Unsupervised Machine Learning - Cryptocurrency Clusters

Background

• One of my clients was interested in offering a new cryptocurrency investment portfolio for its customers. The company, however, was lost in the vast universe of cryptocurrencies. They’ve asked me to create a report that includes what cryptocurrencies are on the trading market and determine whether they can be grouped to create a classification system for this new investment.

• I took raw data and processed it to fit the machine learning models. Unsupervised machine learning was used since there was no known classification system. I used several clustering algorithms to explore whether the cryptocurrencies could be grouped together with other similar cryptocurrencies. I used data visualization to share my findings with client.

Data Preparation

• Read crypto_data.csv into Pandas. The dataset was obtained from CryptoCompare.

• Discarded all cryptocurrencies that are not being traded. Filtered for currencies that are currently being traded. Dropped the IsTrading column from the dataframe.

• Removed all rows that have at least one null value.

• Filtered for cryptocurrencies that have been mined. That is, the total coins mined should be greater than zero.

• In order for the dataset to be comprehensible to a machine learning algorithm, its data should be numeric. Since the coin names do not contribute to the analysis of the data, I deleted the CoinName from the original dataframe.

• Next I converted the remaining features with text values, Algorithm and ProofType, into numerical data. I used Pandas to create dummy variables.

• Last I standardize the dataset so that columns that contain larger values do not unduly influence the outcome.

Dimensionality Reduction

• I created a dummy variables above which dramatically increased the number of features in the dataset. I performed dimensionality reduction with PCA. Rather than specify the number of principal components when I instantiated the PCA model, it is possible to state the desired explained variance. I preserved 90% of the explained variance in dimensionality reduction.

• I further reduced the dataset dimensions with t-SNE and visually inspected the results. I ran t-SNE on the principal components: the output of the PCA transformation. Then created a scatter plot of the t-SNE output.

Cluster Analysis with k-Means

• I used an elbow plot to identify the best number of clusters. A a for-loop was used to determine the inertia for each k between 1 through 10. The elbow plow appears at which value of k it appears.

• Based on Elbow curve using t-sne data we could divide cryptocurrencies data into 5 clusters