def test(file, period=''):
import matplotlib.pyplot as plt
JCR = JamendoCsvReader_album(file)
if period=='':
period = JCR.period
if not period: raise Exception("can't find the period. define it manually putting it as second par")
AllColumns = WeightedDownloadsRate(JCR, period, True)
AllColumns['old_rating_rank'] = getRanks(AllColumns['rating'], reverse=True)
del AllColumns['rating']
#prepare a list for sorting
AllColumnsList = []
for i in range(len(AllColumns['id'])):
AllColumnsList += [dict([(key,AllColumns[key][i]) for key in AllColumns])]
chart = sorted(AllColumnsList, key=lambda x:x['rate'], reverse=True)
printthesecols = ['rate', 'downloads_all', 'weighted_avg_agreed_note', 'old_rating_rank']
if period=='total': printthesecols += ['days_since_publication']
for i in range(0,30):
print str(chart[i]['id'])+': '+str(['%s:%s' % (p[0],round(p[1],4)) for p in chart[i].items() if p[0] in printthesecols])
plt.figure()
rates = [col['rate'] for col in chart]
dl= nomramlizeTo0_1([col['downloads_all']+col['downloads_tracks'] for col in chart])
ls= nomramlizeTo0_1([col['listened_logged']+col['listened_anon'] for col in chart])
plt.plot(ls, 'g-', dl, 'b-', rates, 'r-')
plt.legend(['ls', 'dl', 'rate'])
if period=='total':
plt.figure()
days= nomramlizeTo0_1([col['days_since_publication'] for col in chart])
plt.plot(days, 'y-', rates, 'r-')
plt.legend(['days', 'rate'])
plt.figure()
revavg = nomramlizeTo0_1([col['weighted_avg_agreed_note'] for col in chart])
plt.plot(revavg, 'y-', rates, 'r-')
plt.legend(['weighted_avg_agreed_note', 'rate'])
plt.figure()
plt.hist(nomramlizeTo0_1([col['rate'] for col in chart if col['rate']>0]), bins=100, range=(0,1))
plt.show()
#test('stats_album_week.csv')
def WeightedDownloadsRate(JCR, period, istest=False):
print 'start WeightedDownloadsRate algo'
cols = ['id', 'downloads_all', 'listened_logged', 'listened_anon']
if period == 'total': cols += ['days_since_publication']
if istest: cols += ['rating', 'weighted_avg_agreed_note']
COLUMNS = JCR.getColumns(cols)
COLUMNS['downloads_tracks'] = np.zeros(JCR.len)
#fetch on albums
for i in range(JCR.len):
tracks = JCR.RelatedTracks.iterTrackJoinAlbum(COLUMNS['id'][i])
#fetch on tracks
trackdownloads, numoftracks = 0., 0.
for track in tracks:
numoftracks += 1.
trackdownloads += track['downloads_all']
COLUMNS['downloads_tracks'][i] = 0 if numoftracks==0 else trackdownloads/numoftracks
downloads, listens = COLUMNS['downloads_tracks'] + COLUMNS['downloads_all'], COLUMNS['listened_logged']+COLUMNS['listened_anon']
downloads, listens = nomramlizeTo0_1(np.log2( downloads +2)), nomramlizeTo0_1(np.log2( listens +2))
#as an absolute quantity, downloads are 3 times more important than listens
#weighted_quantity_index = nomramlizeTo0_1(np.true_divide(3*downloads + 1*listens, 4))
weighted_quantity_index = nomramlizeTo0_1(downloads)
if period=='total':
days = 1 + nomramlizeTo0_1(np.log2( COLUMNS['days_since_publication'] + 2 ))
rate = np.true_divide( weighted_quantity_index, days )
else: rate = weighted_quantity_index
if istest:
COLUMNS['rate'] = rate
return COLUMNS
else: return {'id': COLUMNS['id'], 'rate': rate}