Abdelrhman-Elruby's picture
Add application file
9b6c8d0
raw
history blame
26.6 kB
import pandas as pd
import spotipy
from spotipy.oauth2 import SpotifyOAuth, SpotifyClientCredentials
import yaml
import re
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
from sklearn.preprocessing import MinMaxScaler
import pickle
def playlist_model(url, model, max_gen=3, same_art=5):
log = []
Fresult = []
try:
log.append('Start logging')
uri = url.split('/')[-1].split('?')[0]
stream = open("Spotify/Spotify.yaml")
spotify_details = yaml.safe_load(stream)
auth_manager = SpotifyClientCredentials(
client_id=spotify_details['Client_id'], client_secret=spotify_details['client_secret'])
sp = spotipy.client.Spotify(auth_manager=auth_manager)
if model == 'Spotify Model':
def get_IDs(user, playlist_id):
try:
log.append('start playlist extraction')
track_ids = []
playlist = sp.user_playlist(user, playlist_id)
for item in playlist['tracks']['items']:
track = item['track']
track_ids.append(track['id'])
return track_ids
except Exception as e:
log.append('Failed to load the playlist')
log.append(e)
track_ids = get_IDs('Ruby', uri)
track_ids_uni = list(set(track_ids))
log.append('Starting Spotify Model')
Spotifyresult = pd.DataFrame()
for i in range(len(track_ids_uni)-5):
if len(Spotifyresult) >= 50:
break
try:
ff = sp.recommendations(seed_tracks=list(track_ids_uni[i:i+5]), limit=5)
except Exception as e:
log.append(e)
continue
for z in range(5):
result = pd.DataFrame([z+(5*i)+1])
result['uri'] = ff['tracks'][z]['id']
Spotifyresult = pd.concat([Spotifyresult, result], axis=0)
Spotifyresult.drop_duplicates(subset=['uri'], inplace=True,keep='first')
Fresult = Spotifyresult.uri[:50]
log.append('Model run successfully')
return Fresult, log
lendf=len(pd.read_csv('Data/streamlit.csv',usecols=['track_uri']))
dtypes = {'track_uri': 'object', 'artist_uri': 'object', 'album_uri': 'object', 'danceability': 'float16', 'energy': 'float16', 'key': 'float16',
'loudness': 'float16', 'mode': 'float16', 'speechiness': 'float16', 'acousticness': 'float16', 'instrumentalness': 'float16',
'liveness': 'float16', 'valence': 'float16', 'tempo': 'float16', 'duration_ms': 'float32', 'time_signature': 'float16',
'Track_release_date': 'int8', 'Track_pop': 'int8', 'Artist_pop': 'int8', 'Artist_genres': 'object'}
col_name= ['track_uri', 'artist_uri', 'album_uri', 'danceability', 'energy', 'key',
'loudness', 'mode', 'speechiness', 'acousticness', 'instrumentalness',
'liveness', 'valence', 'tempo', 'duration_ms', 'time_signature',
'Track_release_date', 'Track_pop', 'Artist_pop', 'Artist_genres']
try:
def get_IDs(user, playlist_id):
log.append('start playlist extraction')
track_ids = []
artist_id = []
playlist = sp.user_playlist(user, playlist_id)
for item in playlist['tracks']['items']:
track = item['track']
track_ids.append(track['id'])
artist = item['track']['artists']
artist_id.append(artist[0]['id'])
return track_ids, artist_id
except Exception as e:
log.append('Failed to load the playlist')
log.append(e)
track_ids, artist_id = get_IDs('Ruby', uri)
log.append("Number of Track : {}".format(len(track_ids)))
artist_id_uni = list(set(artist_id))
track_ids_uni = list(set(track_ids))
log.append("Number of unique Artists : {}".format(len(artist_id_uni)))
log.append("Number of unique Tracks : {}".format(len(track_ids_uni)))
def extract(track_ids_uni, artist_id_uni):
err = []
err.append('Start audio features extraction')
audio_features = pd.DataFrame()
for i in range(0, len(track_ids_uni), 25):
try:
track_feature = sp.audio_features(track_ids_uni[i:i+25])
track_df = pd.DataFrame(track_feature)
audio_features = pd.concat([audio_features, track_df], axis=0)
except Exception as e:
err.append(e)
continue
err.append('Start track features extraction')
track_ = pd.DataFrame()
for i in range(0, len(track_ids_uni), 25):
try:
track_features = sp.tracks(track_ids_uni[i:i+25])
for x in range(25):
track_pop = pd.DataFrame([track_ids_uni[i+x]], columns=['Track_uri'])
track_pop['Track_release_date'] = track_features['tracks'][x]['album']['release_date']
track_pop['Track_pop'] = track_features['tracks'][x]["popularity"]
track_pop['Artist_uri'] = track_features['tracks'][x]['artists'][0]['id']
track_pop['Album_uri'] = track_features['tracks'][x]['album']['id']
track_ = pd.concat([track_, track_pop], axis=0)
except Exception as e:
err.append(e)
continue
err.append('Start artist features extraction')
artist_ = pd.DataFrame()
for i in range(0, len(artist_id_uni), 25):
try:
artist_features = sp.artists(artist_id_uni[i:i+25])
for x in range(25):
artist_df = pd.DataFrame([artist_id_uni[i+x]], columns=['Artist_uri'])
artist_pop = artist_features['artists'][x]["popularity"]
artist_genres = artist_features['artists'][x]["genres"]
artist_df["Artist_pop"] = artist_pop
if artist_genres:
artist_df["genres"] = " ".join([re.sub(' ', '_', i) for i in artist_genres])
else:
artist_df["genres"] = "unknown"
artist_ = pd.concat([artist_, artist_df], axis=0)
except Exception as e:
err.append(e)
continue
try:
test = pd.DataFrame(
track_, columns=['Track_uri', 'Artist_uri', 'Album_uri'])
test.rename(columns={'Track_uri': 'track_uri',
'Artist_uri': 'artist_uri', 'Album_uri': 'album_uri'}, inplace=True)
audio_features.drop(
columns=['type', 'uri', 'track_href', 'analysis_url'], axis=1, inplace=True)
test = pd.merge(test, audio_features,
left_on="track_uri", right_on="id", how='outer')
test = pd.merge(test, track_, left_on="track_uri",
right_on="Track_uri", how='outer')
test = pd.merge(test, artist_, left_on="artist_uri",
right_on="Artist_uri", how='outer')
test.rename(columns={'genres': 'Artist_genres'}, inplace=True)
test.drop(columns=['Track_uri', 'Artist_uri_x',
'Artist_uri_y', 'Album_uri', 'id'], axis=1, inplace=True)
test.dropna(axis=0, inplace=True)
test['Track_pop'] = test['Track_pop'].apply(lambda x: int(x/5))
test['Artist_pop'] = test['Artist_pop'].apply(lambda x: int(x/5))
test['Track_release_date'] = test['Track_release_date'].apply(lambda x: x.split('-')[0])
test['Track_release_date'] = test['Track_release_date'].astype('int16')
test['Track_release_date'] = test['Track_release_date'].apply(lambda x: int(x/50))
test[['danceability', 'energy', 'key', 'loudness', 'mode', 'speechiness', 'acousticness', 'instrumentalness', 'liveness', 'valence', 'tempo', 'time_signature']] = test[[
'danceability', 'energy', 'key', 'loudness', 'mode', 'speechiness', 'acousticness', 'instrumentalness', 'liveness', 'valence', 'tempo', 'time_signature']].astype('float16')
test[['duration_ms']] = test[['duration_ms']].astype('float32')
test[['Track_release_date', 'Track_pop', 'Artist_pop']] = test[[
'Track_release_date', 'Track_pop', 'Artist_pop']].astype('int8')
except Exception as e:
err.append(e)
err.append('Finish extraction')
return test, err
test, err = extract(track_ids_uni, artist_id_uni)
for i in err:
log.append(i)
del err
grow = test.copy()
test['Artist_genres'] = test['Artist_genres'].apply(lambda x: x.split(" "))
tfidf = TfidfVectorizer(max_features=max_gen)
tfidf_matrix = tfidf.fit_transform(test['Artist_genres'].apply(lambda x: " ".join(x)))
genre_df = pd.DataFrame(tfidf_matrix.toarray())
genre_df.columns = ['genre' + "|" +i for i in tfidf.get_feature_names_out()]
genre_df = genre_df.astype('float16')
test.drop(columns=['Artist_genres'], axis=1, inplace=True)
test = pd.concat([test.reset_index(drop=True),genre_df.reset_index(drop=True)], axis=1)
Fresult = pd.DataFrame()
x = 1
for i in range(int(lendf/2), lendf+1, int(lendf/2)):
try:
df = pd.read_csv('Data/streamlit.csv',names= col_name,dtype=dtypes,skiprows=x,nrows=i)
log.append('reading data frame chunks from {} to {}'.format(x,i))
except Exception as e:
log.append('Failed to load grow')
log.append(e)
grow = grow[~grow['track_uri'].isin(df['track_uri'].values)]
df = df[~df['track_uri'].isin(test['track_uri'].values)]
df['Artist_genres'] = df['Artist_genres'].apply(lambda x: x.split(" "))
tfidf_matrix = tfidf.transform(df['Artist_genres'].apply(lambda x: " ".join(x)))
genre_df = pd.DataFrame(tfidf_matrix.toarray())
genre_df.columns = ['genre' + "|" +i for i in tfidf.get_feature_names_out()]
genre_df = genre_df.astype('float16')
df.drop(columns=['Artist_genres'], axis=1, inplace=True)
df = pd.concat([df.reset_index(drop=True),
genre_df.reset_index(drop=True)], axis=1)
del genre_df
try:
df.drop(columns=['genre|unknown'], axis=1, inplace=True)
test.drop(columns=['genre|unknown'], axis=1, inplace=True)
except:
log.append('genre|unknown not found')
log.append('Scaling the data .....')
if x == 1:
sc = pickle.load(open('Data/sc.sav','rb'))
df.iloc[:, 3:19] = sc.transform(df.iloc[:, 3:19])
test.iloc[:, 3:19] = sc.transform(test.iloc[:, 3:19])
log.append("Creating playlist vector")
playvec = pd.DataFrame(test.sum(axis=0)).T
else:
df.iloc[:, 3:19] = sc.transform(df.iloc[:, 3:19])
x = i
if model == 'Model 1':
df['sim']=cosine_similarity(df.drop(['track_uri', 'artist_uri', 'album_uri'], axis = 1),playvec.drop(['track_uri', 'artist_uri', 'album_uri'], axis = 1))
df['sim2']=cosine_similarity(df.iloc[:,16:-1],playvec.iloc[:,16:])
df['sim3']=cosine_similarity(df.iloc[:,19:-2],playvec.iloc[:,19:])
df = df.sort_values(['sim3','sim2','sim'],ascending = False,kind='stable').groupby('artist_uri').head(same_art).head(50)
Fresult = pd.concat([Fresult, df], axis=0)
Fresult = Fresult.sort_values(['sim3', 'sim2', 'sim'],ascending=False,kind='stable')
Fresult.drop_duplicates(subset=['track_uri'], inplace=True,keep='first')
Fresult = Fresult.groupby('artist_uri').head(same_art).head(50)
elif model == 'Model 2':
df['sim'] = cosine_similarity(df.iloc[:, 3:16], playvec.iloc[:, 3:16])
df['sim2'] = cosine_similarity(df.loc[:, df.columns.str.startswith('T') | df.columns.str.startswith('A')], playvec.loc[:, playvec.columns.str.startswith('T') | playvec.columns.str.startswith('A')])
df['sim3'] = cosine_similarity(df.loc[:, df.columns.str.startswith('genre')], playvec.loc[:, playvec.columns.str.startswith('genre')])
df['sim4'] = (df['sim']+df['sim2']+df['sim3'])/3
df = df.sort_values(['sim4'], ascending=False,kind='stable').groupby('artist_uri').head(same_art).head(50)
Fresult = pd.concat([Fresult, df], axis=0)
Fresult = Fresult.sort_values(['sim4'], ascending=False,kind='stable')
Fresult.drop_duplicates(subset=['track_uri'], inplace=True,keep='first')
Fresult = Fresult.groupby('artist_uri').head(same_art).head(50)
del test
try:
del df
log.append('Getting Result')
except:
log.append('Getting Result')
if model == 'Model 1':
Fresult = Fresult.sort_values(['sim3', 'sim2', 'sim'],ascending=False,kind='stable')
Fresult.drop_duplicates(subset=['track_uri'], inplace=True,keep='first')
Fresult = Fresult.groupby('artist_uri').head(same_art).track_uri.head(50)
elif model == 'Model 2':
Fresult = Fresult.sort_values(['sim4'], ascending=False,kind='stable')
Fresult.drop_duplicates(subset=['track_uri'], inplace=True,keep='first')
Fresult = Fresult.groupby('artist_uri').head(same_art).track_uri.head(50)
log.append('{} New Tracks Found'.format(len(grow)))
if(len(grow)>=1):
try:
new=pd.read_csv('Data/new_tracks.csv',dtype=dtypes)
new=pd.concat([new, grow], axis=0)
new=new[new.Track_pop >0]
new.drop_duplicates(subset=['track_uri'], inplace=True,keep='last')
new.to_csv('Data/new_tracks.csv',index=False)
except:
grow.to_csv('Data/new_tracks.csv', index=False)
log.append('Model run successfully')
except Exception as e:
log.append("Model Failed")
log.append(e)
return Fresult, log
def top_tracks(url,region):
uri = url.split('/')[-1].split('?')[0]
stream= open("Spotify/Spotify.yaml")
spotify_details = yaml.safe_load(stream)
auth_manager = SpotifyClientCredentials(client_id=spotify_details['Client_id'],client_secret=spotify_details['client_secret'])
sp = spotipy.client.Spotify(auth_manager=auth_manager)
log = []
Fresult = []
try:
log.append('Starting Spotify Model')
top=sp.artist_top_tracks(uri,country=region)
for i in range(10) :
Fresult.append(top['tracks'][i]['id'])
log.append('Model run successfully')
except Exception as e:
log.append("Model Failed")
log.append(e)
return Fresult,log
def song_model(url, model, max_gen=3, same_art=5):
log = []
Fresult = []
try:
log.append('Start logging')
uri = url.split('/')[-1].split('?')[0]
stream = open("Spotify/Spotify.yaml")
spotify_details = yaml.safe_load(stream)
auth_manager = SpotifyClientCredentials(
client_id=spotify_details['Client_id'], client_secret=spotify_details['client_secret'])
sp = spotipy.client.Spotify(auth_manager=auth_manager)
if model == 'Spotify Model':
log.append('Starting Spotify Model')
aa=sp.recommendations(seed_tracks=[uri], limit=25)
for i in range(25):
Fresult.append(aa['tracks'][i]['id'])
log.append('Model run successfully')
return Fresult, log
lendf=len(pd.read_csv('Data/streamlit.csv',usecols=['track_uri']))
dtypes = {'track_uri': 'object', 'artist_uri': 'object', 'album_uri': 'object', 'danceability': 'float16', 'energy': 'float16', 'key': 'float16',
'loudness': 'float16', 'mode': 'float16', 'speechiness': 'float16', 'acousticness': 'float16', 'instrumentalness': 'float16',
'liveness': 'float16', 'valence': 'float16', 'tempo': 'float16', 'duration_ms': 'float32', 'time_signature': 'float16',
'Track_release_date': 'int8', 'Track_pop': 'int8', 'Artist_pop': 'int8', 'Artist_genres': 'object'}
col_name= ['track_uri', 'artist_uri', 'album_uri', 'danceability', 'energy', 'key',
'loudness', 'mode', 'speechiness', 'acousticness', 'instrumentalness',
'liveness', 'valence', 'tempo', 'duration_ms', 'time_signature',
'Track_release_date', 'Track_pop', 'Artist_pop', 'Artist_genres']
log.append('Start audio features extraction')
audio_features = pd.DataFrame(sp.audio_features([uri]))
log.append('Start track features extraction')
track_ = pd.DataFrame()
track_features = sp.tracks([uri])
track_pop = pd.DataFrame([uri], columns=['Track_uri'])
track_pop['Track_release_date'] = track_features['tracks'][0]['album']['release_date']
track_pop['Track_pop'] = track_features['tracks'][0]["popularity"]
track_pop['Artist_uri'] = track_features['tracks'][0]['artists'][0]['id']
track_pop['Album_uri'] = track_features['tracks'][0]['album']['id']
track_ = pd.concat([track_, track_pop], axis=0)
log.append('Start artist features extraction')
artist_id_uni=list(track_['Artist_uri'])
artist_ = pd.DataFrame()
artist_features = sp.artists(artist_id_uni)
artist_df = pd.DataFrame(artist_id_uni, columns=['Artist_uri'])
artist_pop = artist_features['artists'][0]["popularity"]
artist_genres = artist_features['artists'][0]["genres"]
artist_df["Artist_pop"] = artist_pop
if artist_genres:
artist_df["genres"] = " ".join([re.sub(' ', '_', i) for i in artist_genres])
else:
artist_df["genres"] = "unknown"
artist_ = pd.concat([artist_, artist_df], axis=0)
try:
test = pd.DataFrame(track_, columns=['Track_uri', 'Artist_uri', 'Album_uri'])
test.rename(columns={'Track_uri': 'track_uri','Artist_uri': 'artist_uri', 'Album_uri': 'album_uri'}, inplace=True)
audio_features.drop(columns=['type', 'uri', 'track_href', 'analysis_url'], axis=1, inplace=True)
test = pd.merge(test, audio_features,left_on="track_uri", right_on="id", how='outer')
test = pd.merge(test, track_, left_on="track_uri",right_on="Track_uri", how='outer')
test = pd.merge(test, artist_, left_on="artist_uri",right_on="Artist_uri", how='outer')
test.rename(columns={'genres': 'Artist_genres'}, inplace=True)
test.drop(columns=['Track_uri', 'Artist_uri_x','Artist_uri_y', 'Album_uri', 'id'], axis=1, inplace=True)
test.dropna(axis=0, inplace=True)
test['Track_pop'] = test['Track_pop'].apply(lambda x: int(x/5))
test['Artist_pop'] = test['Artist_pop'].apply(lambda x: int(x/5))
test['Track_release_date'] = test['Track_release_date'].apply(lambda x: x.split('-')[0])
test['Track_release_date'] = test['Track_release_date'].astype('int16')
test['Track_release_date'] = test['Track_release_date'].apply(lambda x: int(x/50))
test[['danceability', 'energy', 'key', 'loudness', 'mode', 'speechiness', 'acousticness', 'instrumentalness', 'liveness', 'valence', 'tempo', 'time_signature']] = test[['danceability', 'energy', 'key', 'loudness', 'mode', 'speechiness', 'acousticness', 'instrumentalness', 'liveness', 'valence', 'tempo', 'time_signature']].astype('float16')
test[['duration_ms']] = test[['duration_ms']].astype('float32')
test[['Track_release_date', 'Track_pop', 'Artist_pop']] = test[['Track_release_date', 'Track_pop', 'Artist_pop']].astype('int8')
except Exception as e:
log.append(e)
log.append('Finish extraction')
grow = test.copy()
test['Artist_genres'] = test['Artist_genres'].apply(lambda x: x.split(" "))
tfidf = TfidfVectorizer(max_features=max_gen)
tfidf_matrix = tfidf.fit_transform(test['Artist_genres'].apply(lambda x: " ".join(x)))
genre_df = pd.DataFrame(tfidf_matrix.toarray())
genre_df.columns = ['genre' + "|" +i for i in tfidf.get_feature_names_out()]
genre_df = genre_df.astype('float16')
test.drop(columns=['Artist_genres'], axis=1, inplace=True)
test = pd.concat([test.reset_index(drop=True),genre_df.reset_index(drop=True)], axis=1)
Fresult = pd.DataFrame()
x = 1
for i in range(int(lendf/2), lendf+1, int(lendf/2)):
try:
df = pd.read_csv('Data/streamlit.csv',names= col_name,dtype=dtypes,skiprows=x,nrows=i)
log.append('reading data frame chunks from {} to {}'.format(x,i))
except Exception as e:
log.append('Failed to load grow')
log.append(e)
grow = grow[~grow['track_uri'].isin(df['track_uri'].values)]
df = df[~df['track_uri'].isin(test['track_uri'].values)]
df['Artist_genres'] = df['Artist_genres'].apply(lambda x: x.split(" "))
tfidf_matrix = tfidf.transform(df['Artist_genres'].apply(lambda x: " ".join(x)))
genre_df = pd.DataFrame(tfidf_matrix.toarray())
genre_df.columns = ['genre' + "|" +i for i in tfidf.get_feature_names_out()]
genre_df = genre_df.astype('float16')
df.drop(columns=['Artist_genres'], axis=1, inplace=True)
df = pd.concat([df.reset_index(drop=True),
genre_df.reset_index(drop=True)], axis=1)
del genre_df
try:
df.drop(columns=['genre|unknown'], axis=1, inplace=True)
test.drop(columns=['genre|unknown'], axis=1, inplace=True)
except:
log.append('genre|unknown not found')
log.append('Scaling the data .....')
if x == 1:
sc = pickle.load(open('Data/sc.sav','rb'))
df.iloc[:, 3:19] = sc.transform(df.iloc[:, 3:19])
test.iloc[:, 3:19] = sc.transform(test.iloc[:, 3:19])
log.append("Creating playlist vector")
playvec = pd.DataFrame(test.sum(axis=0)).T
else:
df.iloc[:, 3:19] = sc.transform(df.iloc[:, 3:19])
x = i
if model == 'Model 1':
df['sim']=cosine_similarity(df.drop(['track_uri', 'artist_uri', 'album_uri'], axis = 1),playvec.drop(['track_uri', 'artist_uri', 'album_uri'], axis = 1))
df['sim2']=cosine_similarity(df.iloc[:,16:-1],playvec.iloc[:,16:])
df['sim3']=cosine_similarity(df.iloc[:,19:-2],playvec.iloc[:,19:])
df = df.sort_values(['sim3','sim2','sim'],ascending = False,kind='stable').groupby('artist_uri').head(same_art).head(50)
Fresult = pd.concat([Fresult, df], axis=0)
Fresult = Fresult.sort_values(['sim3', 'sim2', 'sim'],ascending=False,kind='stable')
Fresult.drop_duplicates(subset=['track_uri'], inplace=True,keep='first')
Fresult = Fresult.groupby('artist_uri').head(same_art).head(50)
elif model == 'Model 2':
df['sim'] = cosine_similarity(df.iloc[:, 3:16], playvec.iloc[:, 3:16])
df['sim2'] = cosine_similarity(df.loc[:, df.columns.str.startswith('T') | df.columns.str.startswith('A')], playvec.loc[:, playvec.columns.str.startswith('T') | playvec.columns.str.startswith('A')])
df['sim3'] = cosine_similarity(df.loc[:, df.columns.str.startswith('genre')], playvec.loc[:, playvec.columns.str.startswith('genre')])
df['sim4'] = (df['sim']+df['sim2']+df['sim3'])/3
df = df.sort_values(['sim4'], ascending=False,kind='stable').groupby('artist_uri').head(same_art).head(50)
Fresult = pd.concat([Fresult, df], axis=0)
Fresult = Fresult.sort_values(['sim4'], ascending=False,kind='stable')
Fresult.drop_duplicates(subset=['track_uri'], inplace=True,keep='first')
Fresult = Fresult.groupby('artist_uri').head(same_art).head(50)
del test
try:
del df
log.append('Getting Result')
except:
log.append('Getting Result')
if model == 'Model 1':
Fresult = Fresult.sort_values(['sim3', 'sim2', 'sim'],ascending=False,kind='stable')
Fresult.drop_duplicates(subset=['track_uri'], inplace=True,keep='first')
Fresult = Fresult.groupby('artist_uri').head(same_art).track_uri.head(50)
elif model == 'Model 2':
Fresult = Fresult.sort_values(['sim4'], ascending=False,kind='stable')
Fresult.drop_duplicates(subset=['track_uri'], inplace=True,keep='first')
Fresult = Fresult.groupby('artist_uri').head(same_art).track_uri.head(50)
log.append('{} New Tracks Found'.format(len(grow)))
if(len(grow)>=1):
try:
new=pd.read_csv('Data/new_tracks.csv',dtype=dtypes)
new=pd.concat([new, grow], axis=0)
new=new[new.Track_pop >0]
new.drop_duplicates(subset=['track_uri'], inplace=True,keep='last')
new.to_csv('Data/new_tracks.csv',index=False)
except:
grow.to_csv('Data/new_tracks.csv', index=False)
log.append('Model run successfully')
except Exception as e:
log.append("Model Failed")
log.append(e)
return Fresult, log
def update_dataset():
col_name= ['track_uri', 'artist_uri', 'album_uri', 'danceability', 'energy', 'key',
'loudness', 'mode', 'speechiness', 'acousticness', 'instrumentalness',
'liveness', 'valence', 'tempo', 'duration_ms', 'time_signature',
'Track_release_date', 'Track_pop', 'Artist_pop', 'Artist_genres']
dtypes = {'track_uri': 'object', 'artist_uri': 'object', 'album_uri': 'object', 'danceability': 'float16', 'energy': 'float16', 'key': 'float16',
'loudness': 'float16', 'mode': 'float16', 'speechiness': 'float16', 'acousticness': 'float16', 'instrumentalness': 'float16',
'liveness': 'float16', 'valence': 'float16', 'tempo': 'float16', 'duration_ms': 'float32', 'time_signature': 'float16',
'Track_release_date': 'int8', 'Track_pop': 'int8', 'Artist_pop': 'int8', 'Artist_genres': 'object'}
df = pd.read_csv('Data/streamlit.csv',dtype=dtypes)
grow = pd.read_csv('Data/new_tracks.csv',dtype=dtypes)
cur = len(df)
df=pd.concat([df,grow],axis=0)
grow=pd.DataFrame(columns=col_name)
grow.to_csv('Data/new_tracks.csv',index=False)
df=df[df.Track_pop >0]
df.drop_duplicates(subset=['track_uri'],inplace=True,keep='last')
df.dropna(axis=0,inplace=True)
df.to_csv('Data/streamlit.csv',index=False)
return (len(df)-cur)