Model Pipeline - Content-based filtering
Content-based filtering recommends items by learning what features a user likes from their past choices. It compares item features to suggest similar items.
0
0
Content-based filtering in ML Python - Model Pipeline Trace
Data Flow - 5 Stages
1Raw user-item interactions
1000 rows x 3 columns→Collect user IDs, item IDs, and user ratings→1000 rows x 3 columns
UserID=5, ItemID=23, Rating=4
↓
2Item feature extraction
500 items x 10 features→Extract features like genre, keywords, or attributes for each item→500 items x 10 features
ItemID=23, Features=[Action=1, Comedy=0, Drama=0, Length=120, ...]
↓
3User profile creation
User ratings + item features→Calculate weighted average of features from items user rated highly→1000 users x 10 features
UserID=5, Profile=[Action=0.8, Comedy=0.1, Drama=0.1, Length=110, ...]
↓
4Similarity calculation
User profile (1 x 10) and all item features (500 x 10)→Compute similarity scores between user profile and each item→500 similarity scores
UserID=5 similarity to ItemID=23 = 0.85
↓
5Recommendation generation
500 similarity scores→Sort items by similarity and select top N for recommendation→Top 10 recommended items
Recommended items for UserID=5: [23, 45, 12, 78, ...]
Training Trace - Epoch by Epoch
Loss
0.5 |****
0.4 |***
0.3 |**
0.2 |*
0.1 |
1 2 3 4 5 Epochs| Epoch | Loss ↓ | Accuracy ↑ | Observation |
|---|---|---|---|
| 1 | 0.45 | 0.6 | Initial model starts with moderate loss and accuracy. |
| 2 | 0.35 | 0.7 | Loss decreases and accuracy improves as user profiles better match item features. |
| 3 | 0.28 | 0.78 | Model continues to learn user preferences effectively. |
| 4 | 0.22 | 0.83 | Loss decreases steadily, accuracy rises, showing good convergence. |
| 5 | 0.18 | 0.87 | Training converges with low loss and high accuracy. |
Prediction Trace - 3 Layers
Layer 1: Input user profile
Layer 2: Calculate cosine similarity
Layer 3: Sort and select top items
Model Quiz - 3 Questions
Test your understanding
What does the user profile represent in content-based filtering?
Key Insight
Content-based filtering learns user preferences by analyzing features of items they liked. It then recommends new items with similar features, improving recommendations as it learns more about the user.