Table 6.

Advantages and limitations of different classification models.

Methods	Advantages	Limitations
k-Nearest Neighbor	-Easy to understand and easy to implement -Training is very fast -Robust to noisy training data -It is particularly well suited for multimodal classes	-It is sensitive to the local structure of the data -Memory limitation -Being supervised learning lazy Algorithm e.g., runs slowly
Neural Network	-Efficiently handles noisy inputs -Computational rate is high -When an element of the neural network fails, it can continue without any problem with their parallel nature	-It is semantically poor -Difficult in choosing the type of network architecture -Requires high processing time for large neural networks
Gaussian Mixture Model	-Training is very fast -Performs well on the data of different size and densities	-The result is not stable -Sensitive to violations and distributional assumptions
Hidden Markov Model	-Convenient for modeling sequential data -Learning can take place directly from raw data	-Often has a large number of unstructured parameters -Unable to capture higher order correlation
Decision Tree	-Requires little data preparation -Nonlinear relationships between parameters do not affect tree performance -Easy to interpret and explain -Performs well with large data in a short time	-Complexity -Possibility of duplication with the same sub-tree on different paths
Support Vector Machine	-Produces very accurate classifiers -Less over-fitting, robust to noise -Especially popular in text classification problems where very high-dimensional spaces are the norm -Memory-intensive	-Requires both positive and negative examples -Needs to select a good kernel function -SVM is a binary classifier. To do a multi-class classification, pair-wise classifications can be used (one class against all others, for all classes) -There are some numerical stability problems in solving the constraint, QP (Quadratic programming) -Computationally expensive, thus runs slow
Self-Organizing Map	-Simple and easy-to-understand -A topological clustering unsupervised algorithm that works with a nonlinear data set -The excellent capability to visualize high- dimensional data onto 1 or 2-dimensional space makes it unique especially for dimensionality reduction	-Time consuming algorithm
k-Means	-Low complexity	-Necessity of specifying k -Sensitive to noise and outlier data points -Clusters are sensitive to the initial assignment of centroids
Fuzzy Measure	-Efficiently handles uncertainty -Properties are described by identifying various stochastic relationships -Allows a data point to be in multiple clusters	-Without prior knowledge, the output is not good -Precise solutions depend upon the direction of decision
Expectation-Maximization Meta	-Can easily change the model to adapt to a different distribution of data sets -Parameters number does not increase with the training data increasing	-Slow convergence in some cases
Bayesian Classifier	-Improves the classification performance by removing the irrelevant features -Good performance -Short computational time	-Information theoretically infeasible -Computationally infeasible