| Regression analysis |
It can find regression equations and predict dependent variables |
Deeply developed and widely used in many occasions |
Needs large amounts of data and may cause overfitting in practical applications |
Machine learning with systematic density-functional theory calculations: application to melting temperatures of single-and binary-component solids |
| Naïve Bayes classifier |
It can classify data into several categories following the highest possibility |
Only a small amount of data is needed to obtain essential parameters |
The feature independence hypothesis is not always accurate |
A naïve-Bayes classifier for damage detection in engineering materials |
| Support vector machine |
SVM can find a hyperplane to divide a group of points into two categories |
It has great generalization ability and can properly handle high-dimension datasets |
SVM is not very appropriate for multiple classification problems |
PVP-SVM: sequence-based prediction of phage virion proteins using a support vector machine |
| Decision tree and random forest |
By splitting source datasets into several subsets, all data will be judged and classified |
The calculating processes are easy to comprehend. Also, it can handle large amounts of data |
It is difficult to obtain a high-performance decision tree or a random forest. Also, the overfitting problem may occur |
High-throughput machine-learning-driven synthesis of full-Heusler compounds |
| Artificial neural network |
By imitating neuron activities, ANN can automatically find underlying patterns in inputs |
ANN has great self-improving ability, great robustness and high fault tolerance |
Its inner calculation progresses are very difficult to understand |
Learning from the Harvard Clean Energy Project: the use of neural networks to accelerate materials discovery |
| Deep learning |
Originated from ANN. It aims to build a neural network to analyze data by imitating the human brain |
It has the best self-adjusting and self-improving abilities compared with other ML methods |
As a new trend in ML, deep learning has not yet been well studied. Many defects are still unclear |
Artificial intelligence in neuropathology: deep learning-based assessment of tauopathy |
