The purpose of this project was to utilize unsupervised machine learning to evaluate a cryptocurrency database and provide a report that categorizes traded cryptocurrencies into groups based on their characteristics. First, I preprocess the data and utilize principal component analysis to reduce data dimension. Then, I utilize K-Means with the aid of elbow method in order to cluster the cryptocurrencies. Finally, I visualize the classification results in 2D and 3D scatter plots.

After preprocessing the data, there is a total of 533 cryptocurrencies. I use standard scaler on the data and use three principal component on the existing data. The elbow curve shows that aggregating by 4 clusters is the best approach.

A 3d scatter plot was deployed in order to display the PCA and the clusters.

As it can be seen by the 2D graph, clustering by two features does not effectively separate the classes. However, The 3D plot was able to distinguish these clusters properly.