def recommend_items(paper_symbol_U_1, paper_symbol_U_2, paper_symbol_V,
paper_symbol_H_1, paper_symbol_H_2, train_buy_set, alpha,
feature_k, item_k):
X = paper_symbol_U_1.dot(paper_symbol_V.T)
Y = paper_symbol_U_2.dot(paper_symbol_V.T)
A = paper_symbol_U_1.dot(paper_symbol_U_2.T) + paper_symbol_H_1.dot(
paper_symbol_H_2.T)
features = get_top_columns_index(X, feature_k)
#print features
_change_X_with_features(X, features)
#print 'change OK'
tmp1 = alpha * X.dot(Y.T) / (feature_k * 5)
tmp1.index = range(tmp1.shape[0])
tmp1.columns = range(tmp1.shape[1])
tmp2 = (1 - alpha) * A
tmp2.index = range(tmp2.shape[0])
tmp2.columns = range(tmp2.shape[1])
#print 'tmp1', sorted(tmp1.ix[0,:], reverse=True)[:8]
#print 'tmp2', sorted(tmp2.ix[0,:], reverse=True)[:8]
score = tmp1.add(tmp2, fill_value=0)
#print 'score', score.ix[0,:5]
#score.to_csv('score.csv', index=False)
#print score.ix[0,0]
items = get_top_columns_index(score, score.shape[1])
#print items[0]
result = set([])
for idx in range(len(items)):
val = items[idx]
user_items_num = 0
for cdx in val:
potentional_ui = (idx, cdx)
#if not potentional_ui in train_buy_set:
result.add(potentional_ui)
user_items_num += 1
if user_items_num >= item_k:
break
return result
def recommend_items(paper_symbol_U_1, paper_symbol_U_2, paper_symbol_V, paper_symbol_H_1, paper_symbol_H_2, train_buy_set, alpha, feature_k, item_k):
X = paper_symbol_U_1.dot(paper_symbol_V.T)
Y = paper_symbol_U_2.dot(paper_symbol_V.T)
A = paper_symbol_U_1.dot(paper_symbol_U_2.T) + paper_symbol_H_1.dot(paper_symbol_H_2.T)
features = get_top_columns_index(X, feature_k)
#print features
_change_X_with_features(X, features)
#print 'change OK'
tmp1 = alpha * X.dot(Y.T) / (feature_k * 5)
tmp1.index = range(tmp1.shape[0])
tmp1.columns = range(tmp1.shape[1])
tmp2 = (1 - alpha) * A
tmp2.index = range(tmp2.shape[0])
tmp2.columns = range(tmp2.shape[1])
#print 'tmp1', sorted(tmp1.ix[0,:], reverse=True)[:8]
#print 'tmp2', sorted(tmp2.ix[0,:], reverse=True)[:8]
score = tmp1.add(tmp2, fill_value=0)
#print 'score', score.ix[0,:5]
#score.to_csv('score.csv', index=False)
#print score.ix[0,0]
items = get_top_columns_index(score, score.shape[1])
#print items[0]
result = set([])
for idx in range(len(items)):
val = items[idx]
user_items_num = 0
for cdx in val:
potentional_ui = (idx, cdx)
#if not potentional_ui in train_buy_set:
result.add(potentional_ui)
user_items_num += 1
if user_items_num >= item_k:
break
return result
def recommend_features(paper_symbol_U_1, paper_symbol_V, feature_k):
X = paper_symbol_U_1.dot(paper_symbol_V.T)
features = get_top_columns_index(X, feature_k)
result = {}
for idx in range(len(features)):
result[idx] = set(features[idx])
return result
def recommend_features(paper_symbol_U_1, paper_symbol_V, feature_k):
X = paper_symbol_U_1.dot(paper_symbol_V.T)
features = get_top_columns_index(X, feature_k)
result = {}
for idx in range(len(features)):
result[idx] = set(features[idx])
return result
