TimelyBid'com

About This Presentation

Title:

TimelyBid'com

Description:

Flash MX. Spider. Update AI models. Analyze & DM. Novelty. IDEA ... Data mining collection. Sean. TimelyBid website. Data mining algorithms. Database. Go Team! ... – PowerPoint PPT presentation

Number of Views:32

Avg rating:3.0/5.0

Slides: 14

Provided by: seanhum

more less

Transcript and Presenter's Notes

Title: TimelyBid'com

1
TimelyBid.com

Provide an integration between eBay and Google
Calendar for quick and easy visualization of
timely data regarding eBay bids.
By Katie Varland, Josh Smith, Sean Humpherys
MIS-510B with thanks to Dr. Chen Xin Li

2
(No Transcript)
3
Timely Bid Organization
customer
Registers
Spider
API calls
Spider MySQL DB
Timelybid.com
Update AI models
Analyze DM
API calls
Flash MX
Dynamic reports
Data miner with DM computer
4
Novelty

IDEA
Integrate bid data into Google Calendar from eBay
real time
DATA MINING
AWOL risk assessment using three distinct AI
algorithms
VISUALIZATION
Improved visualization of eBay data
Dynamic flash-based charts
Visualization of data mining predictions
LEVERAGE OPEN SOURCE
Joomla Joomla CMS modules
YALE/WEKA for data mining
XML/SWF Charts for dynamic charts

5
(No Transcript)
6
Open Source and Other Tools

Joomla CMS
Customized user registration module
Available Virtuemart for Pay Pal integration
39.95
Apache/Linux/Php/MySQL
eBay API
Google Calendar API
RapidMinger (YALE) / Weka for data mining
XML/SWF Charts for dynamic charts (limited
version free or full license 45)

7
(No Transcript)
8
Demo RapidMiner

RapidMiner formerly YALE
www.rapidminer.com
Download the OWN-free version for easy off-line
mining
Download GPL version if you desire to dynamically
link the system into your program

9
Business Model

Provide an integration between eBay and Google
Calendar for quick and easy visualization of
timely data regarding eBay bids.
Subscription model 12 / year
Advertising model
Uses Pay Pal as payment facilitator, integrates
with Joomlas user registration module

10
Production Costs
11
ROI

Plan A Slow growth by Owners
Break even after 1050 customers (not including
advertising revenue)
Plan B Fast growth with investors
Break even after 9000 customers (not including
advertising revenue)

12
AI Avatars for AWAL Risk Assessment

AI Avatar Joseph is our risk adverse assessor
using C4.5
AI Avatar Julie is an moderate risk assessor
using AD Tree algorithm
AI Avatar Scott is our risk seeking assessor
using REP Tree.

13
Data Mined

Data collected from eBay for mining purposes
isRegistered as dependent variable
userID
Lifetime Positive Feed Back
Feed Back Score
All Positive FB
Member Since
Location
Items for Sale
of SubCategories
of Main Categories
of One Month Positive Feedback

of Six Months Positive FB
of Twelve Months Positive FB
of One Month Negative FB
of Six Months Negative FB
of Twelve Months Negative FB
of Positive FB
of Negative FB
FB Withdrawn Bids Retracted

14
Code for Avatars

if(userArray'twelveMoNeg' lt 1.5)
if (userArray'twelveMoPos' lt 0.5)
if (userArray'numPosFB' lt 46.5) vote
true
else if (userArray'numPosFB' gt 46.5)
if(userArray'posFB' lt 97.55) vote
false
else if(userArray'posFB' gt 97.55)
if (userArray'allPosFB' lt 23450.5)
if (userArray'allPosFB' lt 13578.5)
vote false
else if (userArray'allPosFB' gt
13578.5) vote true
else if (userArray'allPosFB' gt
23450.5) vote false

15
AWOL Risk Assessment

Collected nearly 150,000 records on eBay users
In experimental phase, used up to 80,000 records
for training. Resulted in over-learning but
identified important attributes.
In production phase, used 13,000 specially
selected records to train and 50,000 to test.
Achieved between 95 and 98 accuracies on
predicting if the user became unregistered in the
last thirty days.

16
How to Chart Data