Hello researchers,
I have a dataset that contains 36733 instances, What are the engineering methods to be used for feature selection in machine learning?
Have a look at https://machinelearningmastery.com/feature-selection-with-real-and-categorical-data/
Article A review of feature selection methods on synthetic data
http://romisatriawahono.net/lecture/rm/survey/machine%20learning/Chandrashekar%20-%20Feature%20Selection%20Methods%20-%202014.pdf
Samar Taha Yousif Feature selection can be performed either via attribute selection or subset selection.
Attribute selection involves selecting each attribute for assessment using a metric following which a ranking algorithm is used to rank their importance.
Subset selection involves a comparison of the assessment of all possible subset of attributes and then recommending the subset with the largest value.
Feature selection can also be assessed as either Filter-based or Wrapper-based.
In the filter-based approach, the relevant attributes are selected based on those attributes with have low intercorrelation with other attributes and higher correlation with the target class.
In the wrapper-based approach, the relevant attributes are selected based on those attributes which have the highest accuracy in predicting the target class based on the performance of a machine learning algorithm of interest.
If it is a classification problem, one strategy would be to choose an initial set of features based on research hypothesis and then use a plot to observe class discrimination based on the selected features. Features might be iteratively modified to improve classification results, as measured using a confusion matrix. For multi-dimensional feature vectors PCA might be used for generating 2-D plots.
Dear Samar Taha Yousif,
Before the methods to be applied in ML, which will depend exclusively on your strategy and your application of these data, is it necessary to know if they are all valid data?
If so, you can choose an ML system that you find most appropriate for your application. Or try artificial neural networks (MLP)?
Good luck!
Wiltgen
Feature selection methods are intended to reduce the number of input variables to those that are believed to be most useful to a model in order to predict the target variable. Feature selection is primarily focused on removing non-informative or redundant predictors from the model.
https://machinelearningmastery.com/feature-selection-with-real-and-categorical-data/#:~:text=Feature%20selection%20methods%20are%20intended,redundant%20predictors%20from%20the%20model.
For feature selection in Machine learning, you can use statistical methods and correlation coefficients to select features that contribute to the output/target value (in supervised learning), but not to have many highly correlated features. You can also perform Random Forest to estimate Features importance.
Samar Taha Yousif, you have asked for suggestions on good feature selection methods. On the other hand, you are mentioning the number of instances in your dataset. Can you first clarify if you are trying to reduce the features or the instances?
- Badges
- Science topic