Data mining as defined by Han Kamber and Pei in their book "Data Mining
Concepts and Techniques" as:
" Data mining is the process of discovering interesting patterns and knowledge from large amounts of data. The data sources can include databases, data warehouses, the Web, other information repositories, or data that are streamed into the system dynamically".
The definition mainly differentiates in that it does not necessarily includes the word learning in it and usually is associated in scope with the structured or semi structured sources of information.
The difference between machine learning and patter recognition may not be so straight forward, especially if you look at Dudas or Theodoris book, yet if you examine closely each of these two books have specific chapters that sets them apart from machine learning: these are mainly:
* Dimensionality reduction- some of the algorithms in this category do not require learning.
* Template matching- this topic also does not include learning . If you look in Duda's book it talks in chapter 8 about Boyer-Moore string matching and parsing.These topics can be done by learning algorithms but in general are not.
While there are more subtle arguments to differentiate them these two examples should give you the idea.
Data mining as defined by Han Kamber and Pei in their book "Data Mining
Concepts and Techniques" as:
" Data mining is the process of discovering interesting patterns and knowledge from large amounts of data. The data sources can include databases, data warehouses, the Web, other information repositories, or data that are streamed into the system dynamically".
The definition mainly differentiates in that it does not necessarily includes the word learning in it and usually is associated in scope with the structured or semi structured sources of information.
The difference between machine learning and patter recognition may not be so straight forward, especially if you look at Dudas or Theodoris book, yet if you examine closely each of these two books have specific chapters that sets them apart from machine learning: these are mainly:
* Dimensionality reduction- some of the algorithms in this category do not require learning.
* Template matching- this topic also does not include learning . If you look in Duda's book it talks in chapter 8 about Boyer-Moore string matching and parsing.These topics can be done by learning algorithms but in general are not.
While there are more subtle arguments to differentiate them these two examples should give you the idea.
For me, pattern recognition focuses a bit more on how to represent the data (with feature vectors, graphs, etc) and what this representation means for different classifiers, whereas machine learning focuses a bit more on how to efficiently learn a classifier from as much data as possible
Pattern recognition is part of machine learning as it deals with specific ML algorithms like neural networks to classify patterns like character (roman/arabic), fingerprints etc to achieve high accuracy/recognition rate. Machine learning offers a wide range of algorithms like fuzzy/rough set/neural networks (those classical ones) and moving to bio-inspired range like ant/bats etc. Data mining is pretty straightforward, it applies ML algorithms to solve its mining problems like classification/prediction/clustering/rule mining/anomaly detection. Then there is another school of statistical-based algorithms like naïve bayes/bayesian networks etc.
Data mining, machine learning, and pattern recognition arose from different fields. Loosely speaking, machine learning grew out of artificial intelligence, pattern recognition out of signal/image processing and data mining out of the database community.
Hence, they had/have somewhat different focuses as well as somewhat different terminology (which can lead to some havoc). That said, there has been a large amount of interchange between the different discipline, which means many techniques are seen in the various groups.
Pattern Recognition use machine learning techniques but not necessarily always involve learning whereas in machine learning, learning is always the aim.
Data mining is about extracting or finding useful data or patterns either through classification or predictions or associations or recognition or pattern recognition or patterns matching.
Machine learning are the techniques that do the above tasks e.g. classification, predictions.... so data mining using all machine learning techniques. .