business intelligence & data mining-17
TRANSCRIPT
-
8/10/2019 Business Intelligence & Data Mining-17
1/17
Personalization of
Supermarket Product
Recommendations
R.D. Lawrence et al.
-
8/10/2019 Business Intelligence & Data Mining-17
2/17
Introduction
Personalized recommender systemdesigned to suggest newproducts to
supermarket shoppers
Based upon their previous purchasebehaviour and expected product appeal
Shoppers use PDAs
Alternative source of new ideas
-
8/10/2019 Business Intelligence & Data Mining-17
3/17
Usual Techniques for Product
Recommendations
Content-based filtering based on what person has liked in the past
measure of distance between vectors representing: Personal preferences
Products
overspecialization
Collaborative filtering
items that similar people have liked
Associations mining (product domain)
Clustering (customer domain)
-
8/10/2019 Business Intelligence & Data Mining-17
4/17
Product Taxonomy
Classes(99)
Subclasses(2302)
Products(~30000)
FreshBeef
Petfoods ..Soft Drinks..
DriedCatFood
DriedDogFood
CannedCatFood
FriskiesLiver(250g)
BeefJoints
-
8/10/2019 Business Intelligence & Data Mining-17
5/17
5
Overview
CustomerPurchaseDatabase
Data MiningAssociations
Data MiningClustering Product
Database
MatchingAlgorithm
Cluster-specificProduct lists
PersonalizedRecommendation
List
Normalized
customer
vectors
Cluster
assignments
Product list
for target customers
cluster
Products eligiblefor recommendation
Productaffinities
-
8/10/2019 Business Intelligence & Data Mining-17
6/17
Customer Model
Customer profile
Vector, C(m)s, for each customer
At subclass level => 2303 dim space
Normalized fractional spending quantifies customers interest in subclass relative
to entire customer database
value of 1 implies average level of interest in a
subclass
-
8/10/2019 Business Intelligence & Data Mining-17
7/17
Clustering Analysis
To identify groups of shoppers with similar spending
histories Cluster-specific list of popular products used as input to
recommender
Clustered at 99-dim product-class level
Neural, demographic clustering algorithms (a type ofSOM)
Clusters evaluated in terms of dominant attributes:products which most distinguish members of the cluster
Cluster 2Frozen foodsCluster 3Wines/Beers/SpiritsCluster 4 - Baby products, household items etc..
-
8/10/2019 Business Intelligence & Data Mining-17
8/17
-
8/10/2019 Business Intelligence & Data Mining-17
9/17
Significant Product Classes
-
8/10/2019 Business Intelligence & Data Mining-17
10/17
Within Cluster Product Popularity
-
8/10/2019 Business Intelligence & Data Mining-17
11/17
Associations Mining
Determine relationships among productclasses or subclasses
Used IBMs Intelligent MinerApriori algorithm
Support, Confidence, Lift factors Rule: Fresh Beef => Pork/Lamb
Support 0.016
Confidence 0.33
Lift 4.9 Rule:Baby:Disposable Nappies => Baby:Wipes
-
8/10/2019 Business Intelligence & Data Mining-17
12/17
-
8/10/2019 Business Intelligence & Data Mining-17
13/17
-
8/10/2019 Business Intelligence & Data Mining-17
14/17
Matching Algorithm
Score each product for a specific customerand select the best matches.
Cosine similarity metric used
C is the customer vector
P is the product vector
mnis the score between customer m and product n
mn= n C(m).P(n)/||C(m)|| ||P(n)||
-
8/10/2019 Business Intelligence & Data Mining-17
15/17
Matching Algorithm
Limit recommendations for each customerto 1 per product subclass, and 2 per class
10 to 20 products returned to PDA
Previously bought products excluded
Data from 20,000 customers
Recommendations for 200
-
8/10/2019 Business Intelligence & Data Mining-17
16/17
Results
Recommendations generated weekly 8 months, 200 customers from each store
Respectable 1.8% boost in revenue frompurchases from the list of recommendedproducts.
Accepted Recommendations from productclasses new to the customer
Certain products more amenable torecommendations.
Interesting recommendations: Wine vs.household care.
-
8/10/2019 Business Intelligence & Data Mining-17
17/17