Title: P1246341506PzZRW
1KDD Cup 2000 Question 5
2 Overview
- Aim
- Given a set of page views, which product brand
(Hanes, Donna Karan, American Essentials, or
none) will the visitor view in the remainder of
the session? - Data
- Raw Data - Clicks
- Aggregated Data - Sessions
- Number viewed later for each brand
- Indicator No Top Brand Viewed Later
- Methods and Tools
- Exploratory Data Analysis - S Plus
- Amdocs Business Insight Tool
- Classification Tree
- Two-Class View or not View
- Multiple-Class 3 Brands
- Rules Extraction
3Decision Chart(Our process of identifying
brand-hunters)
4General guidelines for discriminating between
sessions
- Crawler Agent?
- Many sessions terminated after the first click
due to the fact that it was a crawler agent. If
its a crawler agent, then STOP (No top brand is
viewed). - Distinguish between viewing and not viewing a
top brand - Given that it isnt crawler agent session, we
differentiate between sessions in which none of
the specific brands were viewed and those in
which at least one was. If none will be viewed,
then STOP. - Distinguish between each of the top brands
- Given a surfer will view one of the three product
brands, we differentiate between sessions
according to which of the three is viewed Donna
Karren, Hanes or American Essentials.
5Percentage of sessions with any top-brand view
per referring site- Visitors from FashionMall
are more likely to be interested in brands.-
Visitors referred from MyCoupons probably want to
use the coupons for buying one of the brands
products. - On the contrary, entering from
ShopNow reduces the likelihood for brand-browsing.
6Percentage of sessions with any top-brand view
per product category- DK and Hanes sell only
Pantyhose, so entering this category increases
probability of seeing at least one them.- Some
brands have Basic Products collections, so
entering this category might expose surfer to
these brands.- Viewing Men or Women products
increases probability of viewing one of the
brands to 21.
Product Category
7Percentage of sessions with any top-brand view
per segment A few other interesting phenomena
indicate the tendency to view a top-brand. Among
those are sessions that last beyond the first
click, returning customers and search engine
usage.
8Segments likely to view Hanes Products- Viewing
products that include the dressing room
feature.- Not viewing any boutiques but viewing
last a collection line.- Viewing Oroblu
Products, which are very similar in type and in
cost to Hanes products.
9Segment likely to view Donna Karan
ProductsArriving from the shopping sites
ShopNow or FashionMall result in viewing DK
products.Surfers interested in DK might be heavy
web shoppers.
10Segment likely to view American Essentials
Products- Arriving from MyCoupons can indicate
that there are coupons that can be used while
purchasing AE products.- Free gift templates are
associated with AE.- Viewing sports collections
or men's products is related only to AE (Hanes
and DK offer only pantyhose).
11Behavior Changes along Time
12The behaviour of viewing top brands changed over
time
- Donna Karen (DK) products werent for sale on the
site prior to February 26th. - There is an extreme peak in views of American
Essential (AE) products on March 1st and much
milder peaks in the Hanes and DK views. - On March 1st there were many discounted
purchases. - Most AE purchases are discounted
- On that day the amount of AE views was 10 times
the daily average. - The February 28th Ally Mcbeal ad probably caused
the rise in product views. - During March DK views remain relatively high,
Hanes views drop (still higher than the February
level) and AE views drop substantially.
13Amdocs Business Insight Tool Main Features
- Data Preprocessor
- Extracting, transforming and loading data
- Data Mining Engine
- Segmentation - Variable Selection, Variable
Discretization - Sampling
- Profiling - Classification Tree
- Predictive Modeling
- Scoring and Ranking
- Rule Engine
- Rule Extraction
- Rule pruning
- Similarity checks
- Analyst User Interface
- Create new or modify existing patterns
- Visualization
- Exploring high dimensional patterns
14(No Transcript)
15(No Transcript)