sentiment_analysis.py

# -*- coding: utf-8 -*-
"""Sentiment Analysis.ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/1Tcy0Mn8Sv42LRVddOs0ufYTKx2ML-uev
"""

#CODTECH Internship - TASK 2
# Enhanced Sentiment Analysis with Comprehensive Visualizations and Metrics
# Dataset: Sample IMDb reviews (loaded via sklearn)
# ===============================================================

# Step 1: Import Libraries
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.datasets import fetch_20newsgroups
from sklearn.model_selection import train_test_split, cross_val_score, validation_curve
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import (classification_report, confusion_matrix, accuracy_score,
                           precision_score, recall_score, f1_score, roc_auc_score,
                           roc_curve, precision_recall_curve, matthews_corrcoef,
                           balanced_accuracy_score, log_loss)
from sklearn.naive_bayes import MultinomialNB
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier
from wordcloud import WordCloud
import warnings
warnings.filterwarnings('ignore')

# Set style for better plots
plt.style.use('seaborn-v0_8-darkgrid')
sns.set_palette("husl")

# Step 2: Load Dataset (Using 'rec.autos' vs 'sci.med' as proxy for sentiment)
categories = ['rec.autos', 'sci.med']
data = fetch_20newsgroups(subset='all', categories=categories, remove=('headers', 'footers', 'quotes'))

df = pd.DataFrame({'text': data.data, 'target': data.target})
df['target'] = df['target'].map({0: "Negative", 1: "Positive"})

print("✅ Sample data:")
print(df.head())
print(f"\n📊 Dataset shape: {df.shape}")
print(f"📊 Target distribution:\n{df['target'].value_counts()}")

# Step 3: Data Exploration and Visualization
fig, axes = plt.subplots(2, 2, figsize=(15, 12))

# Dataset distribution
df['target'].value_counts().plot(kind='bar', ax=axes[0,0], color=['#FF6B6B', '#4ECDC4'])
axes[0,0].set_title('Target Distribution', fontsize=14, fontweight='bold')
axes[0,0].set_xlabel('Sentiment')
axes[0,0].set_ylabel('Count')
axes[0,0].tick_params(axis='x', rotation=0)

# Text length distribution
df['text_length'] = df['text'].str.len()
df.boxplot(column='text_length', by='target', ax=axes[0,1])
axes[0,1].set_title('Text Length Distribution by Sentiment', fontsize=14, fontweight='bold')
axes[0,1].set_xlabel('Sentiment')
axes[0,1].set_ylabel('Text Length')

# Word count distribution
df['word_count'] = df['text'].str.split().str.len()
df.boxplot(column='word_count', by='target', ax=axes[1,0])
axes[1,0].set_title('Word Count Distribution by Sentiment', fontsize=14, fontweight='bold')
axes[1,0].set_xlabel('Sentiment')
axes[1,0].set_ylabel('Word Count')

# Histogram of text lengths
axes[1,1].hist(df[df['target']=='Negative']['text_length'], alpha=0.7, label='Negative', bins=30)
axes[1,1].hist(df[df['target']=='Positive']['text_length'], alpha=0.7, label='Positive', bins=30)
axes[1,1].set_title('Text Length Histogram', fontsize=14, fontweight='bold')
axes[1,1].set_xlabel('Text Length')
axes[1,1].set_ylabel('Frequency')
axes[1,1].legend()

plt.tight_layout()
plt.show()

# Step 4: Word Clouds
fig, axes = plt.subplots(1, 2, figsize=(16, 6))

# Negative sentiment word cloud
negative_text = ' '.join(df[df['target'] == 'Negative']['text'])
wordcloud_neg = WordCloud(width=800, height=400, background_color='white',
                         colormap='Reds').generate(negative_text)
axes[0].imshow(wordcloud_neg, interpolation='bilinear')
axes[0].set_title('Most Common Words - Negative Sentiment', fontsize=14, fontweight='bold')
axes[0].axis('off')

# Positive sentiment word cloud
positive_text = ' '.join(df[df['target'] == 'Positive']['text'])
wordcloud_pos = WordCloud(width=800, height=400, background_color='white',
                         colormap='Blues').generate(positive_text)
axes[1].imshow(wordcloud_pos, interpolation='bilinear')
axes[1].set_title('Most Common Words - Positive Sentiment', fontsize=14, fontweight='bold')
axes[1].axis('off')

plt.tight_layout()
plt.show()

# Step 5: Preprocessing
df.dropna(inplace=True)
X = df['text']
y = df['target']

# Step 6: TF-IDF Vectorization
tfidf = TfidfVectorizer(stop_words='english', max_df=0.7, min_df=2, max_features=5000)
X_tfidf = tfidf.fit_transform(X)

print(f"\n🔧 TF-IDF Matrix shape: {X_tfidf.shape}")

# Step 7: Train-Test Split
X_train, X_test, y_train, y_test = train_test_split(X_tfidf, y, test_size=0.2,
                                                   random_state=42, stratify=y)

# Step 8: Model Training and Comparison
models = {
    'Logistic Regression': LogisticRegression(random_state=42, max_iter=1000),
    'Naive Bayes': MultinomialNB(),
    'SVM': SVC(probability=True, random_state=42),
    'Random Forest': RandomForestClassifier(n_estimators=100, random_state=42)
}

model_results = {}

# Train and evaluate each model
for name, model in models.items():
    model.fit(X_train, y_train)
    y_pred = model.predict(X_test)
    y_pred_proba = model.predict_proba(X_test)[:, 1] if hasattr(model, 'predict_proba') else None

    # Calculate comprehensive metrics
    metrics = {
        'Accuracy': accuracy_score(y_test, y_pred),
        'Balanced Accuracy': balanced_accuracy_score(y_test, y_pred),
        'Precision': precision_score(y_test, y_pred, pos_label='Positive'),
        'Recall': recall_score(y_test, y_pred, pos_label='Positive'),
        'F1-Score': f1_score(y_test, y_pred, pos_label='Positive'),
        'Matthews Correlation': matthews_corrcoef(y_test, y_pred),
        'AUC-ROC': roc_auc_score(y_test, y_pred_proba) if y_pred_proba is not None else 'N/A',
        'Log Loss': log_loss(y_test, y_pred_proba) if y_pred_proba is not None else 'N/A'
    }

    model_results[name] = {
        'model': model,
        'metrics': metrics,
        'predictions': y_pred,
        'probabilities': y_pred_proba
    }

# Step 9: Results Visualization
# Model comparison
metrics_df = pd.DataFrame({name: result['metrics'] for name, result in model_results.items()}).T

# Filter numeric metrics for plotting
numeric_metrics = ['Accuracy', 'Balanced Accuracy', 'Precision', 'Recall', 'F1-Score', 'Matthews Correlation']
numeric_df = metrics_df[numeric_metrics].astype(float)

fig, axes = plt.subplots(2, 2, figsize=(16, 12))

# Bar plot of metrics
numeric_df.plot(kind='bar', ax=axes[0,0], width=0.8)
axes[0,0].set_title('Model Performance Comparison', fontsize=14, fontweight='bold')
axes[0,0].set_xlabel('Models')
axes[0,0].set_ylabel('Score')
axes[0,0].legend(bbox_to_anchor=(1.05, 1), loc='upper left')
axes[0,0].tick_params(axis='x', rotation=45)

# Heatmap of metrics
sns.heatmap(numeric_df, annot=True, cmap='YlOrRd', ax=axes[0,1], fmt='.3f')
axes[0,1].set_title('Model Performance Heatmap', fontsize=14, fontweight='bold')

# Best model detailed analysis (Logistic Regression)
best_model = model_results['Logistic Regression']['model']
best_predictions = model_results['Logistic Regression']['predictions']
best_probabilities = model_results['Logistic Regression']['probabilities']

# Confusion Matrix
cm = confusion_matrix(y_test, best_predictions)
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', ax=axes[1,0],
           xticklabels=['Negative', 'Positive'], yticklabels=['Negative', 'Positive'])
axes[1,0].set_title('Confusion Matrix - Logistic Regression', fontsize=14, fontweight='bold')
axes[1,0].set_xlabel('Predicted')
axes[1,0].set_ylabel('Actual')

# ROC Curve
fpr, tpr, _ = roc_curve(y_test, best_probabilities, pos_label='Positive')
auc_score = roc_auc_score(y_test, best_probabilities)
axes[1,1].plot(fpr, tpr, linewidth=2, label=f'ROC Curve (AUC = {auc_score:.3f})')
axes[1,1].plot([0, 1], [0, 1], 'k--', linewidth=1)
axes[1,1].set_xlabel('False Positive Rate')
axes[1,1].set_ylabel('True Positive Rate')
axes[1,1].set_title('ROC Curve - Logistic Regression', fontsize=14, fontweight='bold')
axes[1,1].legend()
axes[1,1].grid(True)

plt.tight_layout()
plt.show()

# Step 10: Additional Visualizations
fig, axes = plt.subplots(2, 2, figsize=(16, 12))

# Precision-Recall Curve
precision, recall, _ = precision_recall_curve(y_test, best_probabilities, pos_label='Positive')
axes[0,0].plot(recall, precision, linewidth=2, color='purple')
axes[0,0].set_xlabel('Recall')
axes[0,0].set_ylabel('Precision')
axes[0,0].set_title('Precision-Recall Curve', fontsize=14, fontweight='bold')
axes[0,0].grid(True)

# Prediction Probability Distribution
axes[0,1].hist(best_probabilities[y_test == 'Negative'], alpha=0.7, label='Negative', bins=30)
axes[0,1].hist(best_probabilities[y_test == 'Positive'], alpha=0.7, label='Positive', bins=30)
axes[0,1].set_xlabel('Prediction Probability')
axes[0,1].set_ylabel('Frequency')
axes[0,1].set_title('Distribution of Prediction Probabilities', fontsize=14, fontweight='bold')
axes[0,1].legend()

# Feature Importance (Top TF-IDF features)
feature_names = tfidf.get_feature_names_out()
coef = best_model.coef_[0]
top_positive_indices = np.argsort(coef)[-10:]
top_negative_indices = np.argsort(coef)[:10]

top_features = np.concatenate([coef[top_negative_indices], coef[top_positive_indices]])
feature_labels = [feature_names[i] for i in np.concatenate([top_negative_indices, top_positive_indices])]
colors = ['red'] * 10 + ['blue'] * 10

axes[1,0].barh(range(len(top_features)), top_features, color=colors, alpha=0.7)
axes[1,0].set_yticks(range(len(feature_labels)))
axes[1,0].set_yticklabels(feature_labels)
axes[1,0].set_xlabel('Coefficient Value')
axes[1,0].set_title('Top 20 Most Important Features', fontsize=14, fontweight='bold')

# Cross-validation scores
cv_scores = cross_val_score(best_model, X_tfidf, y, cv=5, scoring='accuracy')
axes[1,1].boxplot(cv_scores)
axes[1,1].set_ylabel('Accuracy Score')
axes[1,1].set_title(f'5-Fold Cross-Validation\nMean: {cv_scores.mean():.3f} (±{cv_scores.std():.3f})',
                   fontsize=14, fontweight='bold')
axes[1,1].grid(True)

plt.tight_layout()
plt.show()

# Step 11: Comprehensive Results Summary
print("\n" + "="*80)
print("🎯 COMPREHENSIVE RESULTS SUMMARY")
print("="*80)

for name, result in model_results.items():
    print(f"\n📊 {name}:")
    print("-" * 50)
    for metric, value in result['metrics'].items():
        if isinstance(value, float):
            print(f"{metric:20}: {value:.4f}")
        else:
            print(f"{metric:20}: {value}")

print(f"\n🔄 Cross-Validation Results (Logistic Regression):")
print(f"CV Scores: {cv_scores}")
print(f"Mean CV Score: {cv_scores.mean():.4f} (±{cv_scores.std():.4f})")

print(f"\n📈 Best Model: Logistic Regression")
print(f"Test Accuracy: {model_results['Logistic Regression']['metrics']['Accuracy']:.4f}")
print(f"AUC-ROC Score: {model_results['Logistic Regression']['metrics']['AUC-ROC']:.4f}")

# Classification report for best model
print(f"\n📋 Detailed Classification Report (Logistic Regression):")
print(classification_report(y_test, best_predictions))

print("\n✅ Analysis Complete!")
print("="*80)