Web Mining Final

8/2/2019 Web Mining Final

1/21

Click to edit Master subtitle styleCompany nameWeb Mining


2/21

Outline

Hype of the web Difficulties with web

Web Mining Advantages / Disadvantages Categories of Web Mining Web Usage Mining


3/21

Hype Of The Web

Information Services

E-mail Business Communities


4/21

Difficulties With Web

Very Huge Information Semi-Structured Data

Redundant Data Web is noisy Customer Behaviour


5/21

Web Mining

Process of discovering useful and

unknown information or knowledgefrom the web data.


6/21

Web Mining - Sub Tasks

Resource Finding

Information Selection Information Preprocessing Data Mining Techniques Analysis


7/21

Benefits

Business Organization Banking Sector

Customer Satisfaction Researchers Society


8/21

Drawbacks

Privacy Issues

Security Issues Mis-Use of Information Sometimes Expensive


9/21

Categories


10/21

Web Usage Mining

The Process of Automaticdiscovering patterns and profile ofusers interacting with a web site.


11/21

Web Usage Mining Model


12/21

Data Cleaning: Clean the raw data,

-Missing value-Redundant data

-Outliers deletion

Trasaction Derivation:

-according to individual user

transaction


13/21

Data Integration:

- Combines data from multiplesources into a data store

Transformation:

- The data are transformed intoappropriate forms for mining such as

generalization and normalization


14/21

Pattern Discovery

Association Rule: X == > Y (support, confidence) 60% of clients who accessed /products/, also

accessed /products/software/webminer.htm

Sequential Pattern:

Discoveryof frequently occurring orderedevents or subsequences as patterns.


15/21

Clustering

process of grouping a set ofobjects into classes of similar objects

Classification

process of finding a model thatdescribes and distinguishes dataclasses or concepts


16/21

K-means Algorithm

Used for clustering, where each clusterscenter is represented by the mean value ofthe objects in the cluster.

Input:

k: the number of clusters,

D: a data set containing nobjects.

Output:

A set of k clusters.


17/21

Steps:

(1) randomly choose k objects from D as the initialcluster centers;

(2) repeat

(3) (re)assign each object to the cluster to which theobject is the most similar , based on the mean valueof the objects in the cluster;

(4) update the cluster means, i.e., calculate themean value of the objects for each cluster;

(5) until no change;


18/21

K-Medoids Algorithm

-To over come the limitations foundin k-means algorithm

-Cluster is represented by the costvalue of the objects in the cluster.


19/21

Input:

k: the number of clusters,

D: a data set containing n objects.

Output:A set of k clusters.

Steps:

(1) randomly choose k objects in D as the initial

representative objects or seeds;

(2) repeat

(3) assign each remaining object to the cluster with thenearest representative object;

(4) randomly select a nonrepresentative object, orandom;

(5) compute the total cost, S, of swapping representativeobject, oj, with orandom;

(6) if S < 0 then swap oj with orandom to form the new setof k representative objects;

(7) until no change;


20/21

P : non represntative objectOj : current representative object

Oi : another representative objectOrandom : good replacement object (nonrepresetative object)


21/21

Thank You

Web Mining Final

Documents

Transcript of Web Mining Final