LDA creates predictions by calculating the likelihood that a fresh set of inputs falls into each of the classes. The output class is the one with the highest probability, and a forecast is produced.
The goal of LDA is to find a feature subspace that maximises group separability. While principal component analysis is an unsupervised Dimensionality reduction technique, it does not take into account the class label. PCA focuses on capturing the data set's highest variation direction.
The likelihood that a fresh set of inputs belongs to each class is estimated using linear discriminant analysis. The output class with the highest probability is chosen. Here's how Bayes' theorem estimates the chance that the data belongs to each class if the output class is (k) and the input is (x).