Skip to content

Using our dataset to look for any groupings, trends, or other information that could help us pitch cryptocurrencies to a firm with the help of unsupervised machine learning

Notifications You must be signed in to change notification settings

Malvi1497/Cryptocurrencies

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 

Repository files navigation

Cryptocurrencies

The purpose of this analysis was to analyze various cryptocurrencies on the market and create a report based on their features for an investment company to propose an investment portfolio for their clients. The following methods were used for the analysis:

Preprocessing the database

Reducing the data dimension using Principal Component Analysis (PCA) Clustering cryptocurrencies using K-Means Visualizing classification results with 2D and 3D scatter plots Resources

Data Source: crypto_data.csv

Python 3.7.11, Jupyter Notebook 6.4.5 [using mlenv environment].

Results

There were a total of 577 tradable cryptocurrencies after preprocessing the database, reduce the data dimensions using PCA, and then cluster the cryptocurrencies using K-Means.

Finding the best value for k using an elbow curve:

elbow_curve

image

As presented in the elbow curve above, the k value is 4. This k value in inputted into a k-means model to predict cryptocurrencies into 4 clusters.

3-D scatter plot with clusters

3d_graph

image

As presented in the 3-D graph above, using the PCA algorithm, the crytocurrencies dimensions were reduced to three principal components (PC1, PC2, PC3).

2-D plot with clusters

2d_plot

image

As presented in the 2-D scatter plot above, using the PCA algorithm, the crytocurrencies dimensions was reduced to two principal components.

Plotting the scatter plot from two cryptocurrency features does not provide a clear visual of the different classes. The data points are hard to visually decipher as compared to the 3-d scatterplot, which is more visually easier to comprehend using the PCA algorithm.

Summary

With a total of 577 classifications of cryptocurrencies identified based on the similarities of their features, specific features of each classification group should further be analyzed to ensure the portolio proposed from these analyses is as robust as possible for the best interest of the clients of the investment bank.

About

Using our dataset to look for any groupings, trends, or other information that could help us pitch cryptocurrencies to a firm with the help of unsupervised machine learning

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published