def gen_datas(train, test):"""生成数据参数:train: 原始训练数据集test: 原始验证数据集返回:X_train_r: 处理后训练数据集y_train: 处理后训练标签X_test_r: 处理后验证数据集y_test: 处理后验证标签"""label_encoder_train = LabelEncoder().fit(train.Result)labels_train = label_encoder_train.transform(train.Result)label_encoder_test = LabelEncoder().fit(test.Result)labels_test = label_encoder_test.transform(test.Result) classes = list(label_encoder_train.classes_)# classes_test = list(label_encoder_test.classes_)train = train.drop('Result', axis=1)test = test.drop('Result', axis=1)# 标签的种类 = NB_CLASSnumb_class = len(classes)# 数据的归一化处理scaled_train = data_scaler(train)# 同一文件内数据集：10%作为测试集，90%作为训练集# random_state：随机数种子，和random中的seed种子一样，保证每次抽样到的数据一样，便于调试sss = StratifiedShuffleSplit(test_size=0.1, random_state=23)for train_index, test_index in sss.split(scaled_train, labels_train):X_train, X_test = scaled_train[train_index], scaled_train[test_index]y_train, y_test = labels_train[train_index], labels_train[test_index]# 不同文件作为训练集和验证集# y_train = labels_train# X_train = data_scaler(train)# y_test = labels_test# X_test = data_scaler(test)# reshape train data # reshape 30*1# 也可以reshape 10*3X_train_r = np.zeros((len(X_train), NB_FEATURES, 1))X_train_r[:, :, 0] = X_train[:, :NB_FEATURES]# X_train_r[:, :, 1] = X_train[:, NB_PER_LAYER: NB_PER_LAYER * 2]# X_train_r[:, :, 2] = X_train[:, NB_PER_LAYER * 2:]# reshape test dataX_test_r = np.zeros((len(X_test), NB_FEATURES, 1))X_test_r[:, :, 0] = X_test[:, :NB_FEATURES]# X_test_r[:, :, 1] = X_test[:, NB_PER_LAYER: NB_PER_LAYER * 2]# X_test_r[:, :, 2] = X_test[:, NB_PER_LAYER * 2:]y_train = np_utils.to_categorical(y_train, NB_CLASS)y_test = np_utils.to_categorical(y_test, NB_CLASS)return X_train_r, y_train, X_test_r, y_test
在线做网站建站行业深圳营销型网站安徽建站呼和浩特网站建设阿里巴巴java开发手册网页制作用什么软件物流网站建设web开发湘潭网站seo重庆做网站discuz二次开发中文网站建设物联网系统开发网站建设知识北京建站公司开县网站建设系统开发平台宝应网站建设太原建站鸡西网站建设武汉企业建站公司网站制作好网站建设公司小程序开发文档巴中网站建设深圳软件开发公司网站制作有哪些免费建站网站网站建设找睿智骄阳广州app开发外包
猜你喜欢