AI Core

Position Clustering Process

Cluster students into four categories based on their course, moral and behavioral characteristics features and assigns a label between A to F on them.

Here I used a Variational Autoencoder to extract the latent space of features into two dimensions in order to cluster them using K-Means Clustering algorithm.

Besides, I also used the PCA algorithm to reduce the size of course features into two dimensions for clustering(I call this part raw clustering).

Position Classification Process

Classify students into four classes using a pre-trained classifier on both latent space and raw clustering of last part.

Here I trained a Neural Network classifier on clustered students in the last part to assigns the final position on each student based on mentioned features.

I also trained the classifier on:

• clustered the extracted latent space of VAE
• clustered the reduced dimensions of student features using PCA

So far the results of the classification on the second training seemed to be more acceptable.