Pattern recognition is the process of efficientlydetecting any patterns or regularities in the given data. Clustering is anexample of unsupervised machine learning while classification is supervisedlearning. The processes can be parametric where in the data is summarised by aset of parameters or can be non parametric.
Linear discriminant analysis, a parametricclassification algorithm is used in testing the significance of gene pathwayand gene network models.Classification assigns instances to predefined classes based on features.It analyses and learn association between the features from the training datato classify the unknown variables. The common classification technique,decision tree, divides the search space into subsets using divide and conquertechnique. Giving grades for students is a simple classification problem. Inreality classification is teaching computer to do classification from thederived knowledge. Linear regression is simple classification methods where inrelationship between observed variables are modeled 2.
The input data are categorized into training data and test data. Trainingdata comprises of representative data from a known category and the test datais unknown data. A feature extractor is used to extract features from inputdata. Features are the parameters or explanatory variables most relevant to theproblem extracted from observations. It can be either categorical, ordinal,integer or real valued and is represented as a vector. When applied inbioinformatics the vector consist of frequency of nucleotides such as A, T, G,C or its 2-mer, 3-mer etc.
Dimensionality reduction techniques are implementedto reduce the number of features. Feature selection is another pre processing methods used to filterfeatures to remove unwanted and redundant data and include most relevant orquality data to produce reliable output. A trainer/classifier, implements anyof the clustering or classification algorithm and maps input to thecorresponding class. The whole process is represented in the diagram givenbelow.