Calculator Form
Use numeric vectors, optional weights, and optional tags for a fuller similarity analysis.
Example Data Table
| Feature | Item A | Item B | Weight |
|---|---|---|---|
| Rating | 4.5 | 4.3 | 1.5 |
| Relevance | 8.9 | 8.4 | 1.8 |
| Freshness | 7.2 | 7.0 | 1.1 |
| Quality | 9.1 | 8.8 | 1.7 |
| Popularity | 7.8 | 7.4 | 1.0 |
Example tags for Item A: premium, wireless, smart, portable. Example tags for Item B: wireless, smart, portable, budget.
Formula Used
1) Cosine Similarity
cosine = (A · B) / (||A|| × ||B||)
2) Weighted Cosine Similarity
weighted cosine = Σ(wᵢaᵢbᵢ) / (√Σ(wᵢaᵢ²) × √Σ(wᵢbᵢ²))
3) Pearson Similarity
r = cov(A,B) / (σA × σB)
pearson similarity = (r + 1) / 2
4) Euclidean-Based Similarity
distance = √(Σ(wᵢ(aᵢ - bᵢ)²) / Σwᵢ), similarity = 1 / (1 + distance)
5) Jaccard Similarity
jaccard = |A ∩ B| / |A ∪ B|
6) Hybrid Similarity Score
hybrid = Σ(metric score × metric weight) / Σ(metric weights)
Use cosine for direction matching, Pearson for pattern alignment, Euclidean for distance penalties, and Jaccard for shared tags or set overlap.
How to Use This Calculator
- Enter feature labels to describe each vector position.
- Paste numeric vectors for Item A and Item B.
- Add optional feature weights to emphasize important dimensions.
- Provide optional tags for Jaccard similarity analysis.
- Select normalization and choose the main metric to display.
- Adjust hybrid weights if you want a blended score.
- Press Calculate Similarity to see the result above the form.
- Use the CSV or PDF buttons to export the summary and table.
Frequently Asked Questions
1) What does item similarity measure?
It measures how closely two items align across shared features, patterns, distances, or tags. This helps with recommendation engines, search ranking, clustering, duplicate detection, and nearest-neighbor style comparisons.
2) When should I use cosine similarity?
Use cosine similarity when direction matters more than magnitude. It works well for embeddings, text vectors, profile matching, and sparse recommendation features where relative shape is more useful than raw scale.
3) Why add feature weights?
Weights let you emphasize important dimensions such as relevance, quality, revenue, or trust. This is useful when some attributes should influence the match more strongly than others.
4) What does a negative Pearson value mean?
A negative Pearson correlation means the two items move in opposite patterns. One tends to rise where the other falls. The calculator converts that raw value into a 0-to-1 similarity scale for easier comparison.
5) Why would I use normalization?
Normalization helps when features use very different scales. Without it, a large numeric dimension may dominate the result. Min-max and z-score scaling make cross-feature comparisons more balanced.
6) What is Jaccard similarity used for?
Jaccard compares overlap between tag sets, categories, or activated features. It is useful when you want to know how many shared labels exist relative to the total unique labels across both items.
7) What does the hybrid score do?
The hybrid score blends multiple similarity signals into one weighted result. It is helpful when you want a more robust decision that combines vector direction, pattern shape, distance, and set overlap.
8) What does the Plotly graph show?
The graph compares the scaled feature values of both items and highlights weighted gaps. It helps you spot where the items align strongly and where the biggest differences appear.