data for table Ranking

ranking_no

venue_type

rank

1

resturant

1

2

resturant
2

3

resturant

3

4

resturant

4

5

museum

1

6

Mall

1

7

relaxation

1

8

Pub

2

9

Karaoke Club3

10

Gym

1



Table structure for table User

Field

Type

Null

Default

User_id

int(11)

No


Firstname

varchar(255)

Yes

NULL

Lastname

varchar(255)

Yes

NULL

userName

varchar(255)

No


password

varchar(255)

Yes

NULL

DOB

varchar(12)

Yes

NULL

email

varchar(255)

Yes

NULL

Q1

tinyint(11)

Yes

NULL

Q2

tinyint(11)

Yes

NULL

Q3

tinyint(11)

Yes

NULL

Q4

tinyint(11)

Yes

NULL



Project Delivery for IS687


Sightseeing



Table structure for table UserLikes


Field

Type

Null

Default

user_id

int(11)

No


venue_id

int(11)

No



Dumping data for table UserLikes

user_id

venue_id

1

2

2

4

3

3

3

5

4

2

1

3

5

4

6

4

7

4






Table structure for table user_interest


Field

Type

Null

Default

user_id

int(20)

No


interest_id

int(20)

No



Dumping data for table user_interest

user_id

interest_id

1

1

2

2

3

3

4

4

5

5

6

6

7

7

8

8

9

9

10

10

11

11



Project Delivery for IS687


Sightseeing



49


Table structure for table venue_ranking

Field

Type

Null

Default

venue_id

varchar(10)

Yes

NULL

ranking_no

varchar(10)

Yes

NULL


Dumping data for table venue_ranking

venue_id

ranking_no

1

2

2

4

3

1

5

2

3

1

4

3



Bibliography


1.

Apayın, Ethem
-

Introduction into Machine Learning
-

The MIT Press
Cambridge, Massachusetts 2010

2.

The template used is

a free template from a
website:



http://www.bestfreetemplates.info/webtemplates/category
-
19.html

3.

Other web resources include:

-

http://en.wikipedia.org/wiki/Myers
-
Briggs_Type_Indicator

-

http://www.davidmarkley.com/personality/personhome.htm

-

http://www.personalitytest.net/types/descriptions/index.htm