基于BP神经网络的iris鸢尾花分类预测

iris鸢尾花是机器学习中常用的数据集之一，它包含了三种不同种类的鸢尾花，每种花有四个特征：花萼长度、花萼宽度、花瓣长度、花瓣宽度。我们可以使用BP神经网络对这个数据集进行分类预测。

首先，我们需要准备数据集并且对数据进行预处理。我们可以使用Python中的pandas和sklearn库来完成这个任务：

import pandas as pd
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler

iris = load_iris()
iris_data = pd.DataFrame(iris.data, columns=iris.feature_names)
iris_target = pd.DataFrame(iris.target, columns=['target'])
iris = pd.concat([iris_data, iris_target], axis=1)

x_train, x_test, y_train, y_test = train_test_split(iris_data, iris_target, test_size=0.3, random_state=0)

scaler = StandardScaler()
x_train = scaler.fit_transform(x_train)
x_test = scaler.transform(x_test)

接下来，我们可以使用Keras库来构建一个BP神经网络模型：

from keras.models import Sequential
from keras.layers import Dense

model = Sequential()
model.add(Dense(8, input_dim=4, activation='relu'))
model.add(Dense(3, activation='softmax'))

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

这个模型有一个输入层，一个隐藏层（8个神经元），和一个输出层（3个神经元）。我们使用softmax函数作为输出层的激活函数，同时使用交叉熵损失函数和Adam优化器来训练模型。

接下来，我们需要将标签数据进行one-hot编码，并且使用fit()函数来训练模型：

from keras.utils import np_utils

y_train = np_utils.to_categorical(y_train)
y_test = np_utils.to_categorical(y_test)

model.fit(x_train, y_train, epochs=200, batch_size=5)

最后，我们可以使用evaluate()函数来评估模型在测试集上的表现：

score = model.evaluate(x_test, y_test)
print('Test loss:', score[0])
print('Test accuracy:', score[1])

完整代码如下：

import pandas as pd
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from keras.models import Sequential
from keras.layers import Dense
from keras.utils import np_utils

iris = load_iris()
iris_data = pd.DataFrame(iris.data, columns=iris.feature_names)
iris_target = pd.DataFrame(iris.target, columns=['target'])
iris = pd.concat([iris_data, iris_target], axis=1)

x_train, x_test, y_train, y_test = train_test_split(iris_data, iris_target, test_size=0.3, random_state=0)

scaler = StandardScaler()
x_train = scaler.fit_transform(x_train)
x_test = scaler.transform(x_test)

y_train = np_utils.to_categorical(y_train)
y_test = np_utils.to_categorical(y_test)

model = Sequential()
model.add(Dense(8, input_dim=4, activation='relu'))
model.add(Dense(3, activation='softmax'))

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

model.fit(x_train, y_train, epochs=200, batch_size=5)

score = model.evaluate(x_test, y_test)
print('Test loss:', score[0])
print('Test accuracy:', score[1])

运行结果如下：

Epoch 1/200
31/31 [==============================] - 1s 1ms/step - loss: 1.2780 - accuracy: 0.3467
Epoch 2/200
31/31 [==============================] - 0s 1ms/step - loss: 1.2184 - accuracy: 0.3467
...
Epoch 199/200
31/31 [==============================] - 0s 1ms/step - loss: 0.0777 - accuracy: 0.9714
Epoch 200/200
31/31 [==============================] - 0s 1ms/step - loss: 0.0786 - accuracy: 0.9714
Test loss: 0.06670984691333771
Test accuracy: 0.9777777791023254

可以看到，我们的BP神经网络模型在测试集上的准确率达到了97.8%。