app.py

# -*- coding: utf-8 -*-
"""01_clustering_methods.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1mqAGInsaItbKYVUlP9muYz3fpdGBWFz5
"""



import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import sklearn.cluster as cluster


import colormaps as cmaps
import opinionated
plt.style.use("opinionated_rc")
from opinionated.core import download_googlefont
download_googlefont('Quicksand', add_to_cache=True)
plt.rc('font', family='Quicksand')

!wget https://github.com/scikit-learn-contrib/hdbscan/raw/master/notebooks/clusterable_data.npy
!wget https://github.com/mwaskom/seaborn-data/raw/master/penguins.csv

hdbscan_example_data = np.load('clusterable_data.npy')
penguins_dataset = pd.read_csv('penguins.csv')[['bill_length_mm','bill_depth_mm','flipper_length_mm']].dropna().values

from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()
penguins_dataset_standardized = scaler.fit_transform(penguins_dataset)













import gradio as gr
import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import make_blobs, make_moons, load_iris
import seaborn as sns
import pandas as pd
import matplotlib.colors as mcolors


from sklearn.cluster import KMeans
from sklearn.cluster import AgglomerativeClustering
from sklearn.mixture import GaussianMixture
import hdbscan


import genieclust





# Pre-defined datasets
blobs_X, _ = make_blobs(n_samples=300, centers=4, cluster_std=0.60, random_state=0)
moons_X, _ = make_moons(n_samples=300, noise=0.05, random_state=0)

# Penguins dataset (3D example)
# For the purpose of this example, let's simulate the Penguins dataset with iris for simplicity
iris_X, _ = load_iris(return_X_y=True)
# Assuming iris_X to be a placeholder for the Penguins dataset with numerical features

datasets = {
    "Blobs": blobs_X,
    "Moons": moons_X,
    "Penguins": penguins_dataset_standardized,  # Placeholder for Penguins dataset
    "hDBSCAN sample": hdbscan_example_data
}

# Function for plotting the unclustered dataset
def plot_unclustered(dataset_name):
    X = datasets[dataset_name]  # Fetch dataset from the dictionary

    # Check if the dataset has more than 2 dimensions
    if X.shape[1] > 2:
        # Convert dataset to DataFrame for seaborn pairplot
        df = pd.DataFrame(X)
        fig = sns.pairplot(df, plot_kws={'color': 'grey','alpha':0.7}, diag_kws={'color': 'grey'}).fig
    else:
        fig, ax = plt.subplots(figsize=(8, 6))
        ax.scatter(X[:, 0], X[:, 1], color='gray', marker='.',alpha=.7)
        ax.set_xlabel("Feature 1")
        ax.set_ylabel("Feature 2")
        ax.grid(True)
        plt.tight_layout()
        plt.close(fig)

    return fig

def plot_clustered(dataset_name, clustering_method, kmeans_n_clusters, agg_n_clusters, agg_linkage, gmm_n_clusters, covariance_type,
                   genie_n_clusters, gini_threshold, M,hdbscan_min_cluster_size, hdbscan_min_samples):
    X = datasets[dataset_name]

    # Determine the clustering method and fit the model accordingly
    if clustering_method == "K-Means":
        model = KMeans(n_clusters=kmeans_n_clusters)
        model.fit(X)
        labels = model.labels_  # For K-Means, labels are in .labels_

    elif clustering_method == "Agglomerative":
        model = AgglomerativeClustering(n_clusters=agg_n_clusters, linkage=agg_linkage)
        model.fit(X)
        labels = model.labels_  # For Agglomerative Clustering, labels are in .labels_

    elif clustering_method == "Gaussian Mixture":
        model = GaussianMixture(n_components=gmm_n_clusters, covariance_type=covariance_type)
        model.fit(X)
        labels = model.predict(X)  # For Gaussian Mixture, use .predict() to get labels

    elif clustering_method == "Genie":
        model = genieclust.Genie(n_clusters=genie_n_clusters, gini_threshold=gini_threshold, M=M)
        labels = model.fit_predict(X)  # GenieClust uses fit_predict directly for both fitting and label prediction

    elif clustering_method == "h-DBSCAN":
        clusterer = hdbscan.HDBSCAN(min_cluster_size=hdbscan_min_cluster_size, min_samples=hdbscan_min_samples).fit(X)
        labels = clusterer.labels_



    n_clusters= len(np.unique([x for x in labels if x >= 0]))


    if n_clusters <= 10:
        original_cmap = cmaps.greenorange_12
        colors = original_cmap([x for x in range(n_clusters)])
        # Create a new listed colormap with the extracted colors
        new_cmap = mcolors.ListedColormap(colors)
    else:
        new_cmap = cmaps.cet_g_bw_minc

    cluster_colors = [new_cmap(x) if x >= 0
                  else (0.5, 0.5, 0.5)
                  for x in labels]


    # Check if the dataset has more than 2 dimensions
    if X.shape[1] > 2:
        # Convert dataset to DataFrame for seaborn pairplot
        df = pd.DataFrame(X)
      # df['cluster'] = labels
      #  fig = sns.pairplot(df, color = cluster_colors, cmap=new_cmap).fig


            # Create bins for each variable
        n_bins = 10
        bins = {column: np.linspace(df[column].min(), df[column].max(), n_bins+1) for column in df.columns}

        # Create a figure and axes
        n = len(df.columns)
        fig, axes = plt.subplots(nrows=n, ncols=n, figsize=(n*2.3, n*2.3))

        for i in range(n):
            for j in range(n):
                ax = axes[i, j]
                ax.grid(True, which='both', linestyle='--', linewidth=0.5)

                if i != j:
                    ax.scatter(df[df.columns[j]], df[df.columns[i]], c=cluster_colors, alpha=0.8, marker='o',s = 10)
                else:  # Diagonal - Stacked Bar Charts
                    data = df[df.columns[i]]
                    counts = np.zeros((n_bins, n_clusters))
                    for cluster in range(n_clusters):
                        cluster_data = data[labels == cluster]
                        hist, _ = np.histogram(cluster_data, bins=bins[df.columns[i]])
                        counts[:, cluster] = hist
                    for cluster in range(n_clusters):
                        ax.bar(range(n_bins), counts[:, cluster], width=1, align='center',
                              bottom=np.sum(counts[:, :cluster], axis=1), color=cluster_colors[list(labels).index(cluster)] )

                # Explicit axis lines at the bottom and left
                ax.spines['top'].set_visible(False)
                ax.spines['right'].set_visible(False)
                ax.spines['bottom'].set_visible(True)
                ax.spines['left'].set_visible(True)

                # Hide axis marks for inner plots and adjust label size
                if i < n - 1:
                    ax.tick_params(labelbottom=False)  # Hide x-axis labels for all but bottom row
                else:
                    ax.tick_params(axis='x', labelsize=8)  # Smaller labels for x-axis
                if j > 0:
                    ax.tick_params(labelleft=False)  # Hide y-axis labels for all but first column
                else:
                    ax.tick_params(axis='y', labelsize=8)  # Smaller labels for y-axis

                # Set labels for outer plots only
                if i == n - 1:
                    ax.set_xlabel(df.columns[j], rotation=0, fontsize=12)
                if j == 0:
                    ax.set_ylabel(df.columns[i], fontsize=12)




    else:
        fig, ax = plt.subplots(figsize=(8, 6))
        ax.scatter(X[:, 0], X[:, 1], c=cluster_colors,  marker='.')
        ax.grid(True)
        plt.tight_layout()
        plt.close(fig)



    return fig

intro_md = """
   # Cluster-algorithm-explorer

    _by [Max Noichl](https://homepage.univie.ac.at/maximilian.noichl/), for the clustering & data-visualization-workshop, Bremen, 2024_

    Below you can test a number of clustering-algorithms on several easier and harder datasets.

    """



# Gradio interface setup remains the same
with gr.Blocks(theme=gr.themes.Monochrome()) as demo:
  with gr.Column():
    gr.Markdown(intro_md)
    with gr.Row():

        with gr.Column():
            gr.Markdown("# Choose your dataset:")
            dataset_dropdown = gr.Dropdown(label="Select a dataset", choices=list(datasets.keys()), value="Blobs")



            gr.Markdown("# Choose your Clustering algorithm & Parameters:")


            # Update the dropdown for clustering method to include "Genie"
            clustering_method_dropdown = gr.Dropdown(label="Select a clustering method", choices=["K-Means", "Agglomerative", "Gaussian Mixture", "Genie", "h-DBSCAN"], value="K-Means")

            # K-Means parameters
            with gr.Group(visible=True) as kmeans_params_group:
                kmeans_n_clusters_slider = gr.Slider(minimum=2, maximum=10, step=1, label="Number of Clusters (K-Means)", value=4)

            # Agglomerative Clustering parameters
            with gr.Group(visible=False) as agglomerative_params_group:
                agg_n_clusters_slider = gr.Slider(minimum=2, maximum=10, step=1, label="Number of Clusters (Agglomerative)", value=4)
                agg_linkage_dropdown = gr.Dropdown(label="Linkage Type", choices=["ward", "complete", "average", "single"], value="ward")

            # Gaussian Mixture Model parameters
            with gr.Group(visible=False) as gmm_params_group:
                gmm_n_clusters_slider = gr.Slider(minimum=2, maximum=10, step=1, label="Number of Components (GMM)", value=4)
                covariance_type_dropdown = gr.Dropdown(label="Covariance Type", choices=["full", "tied", "diag", "spherical"], value="full")

            # GenieClust parameters
            with gr.Group(visible=False) as genie_params_group:
                genie_n_clusters_slider = gr.Slider(minimum=2, maximum=10, step=1, label="Number of Clusters (Genie)", value=4)
                gini_threshold_slider = gr.Slider(minimum=0.0, maximum=1.05, step=0.05, label="Gini Threshold (Genie)", value=.3)
                M_slider = gr.Slider(minimum=0.5, maximum=2.0, step=0.1, label="M Parameter (Genie)", value=1.0)

            with gr.Group(visible=False) as hdbscan_params_group:
                hdbscan_min_cluster_size = gr.Slider(minimum=2, maximum=200, step=1, label="Minimal Cluster Size (hDBSCAN)", value=10)
                hdbscan_min_samples = gr.Slider(minimum=2, maximum=200, step=1, label="Min. Samples (hDBSCAN)", value=10)



            # Update the function that changes visible parameter groups based on selected clustering method
            def update_method_params(clustering_method):
                return {
                    kmeans_params_group: gr.Group(visible=clustering_method == "K-Means"),
                    agglomerative_params_group: gr.Group(visible=clustering_method == "Agglomerative"),
                    gmm_params_group: gr.Group(visible=clustering_method == "Gaussian Mixture"),
                    genie_params_group: gr.Group(visible=clustering_method == "Genie"),
                    hdbscan_params_group: gr.Group(visible=clustering_method == "h-DBSCAN"),


                }


            clustering_method_dropdown.change(update_method_params, inputs=[clustering_method_dropdown], outputs=[kmeans_params_group, agglomerative_params_group,
                                                                                                                  gmm_params_group, genie_params_group,hdbscan_params_group])

            button = gr.Button("Run Clustering!")


        with gr.Column():
            unclustered_plot_output = gr.Plot(label=None)
            clustered_plot_output = gr.Plot(label=None)


        dataset_dropdown.change(plot_unclustered, inputs=[dataset_dropdown], outputs=[unclustered_plot_output])
        demo.load(plot_unclustered, inputs=[dataset_dropdown], outputs=[unclustered_plot_output])
        # Update the button click event to include new parameters for GenieClust
        button.click(
            plot_clustered,
            inputs=[
                dataset_dropdown,
                clustering_method_dropdown,
                kmeans_n_clusters_slider,
                agg_n_clusters_slider,
                agg_linkage_dropdown,
                gmm_n_clusters_slider,
                covariance_type_dropdown,
                genie_n_clusters_slider,  # Add Genie parameters
                gini_threshold_slider,
                M_slider,
                hdbscan_min_cluster_size,
                hdbscan_min_samples
            ],
            outputs=[clustered_plot_output]
        )

if __name__ == "__main__":
    demo.launch(debug=True)