| # 🎵 Music Feature Dataset Analysis | |
| Welcome to the Exploratory Data Analysis (EDA) of a music feature dataset. This project aims to uncover insights into how audio features of songs influence user preferences, using the `liked` column as the target variable. | |
| --- | |
| ## 📁 Dataset Overview: `train.csv` | |
| This dataset contains various musical/audio features for tracks, along with a `liked` score representing user preference. | |
| ### 🎼 Features Description | |
| | Feature | Description | | |
| |-------------------|------------------------------------------------------------------------| | |
| | `danceability` | Suitability of a track for dancing. | | |
| | `energy` | Intensity and activity level of the track. | | |
| | `key` | Musical key as an integer (0 to 11). | | |
| | `loudness` | Overall loudness in decibels (dB). | | |
| | `mode` | Tonality: major (1) or minor (0). | | |
| | `speechiness` | Presence of spoken words in the track. | | |
| | `acousticness` | Likelihood the track is acoustic. | | |
| | `instrumentalness`| Predicts whether a track contains no vocals. | | |
| | `liveness` | Presence of an audience in the recording. | | |
| | `valence` | Positiveness conveyed by the music. | | |
| | `tempo` | Tempo in beats per minute (BPM). | | |
| | `duration_ms` | Duration of the track in milliseconds. | | |
| | `time_signature` | Estimated time signature (usually 3, 4, or 5). | | |
| | `liked` | User preference score (continuous numerical value). | | |
| --- | |
| ## 🧼 Data Preprocessing | |
| - ✅ Handled missing values | |
| - ✅ Removed duplicates | |
| - ✅ Normalized numeric features using **Z-score normalization** | |
| - ✅ Detected outliers using box plots | |
| - ✅ Computed summary statistics (mean, std, etc.) | |
| --- | |
| ## 📊 Visual Analysis | |
| Visualizations included in the analysis: | |
| | Visualization | Description | | |
| |----------------------------------|----------------------------------------------| | |
| |  | Heatmap of missing values | | |
| |  | Count of unique and duplicate rows | | |
| |  | Normalized feature distributions | | |
| |  | Most and least liked tracks | | |
| |  | Histograms of key features | | |
| | | Focused view on `danceability` distribution | | |
| |  | Feature-wise mean and standard deviation | | |
| |  | Feature correlation with `liked` | | |
| |  | Outlier detection via box plots | | |
| --- | |
| ## 📈 Key Findings | |
| - 🔹 Features like `energy`, `valence`, and `danceability` have **positive correlation** with the `liked` score. | |
| - 🔹 Features like `key` and `mode` show **low or no correlation** with user preference. | |
| - 🔹 Most features are approximately **normally distributed**. | |
| - 🔹 Tracks that are **highly energetic, positive, and danceable** are generally more liked. | |
| --- | |
| ## 🛠️ Tools & Libraries Used | |
| - **Language**: Python 🐍 | |
| - **Libraries**: | |
| - `pandas`, `numpy` for data handling | |
| - `matplotlib`, `seaborn` for visualizations | |
| - `scikit-learn` for preprocessing | |
| --- | |
| ## 📌 Conclusion | |
| This analysis helps identify which audio features most impact user preferences. These insights can guide the development of: | |
| - 🎧 Music recommendation systems | |
| - 📊 User behavior models | |
| - 🤖 Feature engineering in ML projects | |
| --- | |
| ## 👤 Author | |
| **Sujal Thakkar** | |
| --- | |