1、鸢尾花预测
1、特征选择
2、对特征处理
train=pd.read_excel("鸢尾花训练数据.xlsx")
test=pd.read_excel("鸢尾花测试数据.xlsx")
x_train=train[["萼片长(cm)","萼片宽(cm)","花瓣长(cm)","花瓣宽(cm)"]]
y_train=train[["类型_num"]].values.ravel()
3、选择逻辑回归作为分类模型。
from sklearn.linear_model import LogisticRegression
lr=LogisticRegression()
4、模型训练
lr.fit(x_train,y_train)
train_predicted=lr.predict(x_train)
5、评估模型的性能。常用的评估指标包括准确率、召回率、精确度、F1分数
from sklearn import metrics
print(metrics.classification_report(y_train,train_predicted))
完整代码
import pandas as pd
from sklearn.linear_model import LogisticRegression
train=pd.read_excel("鸢尾花训练数据.xlsx")
test=pd.read_excel("鸢尾花测试数据.xlsx")
x_train=train[["萼片长(cm)","萼片宽(cm)","花瓣长(cm)","花瓣宽(cm)"]]
y_train=train[["类型_num"]].values.ravel()lr=LogisticRegression()
lr.fit(x_train,y_train)
train_predicted=lr.predict(x_train)
from sklearn import metrics
print(metrics.classification_report(y_train,train_predicted))x_test=test[["萼片长(cm)","萼片宽(cm)","花瓣长(cm)","花瓣宽(cm)"]]
y_test=test[["类型_num"]].values.ravel()
test_predicted=lr.predict(x_test)
print(metrics.classification_report(y_test,test_predicted))predict=pd.read_excel("鸢尾花预测数据.xlsx")
x_predict=predict[["萼片长(cm)","萼片宽(cm)","花瓣长(cm)","花瓣宽(cm)"]]
predict_predicted=lr.predict(x_predict)
print(predict_predicted)
print(lr.predict_proba(x_predict))
2、银行数据
-
导入必要的库:
numpy
和pandas
用于数据处理。LogisticRegression
用于创建逻辑回归模型。StandardScaler
用于特征缩放。train_test_split
用于分割数据集。cross_val_score
用于交叉验证。metrics
用于评估模型性能。
-
读取数据:
使用pandas
的read_csv
函数读取 "creditcard.csv" 文件。 -
特征缩放:
使用StandardScaler
对 "Amount" 特征进行标准化。data["Amount"]=scaler.fit_transform(data[["Amount"]]) data=data.drop(["Time"],axis=1)
-
删除不需要的列:
删除 "Time" 列,因为它对于分类问题可能没有帮助。 -
分割数据集:
使用train_test_split
将数据集分割为训练集和测试集。from sklearn.model_selection import train_test_split x=data.drop(["Class"],axis=1) y=data["Class"] x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,random_state=0)
-
交叉验证寻找最佳正则化参数:
循环不同的正则化参数C
值,使用交叉验证计算召回率,并找到最佳C
值。from sklearn.model_selection import cross_val_score c=[0.01,0.1,1,10,100] d=[] for i in c:lr=LogisticRegression(penalty="l2",C=i,solver="lbfgs",max_iter=1000)score=cross_val_score(lr,x_train,y_train,cv=8,scoring="recall")scores=sum(score)/len(score)d.append(scores)
-
创建并训练逻辑回归模型:
使用找到的最佳C
值创建逻辑回归模型,并用训练集数据训练模型。best_c=c[np.argmax(d)] lr=LogisticRegression(penalty="l2",C=best_c,solver="lbfgs",max_iter=1000) lr.fit(x_train,y_train)
-
评估模型:
使用训练集和测试集数据评估模型性能,并打印分类报告。from sklearn import metrics predicted_train=lr.predict(x_train) print(metrics.classification_report(y_train,predicted_train))predicted_test=lr.predict(x_test) print(metrics.classification_report(y_test,predicted_test))
-
调整分类阈值并评估召回率:
循环不同的分类阈值thresh
,根据阈值调整预测结果,并计算每个阈值下的召回率。thresh=[0.1,0.2,0.3,0.4,0.5,0.6] recalls=[] for j in thresh:y_predic_proba=lr.predict_proba(x_test)y_predic_proba=pd.DataFrame(y_predic_proba).drop([0],axis=1)y_predic_proba[y_predic_proba[[1]]>j]=1y_predic_proba[y_predic_proba[[1]]<=j]=0recall=metrics.recall_score(y_test,y_predic_proba[1])recalls.append(recall)print(j,recall)
-
打印结果:
-
完整代码:
import numpy as np import pandas as pd from sklearn.linear_model import LogisticRegression from sklearn.preprocessing import StandardScaler scaler=StandardScaler() data=pd.read_csv("creditcard.csv") data["Amount"]=scaler.fit_transform(data[["Amount"]]) data=data.drop(["Time"],axis=1)from sklearn.model_selection import train_test_split x=data.drop(["Class"],axis=1) y=data["Class"] x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,random_state=0)from sklearn.model_selection import cross_val_score c=[0.01,0.1,1,10,100] d=[] for i in c:lr=LogisticRegression(penalty="l2",C=i,solver="lbfgs",max_iter=1000)score=cross_val_score(lr,x_train,y_train,cv=8,scoring="recall")scores=sum(score)/len(score)d.append(scores) best_c=c[np.argmax(d)] lr=LogisticRegression(penalty="l2",C=best_c,solver="lbfgs",max_iter=1000) lr.fit(x_train,y_train) from sklearn import metrics predicted_train=lr.predict(x_train) print(metrics.classification_report(y_train,predicted_train))predicted_test=lr.predict(x_test) print(metrics.classification_report(y_test,predicted_test))thresh=[0.1,0.2,0.3,0.4,0.5,0.6] recalls=[] for j in thresh:y_predic_proba=lr.predict_proba(x_test)y_predic_proba=pd.DataFrame(y_predic_proba).drop([0],axis=1)y_predic_proba[y_predic_proba[[1]]>j]=1y_predic_proba[y_predic_proba[[1]]<=j]=0recall=metrics.recall_score(y_test,y_predic_proba[1])recalls.append(recall)print(j,recall)