Classification by Learning Method
Have a training set (
Contains data and its classification) and test sets (
There is no classification of data)
From binary classification (
support vector machine) to multivariate classification
The results of machine learning will be rewarded and punished accordingly.
The goal of machine learning is to maximize rewards (
Similar to operations research)
There is no training set and no reward and punishment.
Mainly by clustering (
K- means, Gaussian mixture, hierarchical clustering, maximum expectation algorithm) and dimension reduction (
Factor analysis, principal component analysis, independent component analysis, hidden Markov model) two main categories
Data mining algorithms mainly include classification, clustering, prediction and association. Classification belongs to supervised learning and clustering belongs to unsupervised learning.
Big Data Analytics Beyond Hadoop