What is Data Mining?
Data mining is a poorly understood term. As a
consequence of this misunderstanding there is
prejudice and resistance to its use commercially,
although this is diminishing as organisations
begin to understand first-hand how analytics benefits
them.
Data mining is a process, not a
technique. Consequently at Offlode
we use the terms analytics and data mining interchangeably.
Data mining practitioners will use a collection
of techniques from a variety of fields such
as statistics, mathematics, machine learning,
and operations research. The goal of data mining
is to extract useful, but previously unknown
information, from typically massive collections
of non-experimental, sometimes non-traditional
data.(Mackinon et all, 1999)
Techniques used in data mining generally perform
descriptive or predictive
modelling. The most appropriate technique
for any given problem is governed by a number
of factors including the purpose of the exercise,
nature and/or quality of the data, time available,
intended use of results, associated costs and
eventual deployment. Techniques come from two
broad groups, supervised and unsupervised learning.
With supervised learning
there exists some outcome that is modelled.
The model seeks to describe or predict the outcome
of interest. Examples of this are predicting
which credit card transactions are fraudulent,
predicting which customers will respond to a
marketing campaign and forecasting price of
a used car in two years time. Interpretation
of the model forms the descriptive component
of the modelling process.
Unsupervised learning seeks
to summarise a collection of behaviours. Typically
this is in the form of clustering or dimension
reduction. Clustering tries to identify groups
with similar behaviours, such as customers who
use their credit card in a similar fashion or
business who use certain technology products.
Dimension reduction aims to summarise the key
elements of behaviour; for instance, summarising
a collection of documents describing a project
to indicate the key phrases that indicate if
a project has been successful.
