def genre_popularity(self):
self.text.delete(1.0, END)
self.ax.cla()
data = DB_CONNECTION.retrieve_data("genre_popularity")
Popularity.calculate(self, data)
self.f.canvas.draw()
self.f.savefig("Plots/user_rating_genre_dist")
def requires(self):
config = Config.get()
return (Coordinates.CreateSampleCoordinates(),
Popularity.SampleCreator(
config.get("ExternalFiles",
"vecs_with_id")), ContourCode(),
CreateContinents(), MakeRegions())
def predict():
print(request.form)
feature=[]
feature.append(731)
feature.append(len(request.form["title"].split())+2)
feature.append(len(request.form["text"].split()))
print(len(set(request.form["text"].split())))
print((len(request.form["text"].split()))+2)
feature.append(len(set(request.form["text"].split()))/(len(request.form["text"].split())+13))
feature.append(len(set(request.form["title"].split()))/len(request.form["title"].split()))
feature.append(len(set(request.form["title"].split()))/len(request.form["title"].split()))
feature.append(int(request.form["external_links"])+int(request.form["self_links"]));
feature.append(int(request.form["self_links"]));
feature.append(int(request.form["image"]));
feature.append(int(request.form["video"]));
words = request.form["text"].split()
feature.append(sum(len(word) for word in words) / (len(words)+13))
feature.append(0)
print("Fyk")
feature.append(1 if request.form["channel"]=="1" else 0)
feature.append(1 if request.form["channel"]=="2" else 0)
feature.append(1 if request.form["channel"]=="3" else 0)
feature.append(1 if request.form["channel"]=="4" else 0)
feature.append(1 if request.form["channel"]=="5" else 0)
feature.append(1 if request.form["channel"]=="6" else 0)
feature.append(int(request.form["max_external_share"]))
feature.append(int(request.form["avg_self_share"]))
print("Shik")
feature.append(1 if request.form["day"]=="1" else 0)
feature.append(1 if request.form["day"]=="2" else 0)
feature.append(1 if request.form["day"]=="3" else 0)
feature.append(1 if request.form["day"]=="4" else 0)
feature.append(1 if request.form["day"]=="5" else 0)
feature.append(1 if request.form["day"]=="6" else 0)
feature.append(1 if request.form["day"]=="7" else 0)
feature.append(1 if request.form["day"]=="6" or request.form["day"]=="7" else 0 )
feature.append(TextBlob(request.form["text"]).sentiment.subjectivity)
feature.append(TextBlob(request.form["text"]).sentiment.polarity)
feature.append(TextBlob(request.form["title"]).sentiment.subjectivity)
feature.append(TextBlob(request.form["title"]).sentiment.polarity)
feature.append(abs(TextBlob(request.form["title"]).sentiment.subjectivity))
feature.append(abs(TextBlob(request.form["title"]).sentiment.polarity))
print(len(feature))
result = ""
if Popularity.predictPopularity(feature) == 1.0:
result = "Popular"
else:
result = "Not Popular"
result = " Your News Article will be " +result
return render_template('Result.html',message =result)
def k_means_clustering(no_of_clusters, item_popularity_df, articles_df):
x_df = pd.read_csv('shared_articles.csv')
x = x_df['text']
model = KMeans(n_clusters=no_of_clusters,
init='k-means++',
max_iter=100,
n_init=1)
#model.fit(x)
print("\n")
popularityModel = Popularity.PopularityRecommender(item_popularity_df,
articles_df)
list_of_popular_items_clusterwise = popularityModel.recommend_items(
user_id=1, topn=no_of_clusters, verbose=True)
return list_of_popular_items_clusterwise
def popularity(user_id, no=20):
global songs_to_ix, ix_to_songs, id_to_no, no_to_id, train_data, test_data, users, songs, song_db_subset
User_no = id_to_no[user_id]
song_db_1 = pd.read_csv('/home/niket/Desktop/triplet.csv')
song_db_2 = pd.read_csv('/home/niket/Downloads/song_data.csv', )
song_db = pd.merge(song_db_1,
song_db_2.drop_duplicates(['song_id']),
on="song_id",
how="left")
song_db['song'] = song_db['title'] + "-" + song_db['artist_name']
song_grouped = song_db.groupby(['song']).agg({
'listen_count': 'count'
}).reset_index()
grouped_sum = song_grouped['listen_count'].sum()
train_data, test_data = train_test_split(song_db_subset,
test_size=.20,
random_state=0)
#print(train_data)
pm = Popularity.popularity_recommender()
pm.make(train_data, 'user_id', 'song', no)
return (list(pm.recommend_songs(user_id)['song']))
ran_N[0] += 10
for N in ran_N:
UID_N_rank = {}
for fn in UID_rank.keys():
UID_N_rank[fn] = {}
for i in UID_rank[fn]:
UID_N_rank[fn][i] = UID_rank[fn][i][:N]
div_N_val[N] = {}
fre_N_val[N] = {}
adv_N_val[N] = {}
pop_N_val[N] = {}
print 'N = %d' % (N)
print 'fn--div--fer--pop--adv\n'
for fn in UID_N_rank.keys():
fre_val = FRE.us_fre(UID_N_rank[fn], song_info, real_sid)
pop_val = POP.us_pop(UID_N_rank[fn], SID_pop)
div_val = DIV.us_div(UID_N_rank[fn], song_info, real_sid)
adv_val = ADV.us_adv(UID_AID_tr_dict, UID_N_rank[fn], song_info,
real_sid)
print '%s' % (fn.split('/')[-1]),
print '--%.4f' % (div_val.mean()),
print '--%d' % (fre_val.mean()),
print '--%d' % (pop_val.mean()),
print '--%.3f' % (adv_val.mean())
div_N_val[N][fn] = div_val.mean()
fre_N_val[N][fn] = fre_val.mean()
adv_N_val[N][fn] = adv_val.mean()
pop_N_val[N][fn] = pop_val.mean()
'''
print 'fn,div,fer,pop\n'
for fn in pred_fn.keys():
def requires(self):
return (MakeRegions(), CalculateZoomsCode(),
Coordinates.CreateFullCoordinates(),
Popularity.PopularityIdentifier())
#Top-N accuracy metrics consts
#EVAL_RANDOM_SAMPLE_NON_INTERACTED_ITEMS = 100
model_evaluator = ModelEvaluator()
# In[15]:
#Computes the most popular items
item_popularity_df = interactions_full_df.groupby(
'contentId')['eventStrength'].sum().sort_values(
ascending=False).reset_index()
item_popularity_df.head(10)
# In[16]:
popularity_model = Popularity.PopularityRecommender(item_popularity_df,
articles_df)
# In[17]:
#print('Evaluating Popularity recommendation model...')
#pop_global_metrics, pop_detailed_results_df = model_evaluator.evaluate_model(popularity_model)
#print('\nGlobal metrics:\n%s' % pop_global_metrics)
#pop_detailed_results_df.head(10)
# In[18]:
#nltk.download('stopwords')
# In[19]:
#CONTENT BASED
def requires(self):
return (Popularity.PopularityIdentifier(), PopularityLabelSizerCode())
#real sid
song_id_fn = '/home/fearofchou/data/KKBOX/stats/AW_song_id.npy'
real_sid = np.sort(np.load(song_id_fn))
#========================================================
#each user dict
SID_dict = {}
for idx, val in enumerate(SID):
try:
SID_dict[val].append(UID[idx])
except:
SID_dict[val] = [UID[idx]]
#========================================================
import Diversity as DIV
import Popularity as POP
import Freshness as FRE
UID_div = DIV.us_div(UID_dict, song_meta, real_sid)
UID_fre = FRE.us_fre(UID_dict, song_meta, real_sid)
UID_pop = POP.us_pop(SID, UID_dict)
UID_us = {}
UID_us['Freshness'] = UID_fre
UID_us['Popularity'] = UID_pop
UID_us['Diversity'] = UID_div
import sys
sys.path.append('/home/fearofchou/code/Recsys/evul/')
from evul_all import *
song_db_2 = pd.read_csv('/home/niket/Downloads/song_data.csv', )
song_db = pd.merge(song_db_1,
song_db_2.drop_duplicates(['song_id']),
on="song_id",
how="left")
song_db['song'] = song_db['title'] + "-" + song_db['artist_name']
song_grouped = song_db.groupby(['song']).agg({
'listen_count': 'count'
}).reset_index()
grouped_sum = song_grouped['listen_count'].sum()
train_data, test_data = train_test_split(song_db_subset,
test_size=.20,
random_state=0)
#print(train_data)
pm = Popularity.popularity_recommender()
pm.make(train_data, 'user_id', 'song', No)
print(pm.recommend_songs(User))
#l=list(pm.recommend_songs(User)['song'])
#for i in l:
# print(i)
else:
#content based filtering
song = []
if (len(user_data['song'].unique()) < 10):
print("20 % Content based filtering \n")
#print("80 % 20% content \n")
def getFinal():
uid = "1"
file = open("user.txt", "r")
uid = file.read()
query = "select user_history.uid,user_history.pid,phonefeatures.ModelName from user_history,phonefeatures where user_history.uid=" + uid + " and phonefeatures.PhoneID=user_history.pid order by user_history.no_of_times_searched desc"
cursor = db.mydb.cursor()
cursor.execute(query)
result = cursor.fetchall()
global pidlist
pidlist = []
flag = False
phone_to_avoid = 122
if (len(result) == 0):
print("User Has No Search History...")
pr.popmodel()
else:
global user_id
user_id = 0
print("User Has Search History...")
print(result)
for res in result:
uid = res[0]
user_id = uid
pid = res[1]
query = "select * from purchased_phones where uid=" + str(
uid) + " and pid=" + str(pid)
cursor.execute(query)
purchased = cursor.fetchall()
if len(purchased) != 0:
#querystmt = "select * from purchased_phones where uid="+str(uid)+" and pid="+str(pid);#+"and overall>=3";
#cursor.execute(querystmt)
#checkres = cursor.fetchall()
#print(checkres)
if purchased[0][3] >= 3:
pidlist.append(res[1])
else:
print("This phone has rating less than 3....")
#co.collaborative(uid,res[1])
phone_to_avoid = res[1]
flag = True
else:
print(
"Entry for this phone exists in search history but not in purchased phones...."
)
list1 = []
list2 = []
if (flag == True):
list1 = co.collaborative(user_id, phone_to_avoid)
if len(pidlist) != 0:
list2 = cb2.ContentBased(pidlist)
uniquePhones = set()
for l in list1:
uniquePhones.add(l)
for l in list2:
uniquePhones.add(l)
NewList = []
i = 0
while len(NewList) < 10:
if list1[i] in uniquePhones:
NewList.append(list1[i])
uniquePhones.remove(list1[i])
if list2[i] in uniquePhones:
NewList.append(list2[i])
uniquePhones.remove(list2[i])
i = i + 1
return NewList