A Financial Advisory firm is looking for a new way of constructing investment portfolios based on cryptocurrencies.
This is a prototype to trial an approach that considers not only returns and volatility, but also other factors that might impact the crypto market. The prototype is presented as a Jupyter Lab notebook.
- Price change data of some cryptocurrencies in different periods: crypto_market_data.csv
- Importing and preparing the data.
- Find the best value for k by using the original data.
- Clustering of cryptocurrencies with K-means by using the original data.
- Optimising the clusters with principal component analysis.
- Finding the best value for k by using the PCA data.
- Clustering the cryptocurrencies with K-means by using the PCA data.
- Plotting the results to compare the performance of the clusters visually.
This tool utilises the following technologies:
- Pandas DataFrame: Documentation
- hvplot Bar chart, Line plot, Scatter: Documentation
- sklearn Cluster, decomposition, preprocessing: Documentation
Please be aware this is an Academic Case Study. The conclusions from this work should not be considered as financial advice.
Since we can only plot 2 dimensions of the data (PCA 1 and PCA 2), we are not able to get a lot of meaningful information. However, the Cryptocurrency segmentation is clearer when using PCA.