机器学习复习代码
利用sklearn实现knn
import numpy as np
import pandas as pd
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import GridSearchCVdef model_selection(x_train, y_train):## 第一个是网格搜索## p是选择查找方式:1是欧式距离 2是曼哈顿距离params = {'n_neighbors': [3,5,7], 'p': [1,2]}model = KNeighborsClassifier()gs = GridSearchCV(model, params, verbose=2, cv=5)gs.fit(x_train, y_train)print("Best Model:", gs.best_params_, "Accuracy:", gs.best_score_)print(gs.best_estimator_)return gs.best_estimator_def read():filename = r"data/shuixianhua.xlsx"data = pd.read_excel(filename, header=None)## iloc[行,列]x1 = data.iloc[1:, [0, 1]].valuesx2 = data.iloc[1:, [3, 4]].values# print(x2)y1 = data.iloc[1:, 2].valuesy2 = data.iloc[1:, 5].valuesx = np.vstack((x1, x2)) # 竖向合并print("x:")print(x)y = np.hstack((y1, y2)) # 横向合并print("y:")print(y)## 这里是因为我把excel的y理解成string类型了,如果正常读可以不加这个## 将y转为数值的inty = y.astype(int)return x, yif __name__ == '__main__':x, y = read()best_model = model_selection(x, y)
利用sklearn实现线性回归
数据集展示
import pandas as pd
from sklearn.linear_model import LinearRegression
import numpy as np
def MAE(y,y_pre):return np.mean(np.abs(y-y_pre))
def MSE(y,y_pred):return np.mean((y-y_pred)**2)
def RMSE(y,y_pred):return np.sqrt(MSE(y,y_pred))
def MAPE(y,y_pred):return np.mean(np.abs(y-y_pred)/y)
def R2(y,y_pred):u=np.sum((y-y_pred)**2)v=np.sum((y-np.mean(y_pred))**2)return 1-(u/v)
def judege(name,y,y_pre):mae=MAE(y,y_pre)mse=MSE(y,y_pre)rmse=RMSE(y,y_pre)mape=MAPE(y,y_pre)r2=R2(y,y_pre)print(f"{name}的MAE:{mae},MSE:{mse},RMSE:{rmse}.MAPE:{mape},R2:{r2}")def read():filename = r"../data/ComposePlot.xlsx"data=pd.read_excel(filename,header=None)x1 = data.iloc[2:, [0,]].valuesy1 = data.iloc[2:,1].valuesx2 = data.iloc[2:,[2,]].valuesy2 = data.iloc[2:,3].valuesx3 = data.iloc[2:,[4,]].valuesy3 = data.iloc[2:,5].valuesx4 = data.iloc[2:,[6,]].valuesy4 = data.iloc[2:,7].valuesreturn x1,y1,x2,y2,x3,y3,x4,y4def getModel(x,y):model = LinearRegression()model.fit(x,y)return modeldef main(x1, y1, x2, y2, x3, y3, x4, y4):model1 = getModel(x1,y1)model2 = getModel(x2, y2)model3 =getModel(x3,y3)model4 =getModel(x4,y4)judege("mode1",y1,model1.predict(x1))judege("mode2",y2,model2.predict(x2))judege("mode3",y3,model3.predict(x3))judege("mode4",y4,model4.predict(x4))if __name__ == '__main__':x1, y1, x2, y2, x3, y3, x4, y4 = read()main(x1, y1, x2, y2, x3, y3, x4, y4)
利用sklearn实现逻辑回归
数据集展示
import pandas as pd
import numpy as np
from sklearn.linear_model import LogisticRegressiondef main(x,y):model=LogisticRegression()model.fit(x,y)print(model.predict(x))
def read():filename = "data/student.xlsx"data=pd.read_excel(filename,header=None)x=data.iloc[1:,[0,1]].valuesy=data.iloc[1:,2].valuesprint(x)print(y)return x,y
if __name__ =='__main__':x,y=read()main(x,y)
利用sklearn实现SVM(向量机)
from sklearn.svm import SVC
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.model_selection import GridSearchCV
import numpy as np
from sklearn.metrics import confusion_matrix, accuracy_score, precision_score, recall_score, \f1_scoredef load_data(): #导入的尾花data = load_iris()x, y = data.data, data.targetx_train, x_test, y_train, y_test = \train_test_split(x, y, test_size=0.3,shuffle=True, random_state=20)return x, y, x_train, x_test, y_train, y_test## 无脑写这个就行
def model_selection(x_train, y_train):model = SVC()paras = {'C': np.arange(1, 10, 5),# rbf:高斯核函数 linear:线性核函数 poly:多项式核函数'kernel': ['rbf', 'linear', 'poly'],'degree': np.arange(1, 10, 2),'gamma': ['scale', 'auto'],'coef0': np.arange(-10, 10, 5)}gs = GridSearchCV(model, paras, cv=3, verbose=2, n_jobs=3)gs.fit(x_train, y_train)print('best score:', gs.best_score_)print('best parameters:', gs.best_params_)return gs.best_params_def train(x_train, x_test, y_train, y_test, C, gamma, kernel):model = SVC(C=C, kernel=kernel, gamma=gamma)model.fit(x_train, y_train)y_pred = model.predict(x_test)# 生成混淆矩阵confusion = confusion_matrix(y_test, y_pred)print("混淆矩阵:\n", confusion)# 生成分类报告,包括准确率、精度、召回率和 F1 得分accuracy = accuracy_score(y_test, y_pred)precision = precision_score(y_test, y_pred, average='weighted')recall = recall_score(y_test, y_pred, average='weighted')f1 = f1_score(y_test, y_pred, average='weighted')print("准确率Accuracy:", accuracy)print("精度Precision:", precision)print("召回率Recall:", recall)print("FScore:", f1)if __name__ == '__main__':X, Y, x_train, x_test, y_train, y_test = load_data()best_params = model_selection(x_train, y_train)## 这步骤也是无脑写就行train(x_train, x_test, y_train, y_test, best_params["C"], best_params["gamma"], best_params["kernel"])
小知识点扩展
将数据分成训练集和测试集
x_train, x_test, y_train, y_test = \train_test_split(x, y, test_size=0.3,shuffle=True, random_state=20)
来自chargpt的解释
是的,这段代码用于将数据集分成训练集和测试集。具体地说,它使用
train_test_split
函数将输入的特征数据x
和标签数据y
分成四个部分:x_train
(训练集的特征数据)、x_test
(测试集的特征数据)、y_train
(训练集的标签数据)和y_test
(测试集的标签数据)。以下是这段代码的主要参数和作用:
x
是特征数据。y
是标签数据。test_size=0.3
指定了测试集的比例,这里是总数据的30%将被用作测试集。shuffle=True
表示在分割数据之前对数据进行洗牌,以确保数据的随机性。random_state=20
是一个随机种子,用于确保每次运行代码时分割数据的结果都相同,这有助于复现实验结果。
综上所诉,只要背就好了,还有参数的意思
归一化
def hypo(x,w,b):z=np.matmul(x,w)+bh_x=1/(1+np.exp(-z))h_x=(h_x>=0.5)*1return h_x
书上p49,我也不太懂归一化的用法,其中z=wx+b
从0实现线性回归
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt# 添加归一化函数
def normalize_data(data):min_val = np.min(data)max_val = np.max(data)normalized_data = (data - min_val) / (max_val - min_val)return normalized_datadef prediction(X, W, bias):return np.matmul(X, W) + biasdef cost_function(X, y, W, bias):m, n = X.shapey_hat = prediction(X, W, bias)return 0.5 * (1 / m) * np.sum((y - y_hat) ** 2)def gradient_descent(X, y, W, bias, alpha):m, n = X.shapey_hat = prediction(X, W, bias)grad_w = -(1 / m) * np.matmul(X.T, (y - y_hat))grad_b = -(1 / m) * np.sum(y - y_hat)W = W - alpha * grad_wbias = bias - alpha * grad_breturn W, biasdef train(X, y, ite=200):m, n = X.shapeW, b, alpha, costs = np.random.randn(n, 1), 0.1, 0.2, []for i in range(ite):costs.append(cost_function(X, y, W, b))W, b = gradient_descent(X, y, W, b, alpha)return costsdef read():filename = r"../../data/easy_test.xlsx"data = pd.read_excel(filename, header=None)x = data.iloc[2:, [0, ]].valuesy = data.iloc[2:, 1].values# 对特征数据 x 进行归一化x_normalized = normalize_data(x)return x_normalized, yif __name__ == '__main__':x, y = read()costs = train(x, y)# print(costs)# 绘制损失曲线plt.figure()plt.plot(range(len(costs)), costs, marker='o', linestyle='-', color='b', label='Training Loss')plt.xlabel('Iteration')plt.ylabel('Cost')plt.title('Training Loss')plt.legend()plt.grid(True)plt.show()