试图为疾病制作分类器，但它返回错误的输出

Question

I am trying to make an app which can get the patients symptons as inputs and outputs the three most likely diseases.

x_train, x_test, y_train, y_test = sklearn.model_selection.train_test_split(x, y, test_size=0.4)

inputs = keras.Input(shape=(9,))
hidden_1 = keras.layers.Dense(12, activation='selu')(inputs)
hidden_2 = keras.layers.Dense(12, activation='relu')(hidden_1)
outputs = keras.layers.Dense(32, activation='sigmoid')(hidden_2)

model = keras.Model(inputs=inputs, outputs=outputs)

model.compile(optimizer='adam',
              loss='categorical_crossentropy',
              metrics=['categorical_accuracy'])

model.summary()

model.fit(x_train, y_train, epochs=200)  # starts training

prediction = model.predict(x_test)
print(prediction)
scores = model.evaluate(x_test, y_test, verbose=0)
print(scores)

print(prediction[0])
print(y_test[0])

model.save("modeldisease.h5")

but when i save and load this model into another python file, the outputs list looks like this:

  [[0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
  1.000000e+00 0.000000e+00 0.000000e+00 1.842211e-21 0.000000e+00
  0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00
  1.000000e+00 1.000000e+00 1.000000e+00 1.000000e+00 0.000000e+00
  0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00
  0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
  1.000000e+00 0.000000e+00]]
100.0 100.0 100.0
HEPATITE A HEPATITE_ALCOOLICA REFLUXO_GASTROESOFAGICO

Process finished with exit code 0

What am i doing wrong?, shouldnt it just return me this prediction list with all the values adding up to 1? dataset link: https://drive.google.com/file/d/1LFCtuBKodGkm1NNdv3QdXOwzlSqdip5q/view?usp=sharing

full code: `

from tensorflow import keras
import numpy as np
import sklearn
from sklearn import preprocessing
import pandas as pd

data = pd.read_csv("DATA.csv", sep=";")

obj_data = data.select_dtypes(include=["object"]).copy()
obj_data_names = []

for col in obj_data.columns:
    obj_data_names.append(col)

for col_name in obj_data_names:
    data[col_name] = data[col_name].astype('category')
    data[col_name + "_cat"] = data[col_name].cat.codes
    data[col_name] = data[col_name + "_cat"]
    data.drop(col_name + "_cat", 1, inplace=True)
print(data)
data.fillna(0, inplace=True)
x = np.array(data.drop(["Doenca"], 1))
y = np.array(data["Doenca"])

x = preprocessing.normalize(x)
y = keras.utils.to_categorical(y)

x_train, x_test, y_train, y_test = sklearn.model_selection.train_test_split(x, y, test_size=0.4)

inputs = keras.Input(shape=(9,))
hidden_1 = keras.layers.Dense(12, activation='selu')(inputs)
hidden_2 = keras.layers.Dense(12, activation='relu')(hidden_1)
outputs = keras.layers.Dense(32, activation='sigmoid')(hidden_2)

model = keras.Model(inputs=inputs, outputs=outputs)

model.compile(optimizer='adam',
              loss='categorical_crossentropy',
              metrics=['categorical_accuracy'])

model.summary()

model.fit(x_train, y_train, epochs=200)  # starts training

prediction = model.predict(x_test)
print(prediction)
scores = model.evaluate(x_test, y_test, verbose=0)
print(scores)

print(prediction[0])
print(y_test[0])

model.save("modeldisease.h5")

for i in range(len(y_test)):
    max_expected = np.amax(y_test[i])
    #print(max_expected)
    y_expected = np.where(y_test[i] == max_expected)
    print(y_expected[0])
    max_predicted = np.amax(prediction[i])
    #print(max_predicted)
    y_predicted = np.where(prediction[i] == max_predicted)
    print(y_predicted[0])

    classes = ["ACNE", "AIDS", "ALERGIA", "ARTRITE", "ARTROSE", "CATAPORA", "COLESTASE", "DENGUE", "DIABETES", "ENXAQUECA", "ESPONDILOSE",
               "FEBRE_TIFOIDE", "GASTROENTERITE", "GRIPE", "HEPATITE_ALCOOLICA", "HEPATITE A", "HEPATITE B", "HEPATITE C", '"HEPATITE D',
               "HEPATITE E", "HIPERTENSAO", "HIPERTENSAO", "HIPERTIROIDISMO", "HIPOGLICEMIA", "HIPOTIREODISMO", "IMPETIGO", "INFECCA_URINA",
               "MALARIA", "PNEUMONIA", "PSORIASE", "REFLUXO_GASTROESOFAGICO", 'TUBERCULOSE', "ULCERA GASTRICA"]
    print(f"Valor esperado: {classes[int(y_expected[0])]}, Valor previsto: {classes[int(y_predicted[0])]}")

# print(prediction[0][0])
# print(round(prediction[0][0]))
# print(classes[round(int(prediction[0][0]))])
#print(y_test)
#print(prediction)

load code:

from tensorflow import keras

model = keras.models.load_model("modeldisease.h5")

result = model.predict([(12, 40, 39, 17, 0, 0, 0, 0, 0)])
print(result)

sortedshit = result[0].argsort()[-3:][::-1]
p1 = result[0][sortedshit[0]]
p2 = result[0][sortedshit[1]]
p3 = result[0][sortedshit[2]]

classes = ["ACNE", "AIDS", "ALERGIA", "ARTRITE", "ARTROSE", "CATAPORA", "COLESTASE", "DENGUE", "DIABETES", "ENXAQUECA",
           "ESPONDILOSE",
           "FEBRE_TIFOIDE", "GASTROENTERITE", "GRIPE", "HEPATITE_ALCOOLICA", "HEPATITE A", "HEPATITE B", "HEPATITE C",
           '"HEPATITE D',
           "HEPATITE E", "HIPERTENSAO", "HIPERTENSAO", "HIPERTIROIDISMO", "HIPOGLICEMIA", "HIPOTIREODISMO", "IMPETIGO",
           "INFECCA_URINA",
           "MALARIA", "PNEUMONIA", "PSORIASE", "REFLUXO_GASTROESOFAGICO", 'TUBERCULOSE', "ULCERA GASTRICA"]

print(p1*100, p2*100, p3*100)

print(classes[sortedshit[0]], classes[sortedshit[1]], classes[sortedshit[2]])

Answer 1

应用 softmax 激活函数而不是 sigmoid：

outputs = keras.Dense(32,activation='softmax')(hidden_2)