Ohad777/spotify-hit-prediction-analysis

---
license: mit
---

<video controls style="max-width: 100%;">
<source src="https://huggingface.co/datasets/Ohad777/spotify-hit-prediction-analysis/resolve/main/presentation.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>

# 🎵 Spotify Hit Prediction - Exploratory Data Analysis (EDA)

## Project Overview
This project analyzes audio features from Spotify to predict track popularity. Using a sample of **2,000 tracks**, I explored how technical attributes like energy and danceability relate to a song's success.

---

## 🔍 Research Questions & Insights
I addressed several key questions during the EDA:
1. **Is the data balanced?** I analyzed the ratio of popular vs. non-popular tracks to ensure fair modeling.
2. **Energy vs. Loudness:** Confirmed a strong positive correlation (**0.79**), showing energetic tracks are consistently louder.
3. **Does "Happiness" matter?** Using a **Violin Plot**, I found that both sad and happy songs (Valence) can become hits.
4. **Danceability:** Popular tracks tend to have a slightly higher and more consistent danceability range.
5. **Tempo:** Found no significant linear relationship between BPM and popularity.

---

## 🛠️ Data Decisions
* **Sampling:** Worked with 2,000 rows for efficiency.
* **Target:** Created a binary variable `is_popular` (1 for Popularity > 50, 0 otherwise).
* **Cleaning:** Confirmed zero missing values and decided to keep outliers as genuine musical variations.

---

## 📁 Files
* `spotify_sample_2000.csv`: Processed data subset.
* `Ohad_Danon_Assignment_1_EDA_&_Dataset.ipynb`: Full analysis code and visualizations.