四种物料类型(宽料, 窄料, 厚料, 小卷料) 四种冷却方式(自然冷却, 1号风机, 2号风机, 1号和2号风机) 物料温度随时间而下降受到分类变量(物料类型, 冷却方式)的影响拟合规律预测不同-CFANZ编程社区

代码

https://blog.51cto.com/u_16055028/7673781

import numpy as np
from sklearn.linear_model import LinearRegression
from sklearn.preprocessing import OneHotEncoder
from sklearn.ensemble import RandomForestRegressor
from sklearn.tree import DecisionTreeRegressor
from sklearn.svm import SVR
from sklearn.neural_network import MLPRegressor

In [2]:

# 生成模拟数据
np.random.seed(0)
n_samples = 1000
n_features = 3  # 物料类型、冷却方式、时间

In [3]:

# 随机生成物料类型和冷却方式
materials = np.random.choice(['宽料', '窄料', '厚料', '小卷料'], size=n_samples)
cooling_methods = np.random.choice(['自然冷却', '1号风机', '2号风机', '1号和2号风机'], size=n_samples)

# 随机生成时间数据
time = np.random.uniform(0, 10, size=n_samples)

# 生成随机物料温度数据（这里只是示例，实际应用中需要根据模型生成）
temperature = 20 + 5 * time + np.random.normal(0, 2, size=n_samples)

In [26]:

# 将分类变量编码为数值
encoder = OneHotEncoder(sparse_output=False)
materials_encoded = encoder.fit_transform(materials.reshape(-1, 1))
cooling_methods_encoded = encoder.fit_transform(cooling_methods.reshape(-1, 1))

In [27]:

# 合并特征
X = np.column_stack((materials_encoded, cooling_methods_encoded, time))

In [6]:

# 拟合数据和解释模型（六种方法）
# 1. 线性回归
linear_model = LinearRegression()
linear_model.fit(X, temperature)
print("Linear Regression Coefficients:")
print(linear_model.coef_)
print("Linear Regression Intercept:", linear_model.intercept_)

# 2. 多项式回归
from sklearn.preprocessing import PolynomialFeatures
poly_features = PolynomialFeatures(degree=2)
X_poly = poly_features.fit_transform(X)
poly_model = LinearRegression()
poly_model.fit(X_poly, temperature)
print("Polynomial Regression Coefficients:")
print(poly_model.coef_)
print("Polynomial Regression Intercept:", poly_model.intercept_)

# 3. 决策树回归
tree_model = DecisionTreeRegressor()
tree_model.fit(X, temperature)
print("Decision Tree Feature Importances:")
print(tree_model.feature_importances_)

# 4. 随机森林回归
rf_model = RandomForestRegressor()
rf_model.fit(X, temperature)
print("Random Forest Feature Importances:")
print(rf_model.feature_importances_)

# 5. 支持向量机回归
svm_model = SVR()
svm_model.fit(X, temperature)
print("Support Vector Machine Coefficients:")
# print(svm_model.coef_)  # 错误 svm_model.coef_ AttributeError: coef_ is only available when using a linear kernel svm_model.fit(X, temperature)
print(svm_model.support_vectors_)
print(svm_model.dual_coef_)

# 6. 神经网络
nn_model = MLPRegressor(hidden_layer_sizes=(10, 10), max_iter=1000)
nn_model.fit(X, temperature)
# 神经网络没有直接解释性参数，通常使用模型的预测能力
# 这里只是示例代码，实际应用中可能需要更复杂的网络结构

Linear Regression Coefficients:
[ 0.09738255  0.09813695 -0.09434166 -0.10117784 -0.04251578  0.01205836
 -0.02354086  0.05399828  4.97497016]
Linear Regression Intercept: 20.07918549347334
Polynomial Regression Coefficients:
[-3.55663651e+07 -2.46440447e+11 -2.70021102e+11 -7.88059614e+10
 -8.11398359e+10 -2.91102428e+10 -1.20136820e+10  1.13551818e+10
 -1.73739022e+10  7.09367018e+10  5.24750931e+10  2.68453159e+07
  1.29473196e+07 -2.17156558e+07  2.86211077e+10  1.16705735e+10
  8.76292665e+09  1.67218805e+10 -1.05961698e+10  7.10109434e+10
  7.38255522e+00 -4.82728086e-01  3.36659124e+10  1.67153782e+10
  1.38077314e+10  2.17666852e+10 -1.05961698e+10 -8.16416041e+10
  0.00000000e+00 -4.89668096e+09 -2.18472152e+10 -2.47548620e+10
 -1.67959082e+10 -1.05961698e+10 -7.58178180e+10 -8.38659259e+09
 -2.53371268e+10 -2.82447736e+10 -2.02858198e+10 -1.05961698e+10
 -8.29216045e+09  0.00000000e+00  0.00000000e+00  0.00000000e+00
 -6.03405320e+10 -8.43818705e+09  0.00000000e+00  0.00000000e+00
 -6.03405320e+10 -2.88994040e+10  0.00000000e+00 -6.03405320e+10
 -8.12927384e+09 -6.03405320e+10 -2.38037109e-03]
Polynomial Regression Intercept: 202782216111.90613
Decision Tree Feature Importances:
[4.81361663e-04 7.29456493e-04 1.01830280e-03 7.48103052e-04
 6.76530127e-04 5.90113556e-04 6.57851309e-04 7.32742416e-04
 9.94365539e-01]
Random Forest Feature Importances:
[6.87663836e-04 7.56478794e-04 8.67089363e-04 8.31802663e-04
 7.98470513e-04 7.20358675e-04 5.81217235e-04 7.52520218e-04
 9.94004399e-01]

Out[6]:

MLPRegressor

MLPRegressor(hidden_layer_sizes=(10, 10), max_iter=1000)

In [44]:

# 应用模型进行预测
# 例如，给定新的物料类型、冷却方式和时间，可以使用训练好的模型进行温度预测
# materials = np.random.choice(['宽料', '窄料', '厚料', '小卷料'], size=n_samples)
# cooling_methods = np.random.choice(['自然冷却', '1号风机', '2号风机', '1号和2号风机'], size=n_samples)
new_material = np.array([[0, 1, 0, 0]])  # 如何根据指定元素在分类列表中的位置生成对应编码 例如 指定4种物料类型里选择宽料
new_cooling_method = np.array([[0, 0, 0, 1]])
new_time = np.array([5])

# 将分类变量编码为数值
# new_material_encoded = encoder.transform(np.array([[new_material]]))
# new_cooling_method_encoded = encoder.transform(np.array([[new_cooling_method]]))
# transform 应为fit_transform或fit().transfomr()
# new_material_encoded = encoder.fit_transform(np.array([[new_material]]))
# new_cooling_method_encoded = encoder.fit_transform(np.array([[new_cooling_method]]))
# new_material_encoded = encoder.fit_transform(new_material.reshape(-1, 1))
# new_cooling_method_encoded = encoder.fit_transform(new_cooling_method.reshape(-1, 1))  # 已经手动转码过了

# new_input = np.column_stack((new_material_encoded, new_cooling_method_encoded, new_time))
# ValueError: X has 3 features, but * is expecting 9 features as input.
print(materials_encoded.shape, cooling_methods_encoded.shape, time.shape, X.shape)
new_input = np.column_stack((new_material, new_cooling_method, new_time))
print(new_material.shape, new_cooling_method.shape, new_time.shape, new_input.shape)

(1000, 4) (1000, 4) (1000,) (1000, 9)
(1, 4) (1, 4) (1,) (1, 9)

In [45]:

# 使用各个模型进行温度预测
linear_prediction = linear_model.predict(new_input)
poly_prediction = poly_model.predict(poly_features.transform(new_input))
tree_prediction = tree_model.predict(new_input)
rf_prediction = rf_model.predict(new_input)
svm_prediction = svm_model.predict(new_input)
nn_prediction = nn_model.predict(new_input)

# 输出预测结果
print("Linear Regression Prediction:", linear_prediction)
print("Polynomial Regression Prediction:", poly_prediction)
print("Decision Tree Regression Prediction:", tree_prediction)
print("Random Forest Regression Prediction:", rf_prediction)
print("Support Vector Machine Regression Prediction:", svm_prediction)
print("Neural Network Regression Prediction:", nn_prediction)

Linear Regression Prediction: [45.10617151]
Polynomial Regression Prediction: [45.1194458]
Decision Tree Regression Prediction: [43.32600977]
Random Forest Regression Prediction: [42.97839963]
Support Vector Machine Regression Prediction: [45.25602304]
Neural Network Regression Prediction: [45.09695762]

In [ ]: