import streamlit as st
import math
import numpy as np
import psycopg2
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import train_test_split
import pandas as pd
from scipy.spatial import distance
import random
# Initialize connection.
# Uses st.experimental_singleton to only run once.
@st.experimental_singleton
def init_connection():
return psycopg2.connect(**st.secrets["postgres"])
list_selected_songs = []
list_allsongs = []
emotion = []
q1_song = []
q2_song = []
q3_song = []
q4_song = []
conn = init_connection()
# number of recommendation of the song to be made
num_recommendation = 10
# Perform query.
# Uses st.experimental_memo to only rerun when the query changes or after 10 min.
@st.experimental_memo(ttl=600)
def run_query(query):
with conn.cursor() as cur:
cur.execute(query)
return cur.fetchall()
def count_emotion(list_emotion):
s1 = 'q1'
s2 = 'q2'
s3 = 'q3'
s4 = 'q4'
global count_q1
global count_q2
global count_q3
global count_q4
count_q1 = list_emotion.count(s1)
count_q2 = list_emotion.count(s2)
count_q3 = list_emotion.count(s3)
count_q4 = list_emotion.count(s4)
def recommended_songs_q1_q2_q3_q4(num_recommendation_q1, num_recommendation_q2, num_recommendation_q3, num_recommendation_q4):
global q1
global q2
global q3
global g4
q1 = np.array(q1_song[0:int(num_recommendation_q1)])
q2 = np.array(q2_song[0:int(num_recommendation_q2)])
q3 = np.array(q3_song[0:int(num_recommendation_q3)])
q4 = np.array(q4_song[0:int(num_recommendation_q4)])
return q1, q2, q3, q4
def get_number_quadrant_recommendation():
total = count_q1+count_q2+count_q3+count_q4
percentage_q1 = count_q1/total
percentage_q2 = count_q2/total
percentage_q3 = count_q3/total
percentage_q4 = count_q4/total
global num_recommendation_q1
global num_recommendation_q2
global num_recommendation_q3
global num_recommendation_q4
num_recommendation_q1 = (percentage_q1*num_recommendation)
num_recommendation_q2 = (percentage_q2*num_recommendation)
num_recommendation_q3 = (percentage_q3*num_recommendation)
num_recommendation_q4 = (percentage_q4*num_recommendation)
def get_distance(e):
# index 6 consists of euclidian distance between two songs, seleted song and song in the database
return e[7]
def get_distance1(e):
# index 0 consists of euclidian distance between two songs in the case of the repeated song
return e[0]
rows = run_query("SELECT song_title, artist from nepali_songs order by song_title")
st.title('Emotion Based Music Recommendation System')
st.image('music.jpg')
with st.form("my_form"):
options = st.multiselect('Choose the song that you want to listen', [
':'.join(map(str, x)) for x in rows])
submitted = st.form_submit_button("Recommend song to me")
# splitting artist and song title of the song that user have selected
if submitted:
artist_song_list = []
for index, element in enumerate(options):
artist_list = element.split(':')
artist_song_list.append(artist_list)
# storing the information of the audio features of the songs that the user have selected by retrieving through the database
for x in artist_song_list:
with conn.cursor() as cur:
cur.execute(
"select * from nepali_songs where song_title=%s and artist=%s", [x[0], x[1]])
row = cur.fetchall()
list_selected_songs.append(row)
print("selected songs")
happy_songs=[]
tensed_songs=[]
sad_songs=[]
calm_songs=[]
# collecting the information about the emotion quadrant of the each song that the user have selected and storing in the list emotion
for i in list_selected_songs:
print(i[0][12],i[0][13])
emotion.append(i[0][14])
#categorizing the songs into 4 different emotions
if(i[0][14]=='q1'):
happy_songs.append(i[0][12])
elif(i[0][14]=='q2'):
tensed_songs.append(i[0][12])
elif(i[0][14]=='q3'):
sad_songs.append(i[0][12])
else:
calm_songs.append(i[0][12])
# calculating the total number of emotion quadrant of the song that the user have selected
count_emotion(emotion)
total = count_q1+count_q2+count_q3+count_q4
st.write("You have selected ", count_q1, " happy songs")
st.write(happy_songs)
st.write("You have selected ", count_q2, " tensed songs")
st.write(tensed_songs)
st.write("You have selected ", count_q3, " sad songs")
st.write(sad_songs)
st.write("You have selected ", count_q4, " calm songs")
st.write(calm_songs)
# retriving information of all songs in the database and storing in the list list_allsongs
distance_info_song_all = []
with conn.cursor() as cur:
cur.execute("select * from nepali_songs")
row = cur.fetchall()
# print(type(row[0]))
list_allsongs.append(row)
# calculating the euclidean distance of selected songs with all songs in the database
for i in list_selected_songs:
valence_selected_songs = float(i[0][11])
energy_selected_songs = float(i[0][2])
list1 = [valence_selected_songs, energy_selected_songs]
for lists in list_allsongs:
# print("printing the list of songs\n")
distance_info = []
for tuples in lists:
valence_allsongs = float(tuples[11])
energy_allsongs = float(tuples[2])
list2 = [valence_allsongs, energy_allsongs]
dist = distance.euclidean(list1, list2)
# i[0][0] is the spotifyid of the selected song and tuples[0] is the spotifyid
# of the song in the database
distance_info.append(
[i[0][12], i[0][14], i[0][0], tuples[12], tuples[14], tuples[13], tuples[0], dist])
distance_info_song_all.append(distance_info)
# st.write("distance info")
# st.write(distance_info_song_all)
nearest_neighbour = []
# distance_info_song_all consists of the euclidian distance of the selected song with all songs in the database
# recommend_list consist of the top 10 songs having minimum distance with all the selected song
for i in distance_info_song_all:
i.sort(key=get_distance, reverse=False)
#sorting the 10 nearest neighbour of the each selected song selected song
for k in range(0,10):
nearest_neighbour.append(i[k])
# st.write(nearest_neighbour)
# calculating the percentange of song associated with q1, q2, q3 and q4 that the user have selected
# st.write("q1 recommendation number", num_recommendation_q1)
# st.write("q2 recommendation number", num_recommendation_q2)
# st.write("q3 recommendation number", num_recommendation_q3)
# st.write("q4 recommendation numner", num_recommendation_q4)
# st.write(num_recommendation)
# selecting the song of the particular quadrant from the nearest neighbour and storing in the list
for t in nearest_neighbour:
if (t[4] == 'q2' and t[7]!=0.0):
q2_song.append(t)
elif (t[4] == 'q1' and t[7]!=0.0):
q1_song.append(t)
elif (t[4] == 'q3' and t[7]!=0.0):
q3_song.append(t)
elif (t[4] == 'q4' and t[7]!=0.0):
q4_song.append(t)
else:
pass
q1_song.sort(key=get_distance,reverse=False)
q2_song.sort(key=get_distance,reverse=False)
q3_song.sort(key=get_distance,reverse=False)
q4_song.sort(key=get_distance,reverse=False)
get_number_quadrant_recommendation()
q1, q2, q3, q4 = recommended_songs_q1_q2_q3_q4(
num_recommendation_q1, num_recommendation_q2, num_recommendation_q3, num_recommendation_q4)
# st.write("q1")
# st.write(q1)
# st.write("q2")
# st.write(q2)
# st.write("q3")
# st.write(q3)
# st.write("q4")
# st.write(q4)
w = set()
x = set()
y = set()
z = set()
for i in q1:
w.add(i[6])
for j in q2:
x.add(j[6])
for k in q3:
y.add(k[6])
for l in q4:
z.add(l[6])
# Checking whether the set w,x,z which consists of the unique spotifyid of the song in the q1, q2, q3, q4 consists of at least one element
if len(w) != 0:
# a set that consists of the spotify id of the repeated song
spotifyid=set()
#list consist of one copy of the repeated song with minimum distance
q1_=[]
for i in w:
lst1=[]
solutions=np.argwhere(q1==i)
#if the spotify id ofthat songs consists in the q1 more than one time
if(len(solutions)>1):
for i in range(0,len(solutions)):
# print(q1[solutions[i][0]][7])
# since that song is repeated more than once, collecing the info of the euclidian distance of that repeated songs in the lst1
lst1.append([q1[solutions[i][0]][7],q1[solutions[i][0]][0],q1[solutions[i][0]][1],q1[solutions[i][0]][3],q1[solutions[i][0]][2],q1[solutions[i][0]][5],q1[solutions[i][0]][4],q1[solutions[i][0]][6]])
spotifyid.add(q1[solutions[i][0]][6])
lst1.sort(key=get_distance1,reverse=False)
q1_.append(lst1[0])
#excluding all the repeated songs
for i in spotifyid:
q1 = [item for item in q1 if item[6]!= i]
#appending the only one copy of repeated song with minimum distance
for i in q1_:
q1.append(i)
if len(x) != 0:
# a set that consists of the spotify id of the repeated song
spotifyid=set()
q2_=[]
for i in x:
lst2=[]
solutions=np.argwhere(q2==i)
#if the spotify id ofthat songs consists in the q1 more than one time
if(len(solutions)>1):
for i in range(0,len(solutions)):
# print(q1[solutions[i][0]][7])
# since that song is repeated more than once, collecing the info of the euclidian distance of that repeated songs in the lst1
lst2.append([q2[solutions[i][0]][7],q2[solutions[i][0]][0],q2[solutions[i][0]][1],q2[solutions[i][0]][3],q2[solutions[i][0]][2],q2[solutions[i][0]][5],q2[solutions[i][0]][4],q2[solutions[i][0]][6]])
spotifyid.add(q2[solutions[i][0]][6])
lst2.sort(key=get_distance1,reverse=False)
q2_.append(lst2[0])
for i in spotifyid:
q2 = [item for item in q1 if item[6]!= i]
for i in q2_:
q2.append(i)
if len(y) != 0:
# a set that consists of the spotify id of the repeated song
spotifyid=set()
q3_=[]
for i in y:
lst3=[]
solutions=np.argwhere(q3==i)
#if the spotify id ofthat songs consists in the q1 more than one time
if(len(solutions)>1):
for i in range(0,len(solutions)):
# print(q1[solutions[i][0]][7])
# since that song is repeated more than once, collecing the info of the euclidian distance of that repeated songs in the lst1
lst3.append([q3[solutions[i][0]][7],q3[solutions[i][0]][0],q3[solutions[i][0]][1],q3[solutions[i][0]][3],q3[solutions[i][0]][2],q3[solutions[i][0]][5],q3[solutions[i][0]][4],q3[solutions[i][0]][6]])
spotifyid.add(q3[solutions[i][0]][6])
lst3.sort(key=get_distance1,reverse=False)
q3_.append(lst3[0])
for i in spotifyid:
q3 = [item for item in q3 if item[6]!= i]
for i in q3_:
q3.append(i)
if len(z) != 0:
# a set that consists of the spotify id of the repeated song
spotifyid=set()
q4_=[]
for i in z:
lst4=[]
solutions=np.argwhere(q4==i)
#if the spotify id ofthat songs consists in the q1 more than one time
if(len(solutions)>1):
for i in range(0,len(solutions)):
# print(q1[solutions[i][0]][7])
# since that song is repeated more than once, collecing the info of the euclidian distance of that repeated songs in the lst1
lst4.append([q4[solutions[i][0]][7],q4[solutions[i][0]][0],q4[solutions[i][0]][1],q4[solutions[i][0]][3],q4[solutions[i][0]][2],q4[solutions[i][0]][5],q4[solutions[i][0]][4],q4[solutions[i][0]][6]])
spotifyid.add(q4[solutions[i][0]][6])
lst4.sort(key=get_distance1,reverse=False)
q4_.append(lst4[0])
for i in spotifyid:
q4 = [item for item in q4 if item[6]!= i]
for i in q4_:
q4.append(i)
st.markdown(f'<h1 style="font-size:24px;">{"Recommened list of songs"}</h1>', unsafe_allow_html=True)
print("Recommended list of songs")
if(q1!=0):
st.write("Happy Songs\n")
for i in q1:
print(i[3]," ",i[6])
st.write(i[3],":",i[5])
if(q2!=0):
st.write("Tensed Songs\n")
for j in q2:
print(j[3]," ",j[6])
st.write(j[3],":",j[5])
if(q3!=0):
st.write("Sad Songs\n")
for k in q3:
print(k[3]," ",k[6])
st.write(k[3],":",k[5])
if(q4!=0):
st.write("Calm Songs\n")
for l in q4:
print(l[3]," ",l[6])
st.write(l[3],":",l[5])