def nmf_run():
"""
非负矩阵分解 run
:return:
"""
# 1、导入用户商品矩阵
print("----------- 1、load data -----------")
data_matrix = FTool.LoadData(file_name="data.txt").load_data_with_none()
# 2、利用梯度下降法对矩阵进行分解
print("----------- 2、training -----------")
w_matrix, h_matrix = train(data_matrix, 5, 10000, 1e-5)
# 3、保存分解后的结果
print("----------- 3、save decompose -----------")
with FTool.SaveModel(file_name="w_matrix") as save_file:
save_file.save_model_mul(w_matrix)
with FTool.SaveModel(file_name="h_matrix") as save_file:
save_file.save_model_mul(h_matrix)
# 4、预测
print("----------- 4、prediction -----------")
predict = prediction(data_matrix, w_matrix, h_matrix, 0)
# 进行Top-K推荐
print("----------- 5、top_k recommendation ------------")
top_recommend = top_k(predict, 2)
print(top_recommend)
print(w_matrix * h_matrix)
def recommend_run():
"""
系统过滤算法运行(包括基于用户和基于项的协同推荐)
:return:
"""
# 1、导入用户商品数据
print(
"=======================User-based Recommend========================")
print("------------ 1. load data ------------")
data = FTool.LoadData(file_name="data.txt").load_data_with_none()
# 2、计算用户之间的相似性
print("------------ 2. calculate similarity between users -------------")
w = similarity(data)
# 3、利用用户之间的相似性进行推荐
print("------------ 3. predict ------------")
predict = user_based_recommend(data, w, 2)
# 4、进行Top-K推荐
print("------------ 4. top_k recommendation ------------")
top_recommend = top_k(predict, 2)
print(top_recommend)
print(
"\n\n=======================User-based Recommend========================"
)
data = data.T # 将用户商品矩阵转换为商品用户矩阵
w = similarity(data)
predict = item_based_recommend(data, w, 2)
# 4、进行Top-K推荐
print("------------ 4. top_k recommendation ------------")
top_recommend = top_k(predict, 2)
print(top_recommend)
def mf_run():
"""
Matrix Factorization-based Recommend run
:return:
"""
# 1、导入用户商品矩阵
print("----------- 1、load data -----------")
data_matrix = FTool.LoadData(file_name="data.txt").load_data_with_none()
# 2、利用梯度下降法对矩阵进行分解
print("----------- 2、training -----------")
p, q = grad_ascent(data_matrix, 3, 0.0002, 0.02,
5000) # 由于样本过少, 迭代次数增加min loss也不会继续收敛
# 3、保存分解后的结果
print("----------- 3、save decompose -----------")
with FTool.SaveModel(file_name="p_result") as save_file:
save_file.save_model_mul(p)
with FTool.SaveModel(file_name="q_result") as save_file:
save_file.save_model_mul(q)
# 4、预测
print("----------- 4、prediction -----------")
predict = prediction(data_matrix, p, q, 0)
# 进行Top-K推荐
print("----------- 5、top_k recommendation ------------")
top_recommend = top_k(predict, 2)
print(top_recommend)
print(p * q)
def k_means_pp():
"""
K-Means ++ 实现
:return:
"""
k = 4 # 聚类中心的个数
file_path = "data.txt"
# 1、导入数据
print("---------- 1.load data ------------")
data, _, _ = FTool.LoadData(file_name=file_path).load_data(
feature_end=0, need_label_length=True, need_list=True)
x_data = [_data[0] for _data in data]
y_data = [_data[1] for _data in data]
with FTool.PaintingWithList(name="K-Means") as paint:
paint.painting_simple_list(x_data, y_data)
# 2、KMeans++的聚类中心初始化方法
print("---------- 2.K-Means++ generate centers ------------")
data = np.mat(data)
centroids = get_centroids(data, k)
# 3、聚类计算
print("---------- 3.k-Means ------------")
sub_center = k_means_function(data, k, centroids)
# 4、保存所属的类别文件
print("---------- 4.save subCenter ------------")
with FTool.SaveModel(file_name="sub_pp") as save_model:
save_model.save_model_mul(sub_center)
# 5、保存聚类中心
print("---------- 5.save centroids ------------")
with FTool.SaveModel(file_name="center_pp") as save_model:
save_model.save_model_mul(centroids)
def k_means(k_class: int):
"""
The K-Means Function
:param k_class: 聚类中心的个数
:return:
"""
k = k_class # 聚类中心的个数
file_path = "data.txt"
# 1、导入数据
print("---------- 1.load data ------------")
data, _, _ = FTool.LoadData(file_name=file_path).load_data(feature_end=0, need_label_length=True, need_list=True)
x_data = [_data[0] for _data in data]
y_data = [_data[1] for _data in data]
with FTool.PaintingWithList(name="K-Means") as paint:
paint.painting_simple_list(x_data, y_data)
# 2、随机初始化k个聚类中心
print("---------- 2.random center ------------")
data = np.mat(data)
centroids = rand_cent(data, k)
# 3、聚类计算
print("---------- 3.kmeans ------------")
sub_center = k_means_function(data, k, centroids)
# 4、保存所属的类别文件
print("---------- 4.save subCenter ------------")
with FTool.SaveModel(file_name="sub") as save_model:
save_model.save_model_mul(sub_center)
# 5、保存聚类中心
print("---------- 5.save centroids ------------")
with FTool.SaveModel(file_name="center") as save_model:
save_model.save_model_mul(centroids)
def personal_rank_run():
"""
Personal Rank run
:return:
"""
# 1、导入用户商品矩阵
print("------------ 1.load data -------------")
data_mat = FTool.LoadData(file_name="data.txt").load_data_with_none()
# 2、将用户商品矩阵转换成邻接表的存储(二部图)
print("------------ 2.generate dict --------------")
data_dict = generate_dict(data_mat)
# 3、利用PersonalRank计算
print("------------ 3.PersonalRank --------------")
rank = personal_rank(data_dict, 0.85, "U_0", 500)
# 4、根据rank结果进行商品推荐
print("------------ 4.recommend -------------")
result = recommend(data_dict, rank, "U_0")
print(result)
def TestOfLR():
"""
测试逻辑回归算法
:return:
"""
lr_train.TrainOfLR()
print("------------------------1. Load Model-----------------------")
with FTool.LoadModel("weights") as model:
w = model.load_model_mul()
n = np.shape(w)[1]
print("------------------------2. Load Data-----------------------")
test_data = FTool.LoadData(file_name="test_data", feature_type="float").load_data_with_limit(number=n, offset=1)
print("------------------------3. Get Prediction-----------------------")
h = predict(test_data, w)
# TODO painting
print("------------------------4. Save Prediction-----------------------")
save_result("result", h)
def cart_train():
"""
cart_train training
:return:
"""
# 1、导入训练数据
print("----------- 1、load data -------------")
data, label, _ = FTool.LoadData(file_name="sine.txt").load_data(
need_label_length=True, need_list=True, feature_end=0)
feature = [_data[0] for _data in data]
with FTool.PaintingWithList(name="cart_train Train") as paint:
paint.painting_simple_list(feature, label)
# 2、构建CART树
print("----------- 2、build cart_train ------------")
regression_tree = build_tree(data, 30, 0.3)
# 3、评估CART树
print("----------- 3、cal err -------------")
err = cal_error(data, regression_tree)
print("\t--------- err : ", err)
# 4、保存最终的CART模型
print("----------- 4、save result -----------")
save_model(regression_tree, "regression_tree")