딥러닝 keras 파이썬 밸류에러 질문 ㅠㅠ

Question

csv 데이터를 pandas로 가져왔습니다.그리고 conv1d에 넣기 위해 numpy 형태로 바꾼 후 reshape 했습니다.그리고 훈련을 돌렸는데 fit에서 자꾸 return ops.EagerTensor(value, ctx.device_name, dtype)ValueError: Failed to convert a NumPy array to a Tensor (Unsupported object type float).이런 에러가 뜹니다. 그래서 3시간 구글링하면서 문자열을 float로 바꾸는 방법 쳐서# x1_train &#61; np.asarray(x1_train).astype(np.float64)# x2_train &#61; np.asarray(x2_train).astype(np.float64)# y_train &#61; np.asarray(y_train).astype(np.float64)요 코드까지도 넣어보았는데도 ValueError: could not convert string to float: '49,600' 에러가 뜹니디...코딩 시작한지 한달도 안된 왕초보임다,, 저를 도와주세요 ㅠㅠ# import numpy as np import pandas as pd import tensorflow as tf from sklearn.model_selection import train_test_split from tensorflow.python.keras.models import Model from tensorflow.python.keras.layers import Dense, Flatten, Input, Conv1D, Dropout, MaxPooling1D from tensorflow.python.keras.callbacks import EarlyStopping, ModelCheckpoint #1. 데이터 path &#61; '/Applications/study/keras/_data/samsung_jusik/' samsung_csv &#61; pd.read_csv(path &#43; 'samsung.csv', index_col&#61;0, encoding&#61;'cp949') amore_csv &#61; pd.read_csv(path &#43; 'amore.csv', index_col&#61;0, encoding&#61;'cp949') # submission &#61; pd.read_csv(path &#43; 'submission.csv', index_col&#61;0) samsung_csv &#61; samsung_csv.sort_index() # index를 오름차순으로 바꿔준다. amore_csv &#61; amore_csv.sort_index() print(amore_csv) print(samsung_csv) print(samsung_csv.shape) # [1980 rows x 16 columns] print(samsung_csv.columns) # ['시가', '고가', '저가', '종가', '전일비', 'Unnamed: 6', '등락률', '거래량', '금액(백만)', # '신용비', '개인', '기관', '외인(수량)', '외국계', '프로그램', '외인비'], # dtype&#61;'object') print(samsung_csv.info()) # 결측치 있는 것을 확인할 수 있음. # samsung_csv &#61; samsung_csv.dropna() # samsung_csv의 결측치 모두 삭제 print(amore_csv) print(amore_csv.shape) # [2220 rows x 16 columns] print(amore_csv.columns) # ['시가', '고가', '저가', '종가', '전일비', 'Unnamed: 6', '등락률', '거래량', '금액(백만)', # '신용비', '개인', '기관', '외인(수량)', '외국계', '프로그램', '외인비'] print(amore_csv.info()) # amore_csv &#61; amore_csv.dropna() # amore_csv의 결측치 모두 삭제 print(samsung_csv.info()) print(amore_csv.info()) # 결측치 삭제된 것을 확인할 수 있음 x1_datasets &#61; samsung_csv[['시가', '종가', '고가', '저가', '거래량']] print(x1_datasets.shape) # (1980, 5) print(x1_datasets) # (1980, 5) x2_datasets &#61; amore_csv[['시가', '종가', '고가', '저가', '거래량']] print(x2_datasets.shape) # (2220, 5) y &#61; samsung_csv['시가'] print(y) print(y.shape) # (1980, ) x_predict &#61; np.array([[[64400], [65000], [63900], [64600], [18154371]]]) print(x_predict.shape) # (1, 5, 1) x1_train, x1_test, y_train, y_test &#61; train_test_split( x1_datasets, y, train_size&#61;0.7, random_state&#61;1234) print(x1_train.shape, y_train.shape) # (1386, 5) (1386, ) print(x1_test.shape, y_test.shape) # (594, 5) (594, ) # 잘 train,test 잘린거 알 수 있음 x2_train, x2_test &#61; train_test_split(x2_datasets, train_size&#61;0.7, random_state&#61;1234) print(x2_train.shape) # (1554, 5) print(x2_test.shape) # (666, 5) x1_train &#61; x1_train.to_numpy() x1_test &#61; x1_test.to_numpy() x2_train &#61; x2_train.to_numpy() x2_test &#61; x2_test.to_numpy() x1_train &#61; x1_train.reshape(1386, 5, 1) x1_test &#61; x1_test.reshape(594, 5, 1) x2_train &#61; x2_train.reshape(1554, 5, 1) x2_test &#61; x2_test.reshape(666, 5, 1) # x1_train &#61; np.asarray(x1_train).astype(np.float32) # x2_train &#61; np.asarray(x2_train).astype(np.float32) # y_train &#61; np.asarray(y_train).astype(np.float32) #2. 모델구성 #2-1. 모델 1 삼성. input1&#61; Input(shape&#61;(5, 1)) dense1&#61; Conv1D(64, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds11')(input1) dense2&#61; Conv1D(64, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds12')(dense1) dense3&#61; Conv1D(128, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds13')(dense2) dense4&#61; Conv1D(128, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds14')(dense3) dense5&#61; Conv1D(256, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds16')(dense4) dense6&#61; Conv1D(256, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds17')(dense5) dense7&#61; Conv1D(256, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds18')(dense6) dense8&#61; Conv1D(512, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds20')(dense7) dense9&#61; Conv1D(128, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds24')(dense8) dense10&#61; Conv1D(128, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds25')(dense9) dense11&#61; Conv1D(128, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds26')(dense10) dense12&#61; Flatten(name&#61;'ds28')(dense11) output1&#61; Dense(32, activation&#61;'relu', name&#61;'ds29')(dense12) # concatenate 의 input # 머지 할 것이기 때문에 아직 output 내지 않음. #2-2. 모델 2. input2&#61; Input(shape&#61;(5, 1)) dense21&#61; Conv1D(64, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds31')(input2) dense22&#61; Conv1D(64, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds32')(dense21) dense23&#61; Conv1D(128, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds34')(dense22) dense24&#61; Conv1D(128, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds35')(dense23) dense25&#61; Conv1D(256, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds37')(dense24) dense26&#61; Conv1D(256, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds38')(dense25) dense27&#61; Conv1D(256, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds39')(dense26) dense28&#61; Conv1D(512, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds41')(dense27) dense29&#61; Conv1D(512, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds42')(dense28) dense30&#61; Conv1D(512, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds43')(dense29) dense31&#61; Conv1D(128, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds45')(dense30) dense32&#61; Conv1D(128, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds46')(dense31) dense33&#61; Conv1D(128, 2, padding&#61;'same', activation&#61;'relu', name&#61;'ds47')(dense32) dense34&#61; Flatten(name&#61;'ds49')(dense33) output2&#61; Dense(64, activation&#61;'relu', name&#61;'ds50')(dense34) # dense21&#61; Dense(21, activation&#61;'linear', name&#61;'ds21')(input2) # dense22&#61; Dense(22, activation&#61;'linear', name&#61;'ds22')(dense21) # output2&#61; Dense(23, activation&#61;'linear', name&#61;'ds23')(dense22) # concatenate 의 input #2-3. 모델병합 from tensorflow.python.keras.layers import concatenate # concatenate 사슬같이 잇다; 연쇄시키다; merge1 &#61; concatenate([output1, output2], name&#61;'mg1') # 2개 이상은 리스트 형태 merge2 &#61; Dense(128, activation&#61;'relu', name&#61;'mg2')(merge1) merge3 &#61; Dense(64, activation&#61;'relu', name&#61;'mg3')(merge2) merge4 &#61; Dense(32, activation&#61;'relu', name&#61;'mg4')(merge3) merge5 &#61; Dense(16, activation&#61;'relu', name&#61;'mg5')(merge4) merge6 &#61; Dense(8, activation&#61;'relu', name&#61;'mg6')(merge5) last_output &#61; Dense(1, name&#61;'last')(merge6) model &#61; Model(inputs&#61;[input1, input2], outputs&#61;last_output) # 모델이 두개라서 인풋이 2개임. model.summary() #3. 컴파일, 훈련 model.compile(loss&#61;'mse', optimizer&#61;'adam', metrics&#61;['acc']) es &#61; EarlyStopping(monitor&#61;'val_loss', mode&#61;'min', restore_best_weights&#61;True, patience&#61;20, verbose&#61;1) import datetime date &#61; datetime.datetime.now() date &#61; date.strftime('%M%D_%H%M') filepath &#61; '/Applications/study/keras/_save/MCP/' filename &#61; '{epoch:04d}-{val_loss:.4f}.hdf5' mcp &#61; ModelCheckpoint(monitor&#61;'val_loss', mode&#61;'min', save_best_only&#61;True, verbose&#61;1, filepath&#61; filepath &#43; 'k34_1_' &#43; date &#43; '_' &#43; filename) # x1_train &#61; np.asarray(x1_train).astype(np.float64) # x2_train &#61; np.asarray(x2_train).astype(np.float64) # y_train &#61; np.asarray(y_train).astype(np.float64) model.fit([x1_train, x2_train], y_train, epochs&#61;200, batch_size&#61;8) #4. 평가, 예측 loss &#61; model.evaluate([x1_test, x2_test], y_test) print('loss:', loss) result &#61; model.predict(x_predict) print('1/30 삼성전자 시가 예측값:', result)

딴짓약사 · Answer

문자 데이터 값에 comma(,)가 들어있는 데이터가 있어서 오류가 난 거에요!

딥러닝 keras 파이썬 밸류에러 질문 ㅠㅠ

1개의 답변

#keras 목록