def vector_score_function(user_id, item_id, star=0):
"tfidf and vector model with different predictors"
if user_id in user_matrix and item_id in item_matrix:
mf_score = cal_score(user_id, item_id, user_matrix, item_matrix,
user_bias, item_bias, global_bias)
lr_score = vec_predictor.predict(user_id, item_id)
if type(lr_score) != float:
raise Exception('predictor return type error')
print '%s\t%s\t%lf\t%lf\t%lf\t%lf\t%lf' % (
user_id, item_id, mf_score, lr_score, star,
star - mf_score - lr_score, star - mf_score)
#return cal_score(user_id, item_id, user_matrix, item_matrix, user_bias, item_bias, global_bias) + tfidf_predictor.predict(user_id, item_id)
return mf_score + lr_score
elif user_id in user_matrix:
global user_miss
user_miss += 1
return user_score[user_id]['score'] * 1.0 / user_score[user_id]['count']
elif item_id in item_matrix:
global item_miss
item_miss += 1
return item_score[item_id]['score'] * 1.0 / item_score[item_id]['count']
else:
global all_miss
all_miss += 1
#return random.randint(1, 5)
return 0
def vector_score_function(user_id, item_id, star=0):
"tfidf and vector model with different predictors"
if user_id in user_matrix and item_id in item_matrix:
mf_score = cal_score(user_id, item_id, user_matrix, item_matrix, user_bias, item_bias, global_bias)
lr_score = vec_predictor.predict(user_id, item_id)
if type(lr_score) != float:
raise Exception("predictor return type error")
print "%s\t%s\t%lf\t%lf\t%lf\t%lf\t%lf" % (
user_id,
item_id,
mf_score,
lr_score,
star,
star - mf_score - lr_score,
star - mf_score,
)
# return cal_score(user_id, item_id, user_matrix, item_matrix, user_bias, item_bias, global_bias) + tfidf_predictor.predict(user_id, item_id)
return mf_score + lr_score
elif user_id in user_matrix:
global user_miss
user_miss += 1
return user_score[user_id]["score"] * 1.0 / user_score[user_id]["count"]
elif item_id in item_matrix:
global item_miss
item_miss += 1
return item_score[item_id]["score"] * 1.0 / item_score[item_id]["count"]
else:
global all_miss
all_miss += 1
# return random.randint(1, 5)
return 0
def mf_score_function(user_id, item_id, star=0):
if user_id in user_matrix and item_id in item_matrix:
return cal_score(user_id, item_id, user_matrix, item_matrix, user_bias, item_bias, global_bias)
elif user_id in user_matrix:
global user_miss
user_miss += 1
return user_score[user_id]["score"] * 1.0 / user_score[user_id]["count"]
elif item_id in item_matrix:
global item_miss
item_miss += 1
return item_score[item_id]["score"] * 1.0 / item_score[item_id]["count"]
else:
global all_miss
all_miss += 1
return 0
def mf_score_function(user_id, item_id, star=0):
if user_id in user_matrix and item_id in item_matrix:
return cal_score(user_id, item_id, user_matrix, item_matrix, user_bias,
item_bias, global_bias)
elif user_id in user_matrix:
global user_miss
user_miss += 1
return user_score[user_id]['score'] * 1.0 / user_score[user_id]['count']
elif item_id in item_matrix:
global item_miss
item_miss += 1
return item_score[item_id]['score'] * 1.0 / item_score[item_id]['count']
else:
global all_miss
all_miss += 1
return 0
def vector_score_function(user_id, item_id, star = 0):
"tfidf and vector model with different predictors"
if user_id in user_matrix and item_id in item_matrix:
mf_score = cal_score(user_id, item_id, user_matrix, item_matrix, user_bias, item_bias, global_bias)
lr_score = vec_predictor.predict(user_id, item_id)
#return cal_score(user_id, item_id, user_matrix, item_matrix, user_bias, item_bias, global_bias) + tfidf_predictor.predict(user_id, item_id)
return mf_score + lr_score
elif user_id in user_matrix:
global user_miss
user_miss += 1
return user_score[user_id]['score'] * 1.0 / user_score[user_id]['count']
elif item_id in item_matrix:
global item_miss
item_miss += 1
return item_score[item_id]['score'] * 1.0 / item_score[item_id]['count']
else:
global all_miss
all_miss += 1
#return random.randint(1, 5)
return 0
def mf_score_function(user_id, item_id, star = 0):
if user_id in user_matrix and item_id in item_matrix:
mf_score = cal_score(user_id, item_id, user_matrix, item_matrix, user_bias, item_bias, global_bias)
residual_score = 0.0
if user_id in user_vec:
residual_score += np.dot(user_vec[user_id], user_weight)
if item_id in item_vec:
residual_score += np.dot(item_vec[item_id], item_weight)
return mf_score + residual_score
elif user_id in user_matrix:
global user_miss
user_miss += 1
return user_score[user_id]['score'] * 1.0 / user_score[user_id]['count']
elif item_id in item_matrix:
global item_miss
item_miss += 1
return item_score[item_id]['score'] * 1.0 / item_score[item_id]['count']
else:
global all_miss
all_miss += 1
return 0
def vector_score_function(user_id, item_id, star=0):
"tfidf and vector model with different predictors"
if user_id in user_matrix and item_id in item_matrix:
mf_score = cal_score(user_id, item_id, user_matrix, item_matrix,
user_bias, item_bias, global_bias)
lr_score = vec_predictor.predict(user_id, item_id)
#return cal_score(user_id, item_id, user_matrix, item_matrix, user_bias, item_bias, global_bias) + tfidf_predictor.predict(user_id, item_id)
return mf_score + lr_score
elif user_id in user_matrix:
global user_miss
user_miss += 1
return user_score[user_id]['score'] * 1.0 / user_score[user_id]['count']
elif item_id in item_matrix:
global item_miss
item_miss += 1
return item_score[item_id]['score'] * 1.0 / item_score[item_id]['count']
else:
global all_miss
all_miss += 1
#return random.randint(1, 5)
return 0
