Table 1.

Comparison between classical statistics and machine learning methods.

	Classical Statistics	Machine Learning
Approach	Top-down (applied to data)	Bottom-up (extracted by data)
Model	Hypothesized by the researcher	Auto-defined
Power of analysis	Medium	Usually High
Accuracy	Medium	It could be superior or inferior to that of classical statistics
Reliability	The same data always provide the same results	The results are affected by the initialization of parameters
Type of relationships among variables	Often Linear, in general not complex	Complex relationships
Interpretability	Simple	More complex