README.md

🎵 Music Feature Dataset Analysis

This repository contains a comprehensive exploratory data analysis (EDA) on a music features dataset. The primary objective is to understand the patterns in audio features and analyze how they relate to user preferences, providing insights for music recommendation systems and user profiling.

📥 Dataset Overview

The dataset (data.csv) contains audio features extracted from music tracks along with user preference scores. This rich collection of acoustic and musical attributes enables deep analysis of what makes music appealing to listeners.

Total songs: 195

Format: CSV (data.csv)

Source: Spotify API

Target column: liked (1 = liked, 0 = disliked)

Data type: Tabular

Licensing: For academic and personal research use (derived from Spotify API)

🎼 Features Description

Feature	Description	Data Type	Range/Values
danceability	Measures how suitable a track is for dancing based on rhythm, tempo, and beat strength	Float	0.0 - 1.0
energy	Intensity and activity level representing loudness, dynamic range, and general entropy	Float	0.0 - 1.0
key	Musical key using standard Pitch Class notation	Integer	0 - 11
loudness	Overall loudness measured in decibels (dB)	Float	Typically -60 to 0
mode	Modality of the track (Major = 1, Minor = 0)	Integer	0, 1
speechiness	Presence of spoken words in a track	Float	0.0 - 1.0
acousticness	Confidence measure of whether the track is acoustic	Float	0.0 - 1.0
instrumentalness	Predicts whether a track contains no vocals	Float	0.0 - 1.0
liveness	Detects the presence of an audience in the recording	Float	0.0 - 1.0
valence	Musical positiveness conveyed by a track	Float	0.0 - 1.0
tempo	Overall estimated tempo in beats per minute (BPM)	Float	Usually 50-200+
duration_ms	Track duration in milliseconds	Integer	Positive integers
time_signature	Estimated overall time signature	Integer	3, 4, 5, 7
liked	Target Variable: User preference score	Float	Continuous values

📊 EDA Overview: Music Preference Dataset

1️⃣ Null Values Check

✔ The dataset is complete — no missing entries detected.

2️⃣ Target Class Breakdown

Liked Tracks (1): 100 entries

Disliked Tracks (0): 95 entries

Total_Liked_and_Disliked_Songs

🟢 The class distribution is fairly even — no need for balancing.

3️⃣ Feature Types

All input variables are numeric.

The target label liked is a binary flag (0 = dislike, 1 = like).

4️⃣ Key Statistical Insights

Higher average values for energy, danceability, and valence are seen in liked songs.

In contrast, acousticness and instrumentalness are more prominent in disliked tracks.

5️⃣ Correlation Patterns

🔗 Positive: energy strongly correlates with loudness.

🔻 Negative: acousticness shows inverse correlation with energy and valence.

6️⃣ Recommended Visual Explorations

📌 Try the following plots to gain deeper insights:

📦 Boxplots comparing liked vs energy, danceability

📊 Bar chart for distribution of likes/dislikes

🌡️ Heatmap of all feature correlations

🎯 Scatter plot: energy vs valence, with points colored by liked status

Required Libraries

pandas>=1.3.0
numpy>=1.21.0
matplotlib>=3.4.0
seaborn>=0.11.0
scikit-learn>=1.0.0

🧼 Data Preprocessing

Our comprehensive preprocessing pipeline includes:

1. Data Quality Assessment

• ✅ Missing value detection and handling
• ✅ Duplicate record identification and removal
• ✅ Data type validation and conversion
• ✅ Outlier detection using statistical methods

🤖 ML Use Cases

You can use this dataset to train:

1. Logistic Regression

2. Random Forest

3. K-Nearest Neighbors.

4. Support Vector Machine.

5. Artificial Neural Network.

6. Naive Bayes

7. Decision Tree.

📈 Analysis & Visualizations

Pairplot_features_liked

Model_Accuracy_Comparison

Loudness_Distribution_by_Liked_Status

correlation_heatmap

Acousticness_vs_Danceability

🔍 Key Findings

🎯 Primary Insights

1. Feature Distributions

   • Most audio features follow approximately normal distributions
   • valence and danceability show interesting bimodal patterns
   • tempo exhibits a wide range with multiple peaks

2. Correlation Patterns

   • Strong positive correlation between energy, valence, and user preference (liked)
   • Moderate correlation between danceability and liked scores
   • Weak correlation for categorical features like key and mode

3. User Preference Drivers

   • Higher danceability → Higher user preference
   • Higher valence (positivity) → Better ratings
   • Optimal energy levels correlate with user satisfaction
   • acousticness shows inverse relationship with preferences

📊 Results

Model Performance Insights

• Features most predictive of user preference: energy, valence, danceability
• Optimal feature ranges for high user satisfaction identified
• Recommendations for music recommendation system development

🛠 Technologies Used

Core Libraries

• Data Manipulation: pandas, numpy
• Visualization: matplotlib, seaborn, plotly
• Statistical Analysis: scipy, statsmodels
• Machine Learning: scikit-learn

Development Tools

• Environment: Jupyter Lab/Notebook
• Version Control: Git
• Package Management: pip/conda
• Documentation: Markdown