🎬 Live Machine Learning Demo

Find your next obsession,
powered by AI.

Pick any title you love — this recommender compares it against 8,787 Netflix titles using TF-IDF + cosine similarity and instantly finds your closest matches. Built end-to-end with scikit-learn by Muhammad Abdul Rafay, BS Artificial Intelligence, Air University Islamabad.

👋 New here? In plain words: tell this page one movie or show you enjoyed, and its AI finds others like it from the Netflix catalogue — and shows you why each one was picked. No sign-up, nothing to install. Just search below.

8,787Titles indexed
TF-IDFVector engine
Top 8Matches per pick
100%Client-side · instant
Try:

👆 How to start: type a title, pick it from the list, press Recommend — or simply tap one of the example chips above.

⏳ Loading the recommendation data…

⚙️ How it works

Pure content-based filtering — no accounts, no watch history, no user data.

🧬

1 · Title profiles

Every title becomes a text profile from its genres, country, director, type and rating.

🧮

2 · TF-IDF vectors

Profiles are vectorised with TF-IDF, so distinctive words count more than common ones.

🎯

3 · Cosine ranking

Cosine similarity ranks the whole catalogue by closeness to your pick — the top 8 win.

🚀 How to use — 3 easy steps

You do not need to know anything about AI or Netflix. It takes about ten seconds.

1️⃣

Pick a title you like

Type the name of any movie or TV show you enjoyed in the search box at the top and select it from the list that appears.

2️⃣

Press “Recommend”

The AI instantly compares your pick against all 8,787 titles in the catalogue and finds the 8 closest matches.

3️⃣

See why each was matched

Every suggestion shows a match score and a “Why this match” note, so you can see exactly what it shares with the title you picked.

🧠 How the AI works — in simple words

The same journey your pick takes, step by step:

🎬
Your pickOne title you like
→
🏷️
KeywordsGenres, cast, director & description become its keyword profile
→
🔢
NumbersTF-IDF turns keywords into numbers, rare words counting more
→
📐
CompareCosine similarity measures how close every title is to yours
→
🏆
Top 8 matchesRanked by match score, best first
What is a “match score”? It is a percentage that says how similar two titles are, based on the keywords they share. 100% would mean two titles have almost identical profiles; a lower score means they share only some ingredients. Try it: Midnight Mass and The Haunting of Hill House are both TV-MA horror shows with the genres “TV Dramas, TV Horror, TV Mysteries” — so the model scores them as a strong match (72.7%). The score is calculated live by this model for this catalogue only. It is not an official Netflix number, and it is not a quality rating — a 60% match can still be a brilliant show; it just shares fewer keywords with your pick.
70–100% · very similar flavour 50–69% · clearly related 30–49% · some shared ingredients Below 30% · distant cousin

🗂️ About the data

Real titles, real catalogue information — nothing made up.

8,787 titlesMovies and TV shows from the public Netflix catalogue dataset — every single one can be picked and matched.
What each title carriesType (movie / TV show), release year, age rating, duration, genres, director and country.
How matches were builtEach title's genres, cast, director and description were turned into a keyword profile with scikit-learn (TF-IDF), then compared with cosine similarity.
Runs in your browserThe recommendations were computed by the model and this page simply looks them up instantly — no server, no waiting.

❓ Questions, answered simply

Everything a first-time visitor usually asks.

Is this the real Netflix?

No. This is a student machine-learning project built by Muhammad Abdul Rafay to demonstrate how recommendation systems work. It uses public Netflix catalogue data, but it is not affiliated with Netflix, and you cannot watch anything here.

Are the match scores official Netflix scores?

No. The percentages are computed by the AI model from keyword similarity inside this catalogue. Netflix never published them — they exist only to show you how close two titles are to each other.

Do I need an account, sign-up, or my own Netflix history?

No. There is no login and nothing is collected. The system knows nothing about you — it only compares the content of titles (genres, cast, director, description), which is why it is called content-based filtering.

Why did it suggest a title I have never heard of?

Because that title shares keywords with your pick — read the “Why this match” note on its card. Unknown titles are often the best discoveries: they match your taste even though they were never famous.

What if my title is not found?

Start typing and choose from the suggestion list — the title must exist in the 8,787-title catalogue. Very new releases may not be in the dataset yet.