def alternateOptimization(opinion_matrix, opinion_matrix_I, rating_matrix,
NUM_OF_FACTORS, MAX_DEPTH, File):
# Save and print the Number of Users and Movies
NUM_USERS = rating_matrix.shape[0]
NUM_MOVIES = rating_matrix.shape[1]
NUM_FEATURE = opinion_matrix.shape[1]
print("Number of Users", NUM_USERS)
print("Number of Item", NUM_MOVIES)
print("Number of Feature", NUM_FEATURE)
print("Number of Latent Factors: ", NUM_OF_FACTORS)
# Create the user and item profile vector of appropriate size.
# Initialize the item vectors according to MF
user_vectors, item_vectors = MF(20, 0.05, 0.02, 0.02, 100, File)
# user_vectors = np.random.rand(NUM_USERS, NUM_OF_FACTORS)
# item_vectors = np.random.rand(NUM_MOVIES, NUM_OF_FACTORS)
i = 0
print("Entering Main Loop of alternateOptimization")
decTree = dtree.Tree(dtree.Node(None, 1), NUM_OF_FACTORS, MAX_DEPTH)
# Do converge Check
while i < 5:
# Create the decision Tree based on item_vectors
#print("Creating Tree.. for i = ", i, "for user")
#decTree = dtree.Tree(dtree.Node(None, 1), NUM_OF_FACTORS, MAX_DEPTH)
#decTree.fitTree_U(decTree.root, opinion_matrix, rating_matrix, item_vectors, NUM_OF_FACTORS)
#print("print user tree ", i)
#decTree.printtree(decTree.root)
print("Getting the user vectors from tree")
# Calculate the User vectors using dtree
user_vectors_before = user_vectors
#user_vectors = decTree.getVectors_f(opinion_matrix, NUM_OF_FACTORS)
# adding personalized term
for index in range(len(rating_matrix)):
indice = np.array([index])
user_vectors[index] = opt.cf_user(rating_matrix, item_vectors,
user_vectors[index], indice,
NUM_OF_FACTORS)
print("Creating Tree.. for i = ", i, "for item")
decTreeI = dtree.Tree(dtree.Node(None, 1), NUM_OF_FACTORS, MAX_DEPTH)
decTreeI.fitTree_I(decTreeI.root, opinion_matrix_I, rating_matrix,
user_vectors, NUM_OF_FACTORS)
print("print item tree ", i)
decTreeI.printtree(decTreeI.root)
print("Getting the item vectors from tree")
item_vectors_before = item_vectors
item_vectors = decTreeI.getVectors_f(opinion_matrix_I, NUM_OF_FACTORS)
for index in range(len(rating_matrix[0])):
indice = np.array([index])
item_vectors[index] = opt.cf_item(rating_matrix, user_vectors,
item_vectors[index], indice,
NUM_OF_FACTORS)
# Calculate Error for Convergence check
Pred_before = np.dot(user_vectors_before, item_vectors_before.T)
Pred = np.dot(user_vectors, item_vectors.T)
Error = Pred_before - Pred
Error = Error.flatten()
error = np.dot(Error, Error)
if error < 0.1:
break
i = i + 1
return decTree, decTreeI, user_vectors, item_vectors
def fitTree_U(self, current_node, opinion_matrix, rating_matrix,
item_vectors, K):
# rating_matrix only consists of rows which are users corresponding to the current Node
# Check if the maxDepth is reached
print("current depth of the tree", current_node.depth)
t1 = time.time()
if current_node.depth + 1 > self.max_depth:
return
if len(rating_matrix) == 0:
return
# Calulate the Error Before the Split
print("Calculate error")
error_before = opt.lossfunction_all(rating_matrix, item_vectors,
current_node.vector, 1)
print("Error Before: ", error_before)
# Create a numy_array to hold the split_criteria Values
params = {}
feature_splitpoint_matrix = []
count = 0
# pool = mp.Pool(20)
for feature_index in range(len(opinion_matrix[0])):
# Split the rating_matrix into like, dislike and unknown
NUMBER_OF_BIN = 5
# for each feature find the split_points
split_points = self.find_split_point(opinion_matrix, feature_index,
NUMBER_OF_BIN)
feature_splitpoint_matrix.append(split_points)
for split_point in split_points:
(indices_like, indices_dislike,
indices_unknown) = split(opinion_matrix, feature_index,
split_point)
params[count] = []
params[count].extend(
(rating_matrix, item_vectors, current_node.vector,
indices_like, indices_dislike, indices_unknown, K))
count += 1
# Calculate the split criteria value
print("Calculating the split criteria value")
results = []
params_index = 0
for feature_index in range(len(opinion_matrix[0])):
print("feature_index", feature_index)
temp = []
# start = time.time()
# result = pool.apply_async(opt.cal_splitvalue, params[feature_index])
for split_point in feature_splitpoint_matrix[feature_index]:
print("split_point", split_point)
result = opt.cal_splitvalue(
params[params_index][0], params[params_index][1],
params[params_index][2], params[params_index][3],
params[params_index][4], params[params_index][5],
params[params_index][6])
params_index += 1
temp.append(result)
results.append(temp)
#results = np.array(results)
temp_value = []
temp_index = []
for i in range(len(opinion_matrix[0])):
# temp_value is the min results for each feature
temp_value.append(min(results[i]))
# temp_index is the corresponding index of split_point for each feature
temp_index.append(results[i].index(min(results[i])))
# get the best feature index
bestFeature = temp_value.index(min(temp_value))
# get the corresponding value of split_point
best_split_point = feature_splitpoint_matrix[bestFeature][
temp_index[bestFeature]]
#for feature_index in range(len(opinion_matrix[0])):
# # split_values[feature_index] = results[feature_index].get()
# split_values[feature_index] = results[feature_index]
# pool.close()
# pool.join()
#bestFeature = np.argmin(split_values)
print("bestFeature index: ", bestFeature)
print("Split point:", best_split_point)
t2 = time.time()
print("Time used to create the layer: ", t2 - t1)
# Store the feature_index for the current_node
current_node.feature_index = bestFeature
# Store the split_point for the current_node
current_node.split_point = best_split_point
# Split the rating_matrix into like, dislike and unknown
(indices_like, indices_dislike,
indices_unknown) = split(opinion_matrix, bestFeature,
best_split_point)
split_value = opt.cal_splitvalue(rating_matrix, item_vectors,
current_node.vector, indices_like,
indices_dislike, indices_unknown, K)
like = rating_matrix[indices_like]
like_op = opinion_matrix[indices_like]
dislike = rating_matrix[indices_dislike]
dislike_op = opinion_matrix[indices_dislike]
unknown = rating_matrix[indices_unknown]
unknown_op = opinion_matrix[indices_unknown]
# Calculate the User Profile Vector for each of the three classes
# print "optimizing like, dislike and unknown..."
# Calculate the User Profile Vector for each of the three classes
like_vector = current_node.vector
dislike_vector = current_node.vector
unknown_vector = current_node.vector
if len(indices_like) > 0:
like_vector = opt.cf_user(rating_matrix, item_vectors,
current_node.vector, indices_like, K)
if len(indices_dislike) > 0:
dislike_vector = opt.cf_user(rating_matrix, item_vectors,
current_node.vector, indices_dislike,
K)
if len(indices_unknown) > 0:
unknown_vector = opt.cf_user(rating_matrix, item_vectors,
current_node.vector, indices_unknown,
K)
# CONDITION check condition RMSE Error check is CORRECT
if split_value < error_before:
# Recursively call the fitTree_f function for like, dislike and unknown Nodes creation
current_node.like = Node(current_node, current_node.depth + 1)
current_node.like.vector = like_vector
if len(like) != 0:
self.fitTree_U(current_node.like, like_op, like, item_vectors,
K)
current_node.dislike = Node(current_node, current_node.depth + 1)
current_node.dislike.vector = dislike_vector
if len(dislike) != 0:
self.fitTree_U(current_node.dislike, dislike_op, dislike,
item_vectors, K)
current_node.unknown = Node(current_node, current_node.depth + 1)
current_node.unknown.vector = unknown_vector
if len(unknown) != 0:
self.fitTree_U(current_node.unknown, unknown_op, unknown,
item_vectors, K)
else:
print("can't spilt")