We were approached by Accountability Accounting, an investment firm, studying the prospects of new cryptocurrency investments. The dramatic rise and fall of BitCoin has made the endeavor seem initially contrived.
To evaluate the various types of currencies that exist, they requested us to perform an in-depth analysis of the cryptocurrencies that are on the trading market, creating a dataset that will classify and show profitability of each of these cryptocurrencies, and to display the data as tables and graphs that can reveal tendencies.
These were our steps:
-
Preprocessing the data for PCA
-
Using get_dummies() function to create variables for text features:
-
Standardizing data with StandardScaler():
-
-
Reducing Data Dimensions Using PCA
-
Creating a DataFrame with three principal components:
-
First we found the best value for K using the Elbow Curve as seen below:
Subsequently, we initialized the K-Means model that predicted the clusters:
Finally, we created a new DataFrame that contains the predicted clusters, along with the cryptocurrency features, seen here partially:
We created visualizations of our results.
-
3D-Scatter with Clusters
-
Table with tradable cryptocurrencies
We are able to predict that there are 532 tradable cryptocurrencies.
We scaled data using MinMaxScaler() to create a scatterplot with tradable cryptocurrencies:
