0.29.18

pca

Author: bartzbeielstein

pyproject.toml

@@ -7,7 +7,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "spotpython"
-version = "0.29.17"
+version = "0.29.18"
 authors = [
   { name="T. Bartz-Beielstein", email="tbb@bartzundbartz.de" }
 ]

src/spotpython/utils/pca.py

@@ -0,0 +1,98 @@
+import numpy as np
+import pandas as pd
+from sklearn.decomposition import PCA
+from sklearn.preprocessing import StandardScaler
+import matplotlib.pyplot as plt
+
+
+def pca_analysis(
+    df,
+    df_name="",
+    k=10,
+    scaler=StandardScaler(),
+    max_scree=None,
+    figsize=(12, 6),
+) -> tuple:
+    """
+    Perform PCA analysis on a DataFrame with specified scaling.
+
+    Args:
+        df (pd.DataFrame):
+            The input data frame to perform PCA on.
+        df_name (str):
+            The name of the data frame.
+        k (int):
+            The number of top features to select based on their influence on PC1.
+        scaler (obj):
+            An instance of a Scaler from sklearn (e.g., StandardScaler()).
+        max_scree (int):
+            The maximum number of principal components to plot in the scree plot. Default is None, which means all components will be plotted.
+        figsize (tuple):
+            The size of the figure for the plots (width, height).
+
+    Returns:
+        tuple: Two pd.Index objects containing the names of the top k features most influential on PC1 and PC2, respectively.
+
+    Examples:
+        >>> import pandas as pd
+        >>> from spotpython.utils import pca_analysis
+        >>> df = pd.DataFrame({
+        ...     "A": [1, 2, 3],
+        ...     "B": [1, 2, 3],
+        ...     "C": [4, 5, 6]
+        ... })
+        >>> pca_analysis(df)
+    """
+    # Scale the data
+    scaled_data = scaler.fit_transform(df)
+    feature_names = df.columns
+    sample_names = df.index
+
+    # Perform PCA
+    pca = PCA()
+    pca.fit(scaled_data)
+    pca_data = pca.transform(scaled_data)
+
+    # Scree plot
+    per_var = np.round(pca.explained_variance_ratio_ * 100, decimals=1)
+    full_labels = ["PC" + str(x) for x in range(1, len(per_var) + 1)]
+
+    # Limit the number of PCs in the scree plot
+    if max_scree is not None:
+        per_var = per_var[:max_scree]
+        scree_labels = full_labels[:max_scree]
+    else:
+        scree_labels = full_labels
+
+    plt.figure(figsize=figsize)  # Set the figure size for the scree plot
+    plt.bar(x=range(1, len(per_var) + 1), height=per_var, tick_label=scree_labels)
+    plt.ylabel("Percentage of Explained Variance")
+    plt.xlabel("Principal Component")
+    plt.title(f"Scree Plot. {df_name}")
+    plt.show()
+
+    # PCA plot
+    plt.figure(figsize=figsize)  # Set the figure size for the PCA plot
+    pca_df = pd.DataFrame(pca_data, index=sample_names, columns=full_labels)
+
+    plt.scatter(pca_df.PC1, pca_df.PC2)
+    plt.title(f"PCA Graph. {df_name}")
+    plt.xlabel("PC1 - {0}%".format(per_var[0]))
+    plt.ylabel("PC2 - {0}%".format(per_var[1]))
+
+    for sample in pca_df.index:
+        plt.annotate(sample, (pca_df.PC1.loc[sample], pca_df.PC2.loc[sample]))
+
+    plt.show()
+
+    # Determine top k features influencing PC1 and PC2
+    loading_scores_pc1 = pd.Series(pca.components_[0], index=feature_names)
+    loading_scores_pc2 = pd.Series(pca.components_[1], index=feature_names)
+
+    sorted_loading_scores_pc1 = loading_scores_pc1.abs().sort_values(ascending=False)
+    sorted_loading_scores_pc2 = loading_scores_pc2.abs().sort_values(ascending=False)
+
+    top_k_features_pc1 = sorted_loading_scores_pc1.head(k).index
+    top_k_features_pc2 = sorted_loading_scores_pc2.head(k).index
+
+    return top_k_features_pc1, top_k_features_pc2