Cryptocurrency Clustering Analysis

This project aims to analyze and cluster cryptocurrencies using Python and unsupervised machine learning to predict if cryptocurrencies are affected by 24-hour or 7-day price changes. As well as utilizing K-means clustering, employing data normalization and dimensionality reduction techniques such as Principal Component Analysis (PCA). The Cryto_Clustering notebook file with interactive visuals can be easily viewed by clicking on this link: [https://nbviewer.org/github.com/thaychansy/CryptoClustering/blob/main/Crypto_Clustering.ipynb]

Table of Contents

• Cryptocurrency Clustering Analysis
  • Table of Contents
  • Prepare the Data
  • Find the Best Value for k Using the Scaled DataFrame
  • Cluster Cryptocurrencies with K-means Using the Scaled DataFrame
  • Optimize Clusters with Principal Component Analysis
  • Find the Best Value for k Using the PCA DataFrame
  • Cluster Cryptocurrencies with K-means Using the PCA DataFrame
  • Conclusion
  • Contact

Prepare the Data

To normalize the data from the CSV file, we use the StandardScaler() module from scikit-learn. We create a DataFrame with the scaled data and set the "coin_id" from the original DataFrame as the index for the new DataFrame.

from sklearn.preprocessing import StandardScaler
import pandas as pd

# Load data
df_market_data = pd.read_csv(
    "Resources/crypto_market_data.csv",
    index_col="coin_id")

# Normalize data
df_market_data_scaled = StandardScaler().fit_transform(df_market_data[['price_change_percentage_24h', 'price_change_percentage_7d', 
                                                                    'price_change_percentage_14d', 'price_change_percentage_30d', 
                                                                    'price_change_percentage_60d', 'price_change_percentage_200d',
                                                                    'price_change_percentage_1y']])

# Create scaled DataFrame
df_market_data_scaled = pd.DataFrame(price_change_scaled, columns=['price_change_percentage_24h', 'price_change_percentage_7d', 
                                                                    'price_change_percentage_14d', 'price_change_percentage_30d', 
                                                                    'price_change_percentage_60d', 'price_change_percentage_200d',
                                                                    'price_change_percentage_1y'])

Find the Best Value for k Using the Scaled DataFrame

Using the elbow method, we determine the best value for k through the following steps:

1. Create a list of k values from 1 to 11.
2. Compute the inertia for each value of k.
3. Plot the elbow curve.

# Initialize lists
k_values = list(range(1, 12))
inertia = []

# Compute inertia for each k
for i in k:
    k_model = KMeans(n_clusters=i, random_state=0)
    k_model.fit(df_market_data_scaled)
    inertia.append(k_model.inertia_)

# Plot elbow curve
df_elbow.hvplot.line(
    x="k", 
    y="inertia", 
    title="Elbow Curve", 
    xticks=k
)

Cluster Cryptocurrencies with K-means Using the Scaled DataFrame

Using the best k value, we cluster the cryptocurrencies as follows:

1. Initialize and fit the K-means model.
2. Predict clusters and add them to the DataFrame.
3. Create a scatter plot.

# Initialize the K-Means model using the best value for k
model = KMeans(n_clusters=4, random_state=1)

# Fit the K-Means model using the scaled data
model.fit(df_market_data_scaled)

# Predict the clusters to group the cryptocurrencies using the scaled data
k_4 = model.predict(df_market_data_scaled)

# Plot
price_change_scaled_predictions_df.hvplot.scatter(
    x="price_change_percentage_24h",
    y="price_change_percentage_7d",
    by="crypto_segment",  # Color points by the cluster labels from K-Means
    hover_cols=["coinid"],  # Add the cryptocurrency name to hover information
    title="Crypto Segmentation based on K-Means Clustering (k=4)"
)

Optimize Clusters with Principal Component Analysis

We perform PCA to reduce features to three principal components.

1. Fit PCA and retrieve explained variance.
2. Create a new DataFrame with PCA data.

# Create a PCA model instance and set `n_components=3`.
pca = PCA(n_components=3)

# Use the PCA model with `fit_transform` to reduce to 
# three principal components.
crypto_pca = pca.fit_transform(df_market_data_scaled)

# Retrieve the explained variance to determine how much information 
# can be attributed to each principal component.
pca.explained_variance_ratio_

# Create a new DataFrame with the PCA data.
crypto_pca_df = pd.DataFrame(
    crypto_pca,
    columns=["PCA1", "PCA2", "PCA3"]

# Set the coinid column as index
crypto_pca_df.set_index('coinid', inplace=True)
)

Find the Best Value for k Using the PCA DataFrame

Use the elbow method on the PCA DataFrame as previously described to determine the best k value.

Best Value for k with PCA is k = 4.

# Create a list with the number of k-values from 1 to 11
k = list(range(1, 12))

# Create a list with the number of k-values from 1 to 11
k = list(range(1, 12))

# Create an empty list to store the inertia values
inertia = []

# Create a for loop to compute the inertia with each possible value of k
for i in k:
    k_model = KMeans(n_clusters=i, random_state=0)
    k_model.fit(crypto_pca_df)
    inertia.append(k_model.inertia_)

# Plot
df_elbow_pca.hvplot.line(
    x="k",
    y="inertia",
    title="Elbow Curve PCA",
    xticks=k
)

Cluster Cryptocurrencies with K-means Using the PCA DataFrame

Repeat the clustering process using the PCA DataFrame.

Initialize K-means with the best k value from PCA.

1. Fit and predict clusters.
2. Create a scatter plot.

# Initialize K-means with best k from PCA
model = KMeans(n_clusters=4, random_state=0)
model.fit(crypto_pca_df)

# Predict the clusters to group the cryptocurrencies using the PCA data
k_4 = model.predict(crypto_pca_df)

# Create a copy of the DataFrame with the PCA data
crypto_pca_predictions_df = crypto_pca_df.copy()

# Add a new column to the DataFrame with the predicted clusters
crypto_pca_predictions_df["cryto_segments"] = k_4

# Plot
crypto_pca_predictions_df.hvplot.scatter(
    x="PCA1",
    y="PCA2",
    by="cryto_segments",
    hover_cols=["coinid"],  # Add the cryptocurrency name to hover information
    title="PCA Crypto Segmentation based on K-Means Clustering (k=4)"
)

Conclusion

This analysis provides insights into the clustering of cryptocurrencies using both the scaled data and the PCA-transformed data. The results can help in identifying trends and making informed investment decisions. The cryptocurrency clustering project successfully applied unsupervised learning techniques, specifically K-means clustering, to group cryptocurrencies based on their price change patterns over various time periods.

Contact

Thay Chansy - @thaychansy - or [email protected]

Please visit my Portfolio Page: thaychansy.github.io (https://thaychansy.github.io/)