Real Data Experiments:
-Music Recomendation
+--music-exps-survey: web app where users evaluated playlists recommendations
+--music-generator: source code for generating playlist recommendations
+--results: results of the evaluations
-Task Recomendation
+--results
+--Long-Sessions: raw data in .db of the long sessions. Analysis is in long-sessions-summary.
+--Short-Sessions: raw data in .db of the short sessions. Analysis is in short-sessions-summary.
+--music-generator: source code for generating task recommendations
+--task-generator: web app for workers to complete generated tasks
Dataset Description:
These are the new datasets created from various sources. They are Python Pickled files that are in the format of dict
song_id : (first-dimension, second-dimension)
example:
data = {
...
992: (289.2273, 0.499617040318),
993: (235.88526, 0.431673556862),
994: (253.67465, 0.460225344488),
995: (314.17424, 0.14468466091),
996: (366.70649, 0.417017294763),
997: (578.89914, 0.512226738318),
998: (239.56853, 0.542883760596),
999: (325.53751, 0.424360539044),
...
}
Datasets :
d1_duration_hotness.p : Dataset produced from 1-M-Songs Dataset Meta Data containing :
• first_dim : Song duration
• second_dim : artist_hotttnesss
d2_release_artist_familiarity.p: Dataset produced from 1-M-Songs Dataset Meta Data containing :
• first_dim : Song release year
• second_dim : artist_familiarity
d3_tempo_loudness.p : Dataset produced from songsdata.csv :
• first_dim : Song tempo
• second_dim : Song loudness
Codes: Implementation of the algorithms are inside the "qual_analysis" folder
Use the following commands to compile the codes:
Before doing so, ensure your current directory is the "Qual_Analysis" folder. Then type as below:
▪ Python simulationRunner——MMRAtOnce.py (To run mmr)
▪ Python simulationRunner—OurAlgAtOnce.py (To run our algorithms)
▪ Python randomalg_simulation.py (To run the random algorithm)
Provide inputs as per the click commands. The results will be stored in the res folder upon running algorithms.