def add_data(self, env_id):
"""
Start a data generation process.
"""
self.data.append(
DataProcess(self.experience_q[env_id], env_id, len(self.data)))
self.data[-1].start()
def __init__(self, process=DataProcess()):
self.model = load_model('./seq2seq_keras.h5')
self.process = process
self.PAD_INDEX = 0
self.STA_INDEX = 1
self.END_INDEX = 2
self.UNK_INDEX = 3
self.max_sequences = 60
self.ENCODER_INPUT = 0
def load_data(self):
mydata = DataProcess(self.img_rows, self.img_cols)
imgs_train, imgs_mask_train = mydata.load_train_data()
imgs_test = mydata.load_test_data()
return imgs_train, imgs_mask_train, imgs_test
hit += 1
recall += len(tu)
precision += N
# print((hit / (recall * 1.0),hit / (precision * 1.0)))
return (hit / (recall * 1.0), hit / (precision * 1.0))
if __name__ == "__main__":
N = 10
# 输入的数据集
totalData = './ml-100k/u.data' # dataset
trainFile = './ml-100k/u1.base' # train_set
testFile = './ml-100k/u1.test' # test_set
data = DataProcess(totalData)
trainData = DataProcess(trainFile)
testData = DataProcess(testFile)
userCF = UserCFPearson(data, trainData, testData)
# According to the training set and test set, the test
# result set of predicting test result is obtained,
# which is the same as the number of rows of test result set.
# Calculate model accuracy based on test set and prediction result set
# rmse, mae = userCF.calRmseAndMae(K)
# print('rmse: %1.5f\t mae: %1.5f' % (rmse, mae))
TopN, _ = userCF.recommendation(userID=123, N=10, K=10)
i = 1
print('userID = 123, N = 10, K = 10')
from unet import *
from data import DataProcess
mydata = DataProcess(512,512)
imgs_test = mydata.load_test_data()
myunet = myUnet()
model = myunet.get_unet()
model.load_weights('unet.hdf5')
imgs_mask_test = model.predict(imgs_test, verbose=1)
np.save('imgs_mask_test.npy', imgs_mask_test)
return precision, recall, coverage, popularity
def idToTitle(self, movieID, data):
for index, row in data.iterrows():
if movieID == row['MovieID']:
return row['MovieTitle']
if __name__ == '__main__':
# 输入的数据集
totalData = './ml-100k/u.data' #总数据集
trainFile = './ml-100k/u1.base' #训练集
testFile = './ml-100k/u1.test' #测试集
data = DataProcess(totalData)
trainData = DataProcess(trainFile)
testData = DataProcess(testFile)
ItemCF = ItemBasedCF(trainData, testData)
movie_df = data.getMovies()
N = 10
recd = ItemCF.Recommendation(userID=123, K=10, N=10)
i = 1
print("userID=123,K=10,N=10")
for key in recd.items():
print("Top: ", i, ItemCF.idToTitle(key[0], movie_df))
i += 1
print("%5s%5s%20s%20s%20s%20s" %
('K', 'N', 'precision(%)', "recall(%)", 'coverage(%)', 'popularity'))
def run(user_question, seq2seq=Seq2seq(), proc=DataProcess()):
pro_sent = proc.prepocess_sentence(user_question)
txt_to_idx = seq2seq.convert_text_to_index(pro_sent, proc.word2idx)
pred = seq2seq.predict_model()
sentence = seq2seq.idx_to_sentence(txt_to_idx, pred)
return sentence_spacing(sentence)
def run(userID, method):
# 输入的数据集
totalData = './ml-100k/u.data' # 总数据集
trainFile = './ml-100k/u1.base' # 训练集
testFile = './ml-100k/u1.test' # 测试集
# 参数 #用户ID
K = 5 # K为选取相邻用户个数
N = 5 # 推荐没有接触过的物品的个数
threshold = 4 # 评分阀值
data = DataProcess(totalData)
trainData = DataProcess(trainFile)
testData = DataProcess(testFile)
if method == 'userbased-pearson':
#userBased - pearson
userCF = UserCFPearson(data, trainData, testData)
# 根据训练集和测试集，得到预测试结果的测结果集，和测试集结果的行数一样
# 根据测试集和预测结果集，计算模型精确度
# rmse, mae = userCF.calRmseAndMae(K)
# print('rmse: %1.5f\t mae: %1.5f' % (rmse, mae))
TopN, recommend_list = userCF.recommendation(userID, N, K)
trainDict = userCF.classifyMovie(trainData)
testDict = userCF.classifyMovie(testData)
i = 1
for line in TopN:
print("top", i, ": ", line)
i += 1
# 测算recall 和 precision
# print("%5s%5s%20s%20s" % ('K', 'N', "recall", 'precision'))
# K 选取临近的用户数量
# N 输出推荐电影的数量
for k in [5, 10, 20, 40, 80, 160]:
for n in [5, 10, 15, 20]:
recall, precision = userCF.recallAndPrecision(k, n)
# print("%5d%5d%19.3f%%%19.3f%%" % (k,n,recall * 100,precision * 100))
return 0
elif method == 'userbased-euclidean':
userCF = UserCFEuclidean(data, trainData, testData)
TopN, _ = userCF.predict(userID, N, threshold)
precision = userCF.evaluation(N)
i = 1
for line in TopN:
print("top", i, ": ", line[1][1], "\t", line[1][0])
i += 1
return 0
elif method == 'itembased':
#itemBased - 余弦函数
ItemCF = ItemBasedCF(data, trainData, testData)
TopN, _, recommend_list = ItemCF.Recommendation(userID, K, N)
i = 1
for line in TopN:
print("top", i, ": ", line)
i += 1
# 测算recall 和 precision
print("%5s%5s%20s%20s" % ('K', 'N', "recall", 'precision'))
# K 选取临近的用户数量
# N 输出推荐电影的数量
for k in [5, 10, 20, 40, 80, 160]:
for n in [10]:
recall, precision = ItemCF.recallAndPrecision(k, n)
# print("%5d%5d%19.3f%%%19.3f%%" % (k,n,recall * 100,precision * 100))
return 0
elif method == 'exit':
return 0
else:
return 1
print("-------------Completed!!-----------")
return precision, recall, cover, popular
else:
recall = self.Recall(train_set, test_set, N, K)
precision = self.Precision(train_set, test_set, N, K)
cover = self.Coverage(train_set, test_set, N, K, all_item)
popular = self.Popularity(train_set, test_set, N, K,
item_popularity)
return precision, recall, cover, popular
def xxx(self):
a = self.Precision(self.trainMovieDict, self.testMovieDict, 5, 10, 1)
return a
if __name__ == "__main__":
data = DataProcess('./ml-100k/u.data')
trainData = DataProcess('./ml-100k/u1.base')
testData = DataProcess('./ml-100k/u1.test')
trainLr = trainData.merge_lrRating_movies()
trainX = trainLr[[
'UserID', 'MovieID', 'Age', 'Gender', 'Occupation', 'ZipCode'
]]
trainY = trainLr['Rating']
testLr = testData.merge_lrRating_movies()
testX = testLr[[
'UserID', 'MovieID', 'Age', 'Gender', 'Occupation', 'ZipCode'
]]
testY = testLr['Rating']