Added ML model for recommendations -NLP Based

[parinaB]

What does this PR do?

---

Implements a lightweight ML-based Resource Recommendation System for HerStack. Given a user query (e.g. "I want to learn CNN"), the system vectorizes it using TF-IDF and ranks all 53 learning resources by cosine similarity — returning the most relevant ones with optional difficulty and category filters.

Key changes:

• Added clean_text() preprocessing pipeline (lowercase + regex noise removal)
• TF-IDF vectorizer tuned with bigrams, max_df=0.85, max_features=1000, sublinear_tf
• Added dataset of 53 resources across 7 categories.
• KNN classifier on TF-IDF vectors for accuracy evaluation — 82% 3-fold CV accuracy
• UI integration (React "Recommended for You" section) is out of scope for this PR and will be tracked separately.
  Note: Separating the React UI integration helps keep the PR focused, easier to review, and avoids mixing backend ML logic with frontend implementation changes.

---

Type of change

• 🐛 Bug fix
• ✨ New feature
• [] 📝 Content (mentor / resource / storyboard)
• 📖 Documentation
• 🎨 Style / UI improvement
• ♻️ Refactor

---

Checklist

• I've tested my changes locally (npm run dev)
• My code follows the project style guide
• I've added/updated comments where needed
• No console errors introduced

---

Screenshots (if UI change)

---

Adding Accuracy Charts of the model

Related issue

Closes #13

NOTE: I plan to open a separate PR for integrating frontend features and building the React-based recommendation UI for this model. I’d really love to continue contributing to and being part of this project further.

[vercel]

@parinaB is attempting to deploy a commit to the HerStack's projects Team on Vercel.

A member of the Team first needs to authorize it.

[parinaB]

@escher-m could you please review and update accordingly..

[parinaB]

@escher-m would u kindly review once free...