1 Reply Latest reply on Jan 10, 2018 10:16 PM by Intel Corporation

    encoding in Classification


      Can I use Principal Component Analysis, Factor Discriminant Analysis or Linear Discriminant Analysis instead of one hot encoding in classification? what is the different between the 4 models?

        • 1. Re: encoding in Classification
          Intel Corporation
          This message was posted on behalf of Intel Corporation

          Hi Mohamed,

          Whereas One-Hot Encoding is used to convert categorical variables into a number which can be understood by the ML algorithm.
          LDA PCA and FA are used for dimensionality reduction in your feature set. So you cannot use these in place of one-hot encoding.
          The differences between LDA, PCA & FA are:
          Linear Discriminant Analysis(LDA) uses information of classes to find new features in order to maximize the class separability (between class-variance). LDA works when the measurements made on independent variables for each observation are continuous quantities
          Principle Component Analysis(PCA) uses the variance of each feature and identifies the components (within variable variance)
          Factor Analysis(FA) builds the feature combinations based on differences rather than similarities. It tries to uncover the latent factors which account for the variance shared between the observed variables.