基于Python语言的PSO-LSTM实现分类

基于Python语言的PSO-LSTM实现分类，源数据如图所示：

在这里插入代码片
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense, Input
from pyswarm import pso
import matplotlib.pyplot as plt

# 读取数据
data = pd.read_excel('1.xlsx', header=None)

# 假设数据的第1到第6列是特征，第7列是标签
X = data.iloc[:, 1:6].values
y = data.iloc[:, 6].values

# 数据归一化
scaler = StandardScaler()
X = scaler.fit_transform(X)

# 分割训练集和测试集
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# 将输入数据形状调整为 (samples, time_steps, features)
X_train = np.expand_dims(X_train, axis=2)
X_test = np.expand_dims(X_test, axis=2)


# 定义LSTM模型
def create_lstm_model(input_shape, units, dropout):
    model = Sequential()
    model.add(Input(shape=input_shape))
    model.add(LSTM(units=units, dropout=dropout))
    model.add(Dense(5, activation='softmax'))  # 假设有4个类别
    model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
    return model


# 适应度函数
def fitness_function(params):
    units, dropout = params
    units = int(units)
    dropout = float(dropout)

    model = create_lstm_model((X_train.shape[1], X_train.shape[2]), units, dropout)
    model.fit(X_train, y_train, epochs=50, batch_size=5, verbose=0)

    loss, accuracy = model.evaluate(X_test, y_test, verbose=0)
    return 1 - accuracy


# PSO参数
lb = [10, 0.0]  # 最低单元数和最小dropout
ub = [100, 0.5]  # 最高单元数和最大dropout

# 执行PSO
best_params, best_fitness = pso(fitness_function, lb, ub, swarmsize=10, maxiter=10)

# 最优参数
best_units = int(best_params[0])
best_dropout = best_params[1]

# 使用最优参数训练最终模型
final_model = create_lstm_model((X_train.shape[1], X_train.shape[2]), best_units, best_dropout)
final_model.fit(X_train, y_train, epochs=50, batch_size=5, verbose=1)

# 评估最终模型
loss, accuracy = final_model.evaluate(X_test, y_test)
print(f'Final Model Accuracy: {accuracy * 100:.2f}%')

# 生成预测值
y_pred = final_model.predict(X_test)
y_pred_classes = np.argmax(y_pred, axis=1)

# 绘制实际值和预测值对比图
plt.figure(figsize=(14, 7))
plt.plot(y_test, label='Actual Values')
plt.plot(y_pred_classes, label='Predicted Values', linestyle='dashed')
plt.title('Actual vs Predicted Values')
plt.xlabel('Sample Index')
plt.ylabel('Class')
plt.legend()
plt.show()

计算结果如下图所示，没有对图片进行美化。