This project demonstrates a method to automatically choose the number of clusters for clustering geographical coordinates using Gaussian Mixture Models (GMM) and K-means algorithm. It provides visualizations of the cluster distribution and cluster visualization on a map.
- pandas
- matplotlib
- scikit-learn
- numpy
- scipy
-
Ensure that the
coordinates.pklfile is present in the same directory as the code file. -
Run the code to perform clustering on the coordinates.
-
The method will automatically select the optimal number of clusters based on Bayesian Information Criterion (BIC).
-
The resulting clusters will be added to the
coordinatesdataframe. -
The number of clusters will be displayed.
-
The distribution of clusters will be plotted in a bar chart.
-
The clusters will be visualized on a map.
- https://scikit-learn.org/stable/modules/generated/sklearn.mixture.GaussianMixture.html
- https://scikit-learn.org/stable/modules/generated/sklearn.cluster.KMeans.html
- https://www.datacamp.com/tutorial/k-means-clustering-python
# Load the coordinates data from the pickle file
coordinates = pd.read_pickle('coordinates.pkl')
# Cluster the coordinates and automatically choose the number of clusters
coordinates, num_clusters = cluster_coordinates(coordinates)
# Display the number of clusters
print(f"Number of clusters: {num_clusters}")
# Plot the distribution of clusters
cluster_counts = coordinates['cluster'].value_counts()
plt.figure(figsize=(10, 6))
cluster_counts.plot(kind='bar')
plt.xlabel('Cluster')
plt.ylabel('Frequency')
plt.title('Distribution of Clusters')
plt.show()
# Visualize the clusters on a map
colors = ['green', 'blue', 'red', 'yellow']
plt.figure(figsize=(8, 8))
for cluster_id in range(num_clusters):
cluster_points = coordinates[coordinates['cluster'] == cluster_id]
plt.scatter(cluster_points['Composite:GPSLongitude'],
cluster_points['Composite:GPSLatitude'],
color=colors[cluster_id], label=f'Cluster {cluster_id}')
plt.xlabel('Longitude')
plt.ylabel('Latitude')
plt.title('Cluster Visualization')
plt.legend()
plt.grid(True)
plt.show()
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent