Update README.md

Author: sanskarppatil

Models/cosineandSVM/README.md

@@ -1 +1,59 @@
+ Movie Recommender System (Optimized SVM + PCA):
+
+This project implements a content-based movie recommendation system using:
+-  PCA (Principal Component Analysis) via Singular Value Decomposition (SVD) for dimensionality reduction
+- (Support Vector Machine) using a custom fast vectorized implementation
+
+All done from scratch using NumPy and Pandas — no scikit-learn  required.
+
+---
+
+ Dataset :
+
+The input dataset is a file named `cleaned_data.csv`, which includes:
+- title: Title of the movie
+- imdb_id: Movie ID (not used)
+- Other numerical columns representing metadata (genres, language, keywords, etc.)
+
+
+How It Works :
+
+Functions Overview :
+
+| Function Name             | Role                                                                 |
+|---------------------------|----------------------------------------------------------------------|
+ pca_svd(X, n_components)   Reduces the number of dimensions in `X` using SVD to `n_components
+ recommend_svm_optimized(...)   Trains a linear SVM from scratch to identify similar movies 
+ main block               Accepts user input, runs PCA, trains SVM, and outputs top results   
+
+---
+
+Execution Flow :
+
+1. Load Data:
+   - Reads the `cleaned_data.csv`
+   - Extracts features and movie titles
+
+2. **Standardize Data:
+   - Normalizes each feature to have zero mean and unit variance
+
+3. Apply PCA:
+   - Uses SVD to reduce features to the top `100` principal components
+
+4. Train Linear SVM (One-vs-All):
+   - Assigns label `1` to input movie, `0` to all others
+   - Trains linear SVM using hinge loss and gradient descent
+
+5. Score & Recommend:
+   - Calculates similarity scores for all movies
+   - Recommends top `10` similar titles, excluding the input itself
+
+---
+
+
+Requirements :
+
+Only basic libraries:
+```bash
+pip install numpy pandas