# Let us set up Python to hide unnecessary warnings.
import warnings
warnings.filterwarnings('ignore')

# Data setup: Let us load and preview the Iris dataset.
from sklearn.datasets import load_iris
import pandas as pd

iris = load_iris()
X = pd.DataFrame(iris.data, columns=iris.feature_names)
y = pd.Series(iris.target, name='species')
 
print('Shape:', X.shape)
print('First 5 rows:')
print(X.head())

# Let us check how many samples are in each species group.
print('Species counts:')
print(y.value_counts())

# Let us view basic statistics for the features.
print(X.describe())

# Standardize the data to mean 0 and variance 1.
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)

print('Mean of each column after scaling:')
print(X_scaled.mean(axis=0))

# Fit a PCA model and look at explained variance.
from sklearn.decomposition import PCA
pca = PCA()
pca.fit(X_scaled)

print('Explained variance ratio:', pca.explained_variance_ratio_)

# Let us plot the cumulative variance explained.
import matplotlib.pyplot as plt
plt.plot(range(1, 1 + len(pca.explained_variance_ratio_)), pca.explained_variance_ratio_.cumsum(), marker='o')
plt.xlabel('Number of components')
plt.ylabel('Cumulative explained variance')
plt.title('Variance explained by PCA components')
plt.show()

# Transform data using two principal components for visualization.
pca2 = PCA(n_components=2)
X_pca2 = pca2.fit_transform(X_scaled)

print('Shape after PCA:', X_pca2.shape)
print('First few PCA-transformed rows:')
print(X_pca2[:5])

# Visualize the PCA result with a scatter plot.
plt.figure(figsize=(6,4))
for lab, col in zip(range(3), ['red', 'green', 'blue']):
    plt.scatter(X_pca2[y==lab, 0], X_pca2[y==lab, 1], label=iris.target_names[lab], color=col)
plt.xlabel('First principal component')
plt.ylabel('Second principal component')
plt.legend()
plt.title('Iris data after PCA')
plt.show()

# Let us check the actual numbers behind the principal components.
print('PCA components (directions):')
print(pca2.components_)

# Reconstruct original data from two components (inverse transform).
X_reconstructed = pca2.inverse_transform(X_pca2)
print('First few reconstructed rows:')
print(X_reconstructed[:5])

# Sometimes, PCA is useful before running other models.
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression

# Split the PCA data into train and test sets.
X_train, X_test, y_train, y_test = train_test_split(X_pca2, y, test_size=0.3, random_state=42)

# Fit a simple classifier on top of PCA components.
model = LogisticRegression()
model.fit(X_train, y_train)
score = model.score(X_test, y_test)
print('Test set accuracy:', score)

# Challenge: What if we use only 1 principal component?
pca1 = PCA(n_components=1)
X_pca1 = pca1.fit_transform(X_scaled)
print('First few values with one component:')
print(X_pca1[:5])

# Handling non-numeric data: PCA needs numbers only.
# Let us pretend one column is text to see what happens.
X_test_bad = X.copy()
X_test_bad['fake_text'] = ['flower']*len(X_test_bad)
try:
    _ = scaler.fit_transform(X_test_bad)
except Exception as e:
    print('Error:', e)