Using unsupervised machine learning to determine which cryptocurrencies are good investments
The aim of this project was to group the cryptocurrencies in order to determine the products in which to invest using data tables and data visualizations
The input to this analysis is the crypto_data.csv file which contains the following columns:
| Column | Type |
|---|---|
| Unnamed: 0 | object |
| CoinName | object |
| Algorithm | object |
| IsTrading | bool |
| ProofType | object |
| TotalCoinsMined | float64 |
| TotalCoinSupply | object |
Please refer to the notebook for a detailed explanation of the processing.
In order to determine how many clusters to use, the following elbow curve was created.
Based on this curve, it was decided that the appropriate number of clusters is 4.
After the determination to use 4 clusters was made, the following 3-D scatter plot was created to provide a visual of the clustering:
Another visual to display the clustering was the following 2-D scatter plot:
Yet another method to display the clustering of stocks was the data table which comprised the following columns:
- CoinName
- Algorithm
- ProofType
- TotalCoinSupply
- TotalCoinsMined
- Class
The Class column identifies the cluster to which the coin belongs.
It would appear that splitting up the coins into four clusters was quite appropriate. The clusters fall into the following categories:
- High supply, low demand
- Medium supply, low demand
- Low supply, low demand
- High supply, high demand
Beyond these statistics, information about the companies issuing those currencies and a price history of those currencies would be very instructive in determining the cluster of currencies into which it would be most prudent to invest.