Please help with this

Given 100 features, you have to select 10 features for a 3 class classification problem. What is your approach?

1 Like

Two questions:

  • How big is the data set (how many examples do you have)?
  • Why do you need to select only 10 features when you have 100 available?
2 Likes

If you need to get best 10 feature columns out of 100 columns below I have listed few technique which can be used to select 10 features:

Principal Component Analysis (PCA): Transform the data into a lower-dimensional space using PCA to capture the most important patterns while reducing the number of features.

Correlation analysis: Identify and keep the features that have strong correlations with the target variable while eliminating highly correlated features among themselves.

Univariate feature selection: Evaluate each feature’s relationship with the target independently and select the top 10 most relevant ones based on statistical tests like chi-square or ANOVA.

1 Like