The article evaluates the performance of 15 feature selection methods across varying data sizes, focusing on r-squared values. Findings indicate that Var methods exhibit superior performance, followed by stepwise and correlation methods, while the lasso method lags behind. The study emphasizes how r-squared values correlate positively with sample size, showcasing overall better performance at higher sample percentages. Notably, Euclidean distance and DTW methods rank well among similarity methods, while edit distance performs poorly. Visual representations demonstrate these trends and underlying relationships among different methods.
The evaluation of feature selection methods revealed that Var methods outperformed others, while the lasso method showed significantly lower performance across varied sample sizes.
The study highlights that r-squared values generally increased with sample size, but all methods performed similarly at higher sample percentages.
Collection
[
|
...
]