Python from scratch - PAL

import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.patches import Ellipse
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA
from scipy.spatial import ConvexHull

# 1) File paths and column definitions
excel_path = r"C:\Users\Office\Desktop\PCA.xlsx"
sheet_name = 0

id_col = "Subject ID"
group_col = "Group"

feature_cols = ["Variable 1", "Variable 2", "Variable 3", "Variable 4",
"Variable 5", "Variable 6", "Variable 7", "Variable 8"]

group_colors = {"T1": "blue", "T2": "red"}

out_dir = "pca_outputs"
os.makedirs(out_dir, exist_ok=True)

# 2) Read data
df = pd.read_excel(excel_path, sheet_name=sheet_name)

if id_col not in df.columns:
raise ValueError(f"ID column '{id_col}' not found.")
if group_col not in df.columns:
raise ValueError(f"Group column '{group_col}' not found.")

mask = df[feature_cols].notna().all(axis=1) & df[group_col].notna()
df_clean = df.loc[mask].copy()

# 3) Standardize data
X_raw = df_clean[feature_cols].to_numpy(float)
scaler = StandardScaler()
X_z = scaler.fit_transform(X_raw)

# 4) PCA computation
pca = PCA(n_components=2)
scores = pca.fit_transform(X_z)
loadings = pca.components_.T
explained = pca.explained_variance_ratio_

df_clean["PC1"] = scores[:, 0]
df_clean["PC2"] = scores[:, 1]

# 5) Draw convex hull polygons
def draw_convex_hull(ax, points, color, alpha=0.15):
if len(points) < 3:
return
hull = ConvexHull(points)
hull_points = points[hull.vertices]
ax.fill(hull_points[:, 0], hull_points[:, 1],
color=color, alpha=alpha)
ax.plot(hull_points[:, 0], hull_points[:, 1],
color=color, linewidth=2)

# 6) PCA score plot with convex hulls
plt.figure(figsize=(7, 6))
ax = plt.gca()

ax.axhline(0, linestyle="--", color="lightgray")
ax.axvline(0, linestyle="--", color="lightgray")

for grp in df_clean[group_col].unique():
subset = df_clean[df_clean[group_col] == grp]
color = group_colors.get(grp, "black")

ax.scatter(subset["PC1"], subset["PC2"],
color=color, label=grp, alpha=0.9)

pts = subset[["PC1", "PC2"]].to_numpy()
draw_convex_hull(ax, pts, color)

# Add subject labels
for _, row in df_clean.iterrows():
ax.text(row["PC1"], row["PC2"] + 0.04,
str(row[id_col]), fontsize=8)

ax.set_xlabel(f"PC1 ({explained[0]*100:.1f}%)")
ax.set_ylabel(f"PC2 ({explained[1]*100:.1f}%)")
ax.set_title("PCA Score Plot (Convex Hull Groups)")
ax.legend()

plt.tight_layout()
plt.savefig(os.path.join(out_dir, "PCA_score_hull.jpg"), dpi=300)
plt.savefig(os.path.join(out_dir, "PCA_score_hull.svg"))
plt.show()

# 7) PCA Loading plot (PC1 vs PC2)
plt.figure(figsize=(7, 6))
pc1_load = loadings[:, 0]
pc2_load = loadings[:, 1]

plt.axhline(0, linestyle="--", color="lightgray")
plt.axvline(0, linestyle="--", color="lightgray")

plt.scatter(pc1_load, pc2_load, s=80)

for i, feat in enumerate(feature_cols):
plt.text(pc1_load[i], pc2_load[i] + 0.02, feat,
fontsize=9, ha="center")

plt.xlabel("PC1 Loading")
plt.ylabel("PC2 Loading")
plt.title("PCA Loading Plot (PC1 vs PC2)")

plt.tight_layout()
plt.savefig(os.path.join(out_dir, "PCA_loadings.jpg"), dpi=300)
plt.savefig(os.path.join(out_dir, "PCA_loadings.svg"))
plt.show()

# 8) Explained variance plot
plt.figure(figsize=(7, 4))
x = np.arange(1, 3)

plt.bar(x, explained * 100)

for i, v in enumerate(explained * 100):
plt.text(i + 1, v + 0.5, f"{v:.1f}%",
ha="center", fontsize=10)

plt.ylim(0, (explained * 100).max() + 10)
plt.xticks([1, 2], ["PC1", "PC2"])

plt.xlabel("Principal Component")
plt.ylabel("Explained Variance (%)")
plt.title("Explained Variance of PCA Components")

plt.tight_layout()
plt.savefig(os.path.join(out_dir, "PCA_variance.jpg"), dpi=300)
plt.savefig(os.path.join(out_dir, "PCA_variance.jpg"))
plt.show()

print("\n✔️ All PCA plots generated and saved in:", os.path.abspath(out_dir))

61 viewsedited 05:59