|
Methods | Advantages | Disadvantages | Literatures |
|
Logistic regression | Multifunction | Needing of sample size | Luo et al. [46] |
|
Bayesian networks | Utilization of incomplete and inaccurate data | Needing of preceding researches as guidance | Qu et al. [33] |
|
Rough sets theory | Without priori information; simplicity; handling ambiguous and uncertain information | Needing of self-development | Zhang et al. [28] |
|
Association rules mining | Supporting indirect data mining | Nonselectivity; subjectivity | Wu et al. [26] |
|
Set pair analysis | Suitability for changing systems | Handicap in handle relatively precise problems | Li et al. [45] |
|
Structural equation modeling | Analyzing the causality between the latent variables | Needs of 200 samples at least | Chen et al. [44] |
|
Cluster analysis | Minimization errors caused by subjective judgment | Too much calculation; handicap in clustering data with multidimensions and multilevel | Gu et al. [30] |
|
Decision trees | Handling in nonnumeric data; Simplicity | Maybe misleading | Zhong et al. [35] |
|
Principal component analysis | Dimension reduction; holism | Less specificity | Lu et al. [39] |
|
Partial least squares method | Specificity | Handicap in deciding principal component | Van Wietmarschen et al. [17] |
|
Artificial neural network | Simplicity; nonlinear | Handicap in obtaining the hidden information | Sun et al. [37] |
|
Entropy cluster algorithm | Little demand on variances’ types; analysis on any statistical dependence of the variances | Needing of self-development | Wang et al. [47] |
|
Factor analysis | Correction capability; views to latent variables | Absence of domination and relationship between primary and secondary | Wang et al. [42] |
|
Support vector machine | Classification without representing the feature space explicitly | Expressing the more complex prior information; analyzing limited samples | Yang et al. [48] |
|