Music Interaction Dataset

This dataset captures real-world music listening behaviors and contextual metadata from January 2021 to January 2025, sampled at 30-minute intervals. It reflects the streaming activity of 50 anonymized users across a variety of devices and locations. Designed to support the development of intelligent music recommendation systems, this dataset offers rich contextual information suitable for personalization, behavioral analysis, sentiment modeling, and federated learning applications. Each record represents a user-song interaction and is timestamped with detailed user metadata, song descriptors, audio features, emotional indicators, and session-level dynamics. The dataset is especially useful for tasks such as preference prediction, session-based recommendation, genre classification, and user intent modeling. Feature Overview User Profile: user_id, age, gender, location, subscription_type, device_type Behavioral Patterns: listening_time_mins, sessions_per_day, time_of_day, day_of_week, recent_skip_rate, first_time_listening User Preferences: preferred_genre, preferred_artist, repeat_count, added_to_playlist, finished_song Content Metadata: song_id, title, artist, album, genre, release_year, language, duration_sec, explicit, popularity Audio Features: tempo, key, mode, time_signature, energy, danceability, acousticness, instrumentalness, liveness, valence, loudness, speechiness Sentiment and Emotion: lyrics_sentiment, emotion_tag Interaction Logs: play_count, skip_count, time_spent_on_song Session-Level Context: context_type, song_position_in_session, session_duration_mins Target Label: liked (1 if user liked the song, 0 if skipped)