Data mining techniques and applications

2. Data Mining Algorithms and Techniques
Various algorithms and techniques like Classification, Clustering, Regression, Artificial 
Intelligence, Neural Networks, Association Rules, Decision Trees, Genetic Algorithm, Nearest Neighbor 
method etc., are used for knowledge discovery from databases.
2.1. Classification 
Classification is the most commonly applied data mining technique, which employs a set of pre-classified 
examples to develop a model that can classify the population of records at large. Fraud detection and credit-
risk applications are particularly well suited to this type of analysis. This approach frequently employs 
decision tree or neural network-based classification algorithms. The data classification process involves 
learning and classification. In Learning the training data are analyzed by classification algorithm. In 
classification test data are used to estimate the accuracy of the classification rules. If the accuracy is 
acceptable the rules can be applied to the new data tuples. For a fraud detection application, this would 
include complete records of both fraudulent and valid activities determined on a record-by-record basis. 
The classifier-training algorithm uses these pre-classified examples to determine the set of parameters 
required for proper discrimination. The algorithm then encodes these parameters into a model called a 
classifier. 
Types of classification models: 

Classification by decision tree induction 

Bayesian Classification 

Neural Networks 

Support Vector Machines (SVM) 

Classification Based on Associations 
2.2. Clustering 
Clustering can be said as identification of similar classes of objects. By using clustering techniques we can 
further identify dense and sparse regions in object space and can discover overall distribution pattern and 
correlations among data attributes. Classification approach can also be used for effective means of 
distinguishing groups or classes of object but it becomes costly so clustering can be used as preprocessing 
approach for attribute subset selection and classification. For example, to form group of customers based on 
purchasing patterns, to categories genes with similar functionality. 
Types of clustering methods 
