Fix: Implemented Label Encoding and Unified ML Models using Pipelines

[itsvishwasj]

Fix: Implemented Label Encoding and Unified ML Models using Pipelines

🛠️ Pull Request Template

🏷️ PR Type

• 🐞 Bug Fix
• 🛠️ Improvement / Refactor

---

🔗 Related Issue

• Closes Training pipeline breaks without Label Encoding and integrated TF-IDF #68

---

📝 Rationale / Motivation

This PR fully resolves Issue #68, which reported errors and instability in the classical machine learning model training workflow.

1. String Label Error: Fixed runtime errors caused by passing non-numerical string labels (e.g., "half-true") directly to scikit-learn classifiers. This is fixed by applying LabelEncoder.
2. Preprocessing Duplication: Fixed inconsistent and redundant $\text{TfidfVectorizer}$ steps by bundling the vectorizer and the classifier into a unified Pipeline object.

The changes significantly improve stability, data consistency, and code clarity.

---

✨ Description of Changes

• Core Fixes Applied:

  • Label Encoding: Implemented sklearn.preprocessing.LabelEncoder to convert the 6 unique string labels to integers across all affected files.
  • Pipeline Integration: Refactored training logic to use sklearn.pipeline.Pipeline, combining $\text{TfidfVectorizer}$ with the classifier (NB, LR, RF) in a single object.

• Files Modified:

  • scripts/fake_news_logreg_rf.py: Implemented Pipelines for $\text{LogisticRegression}$, $\text{RandomForestClassifier}$, and $\text{MultinomialNB}$.
  • module/liar-data-analysis.py: Implemented Label Encoding and Pipelines for analysis examples.
  • module/fake-news-detection-using-nb.ipynb: Applied Label Encoding and the $\text{MultinomialNB}$ Pipeline logic.

---

🧪 Testing Instructions

1. Pull this branch and ensure your dependencies are installed.
2. Run the main comparison script:

   python scripts/fake_news_logreg_rf.py

3. Expected Results:
   • The script must execute without crashing (no more errors about string labels).
   • Accuracy scores and classification reports for all three models (NB, LR, RF) should be printed to the console.
   • New or updated result files (.md, confusion matrix PNGs) should be generated in the results/ directory.

---

👀 Impact Assessment

• Stability: High impact, resolving core ML runtime errors and inconsistencies reported in Training pipeline breaks without Label Encoding and integrated TF-IDF #68.
• Maintainability: Improves code cleanliness by eliminating redundant vectorization steps.

---

⚡ Checklist

• Code follows project’s coding style and guidelines
• Changes are tested locally
• Automated tests added/updated (N/A)
• Documentation updated (N/A)
• User-facing changes are documented (N/A)
• Related issue linked
• No new warnings/errors introduced

---

⚠️ Breaking Changes

None. This is an internal fix that preserves the input/output of the scripts.

---

🎯 Priority / Impact Level

• Priority: High (Critical bug fix)
• Impact: Medium (Internal code structure)

[vercel]

@itsvishwasj is attempting to deploy a commit to the deepika's projects Team on Vercel.

A member of the Team first needs to authorize it.