机器学习植物叶片识别

植物叶片的识别：给出叶片的数据集”叶子形状.csv”，描述植物叶片的边缘、形状、纹理这三个特征的数值型变量各有64个（共64*3=192个变量）。此外，还有1个记录每片叶片所属植物物种的分类型变量，共193个变量。请采用特征选择方法进行特征选择，并比较各特征选择结果的异同（20分）。通过数据建模，完成叶片形状的识别（30分）。

目录

目录

思路

1导入包

 2画出相关性矩阵（需要根据相关性矩阵，选择特征进行特征工程）

3进行PCA降维

4 KNN网格搜索优化 ，PCA前后

5 SVC

6.逻辑回归

 

 

---

 

 

思路

1.数据分析 可视化
2.建立特征工程（需要根据相关性矩阵，选择特征进行特征工程。包括对数据进行预处理，补充缺失值，归一化数据等）
3.机器学习算法模型去验证分析

 

 

1导入包

import pandas as pd
from sklearn import svm
import numpy as np
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA

 

 2画出相关性矩阵（需要根据相关性矩阵，选择特征进行特征工程）

 

Train= pd.read_csv("叶子形状.csv")
X = Train.drop(['species'], axis=1)
Y = Train['species']

  

 

 

Train['species'].replace(map_dic.keys(), map_dic.values(), inplace=True)
Train.drop(['id'], inplace = True, axis = 1)
Train_ture = Train['species']
#画出相关性矩阵
corr = Train.corr()
f, ax = plt.subplots(figsize=(25, 25))
cmap = sns.diverging_palette(220, 10, as_cmap=True)
sns.heatmap(corr, cmap=cmap, vmax=.3, center=0,
            square=True, linewidths=.5)
plt.show()

补充缺失值

np.all(np.any(pd.isnull(Train)))

#false

                                                                                        

训练集测试集划分（80%训练集、20%测试集）

x_train,x_test,y_train,y_test=train_test_split(X,Y,test_size=0.2,random_state=123)

对数据归一化处理

standerScaler = StandardScaler()
x_train = standerScaler.fit_transform(x_train)
x_test = standerScaler.fit_transform(x_test)

 

 

3进行PCA降维

pca = PCA(n_components=0.9)
x_train_1 = pca.fit_transform(x_train)
x_test_1 = pca.transform(x_test)

## 44个特征

 

 

4 KNN网格搜索优化 ，PCA前后

from sklearn.neighbors import KNeighborsClassifier

knn_clf0 = KNeighborsClassifier()
knn_clf0.fit(x_train, y_train)
print('KNeighborsClassifier')

y_predict = knn_clf0.predict(x_test)
score = accuracy_score(y_test, y_predict)
print("Accuracy: {:.4%}".format(score))

print("PCA后")

knn_clf1 = KNeighborsClassifier()
knn_clf1.fit(x_train_1, y_train)
print('KNeighborsClassifier')

y_predict = knn_clf1.predict(x_test_1)
score = accuracy_score(y_test, y_predict)
print("Accuracy: {:.4%}".format(score))

 

5 SVC

svc_clf = SVC(probability=True)
svc_clf.fit(x_train, y_train)

print("*"*30)
print('SVC')

y_predict = svc_clf.predict(x_test)
score = accuracy_score(y_test, y_predict)
print("Accuracy: {:.4%}".format(score))



svc_clf1 = SVC(probability=True)
svc_clf1.fit(x_train_1, y_train)

print("*"*30)
print('SVC')

y_predict1 = svc_clf1.predict(x_test_1)
score = accuracy_score(y_test, y_predict1)
print("Accuracy: {:.4%}".format(score))

 

 

6.逻辑回归

from sklearn.linear_model import LogisticRegressionCV

lr = LogisticRegressionCV(multi_class="ovr", 
                          fit_intercept=True, 
                          Cs=np.logspace(-2,2,20), 
                          cv=2, 
                          penalty="l2", 
                          solver="lbfgs", 
                          tol=0.01)

lr.fit(x_train,y_train)
print('逻辑回归')

y_predict = lr.predict(x_test)
score = accuracy_score(y_test, y_predict)
print("Accuracy: {:.4%}".format(score))

 

 

 

 

逻辑回归准确率最高98.65

经过特征选择和主成分分析不一定会提高准确率