Iris 鳶尾花數據集是一個經典數據集，在統計學習和機器學習領域都經常被用作示例。數據集內包含 3 類共 150 條記錄，每類各 50 個數據，每條記錄都有 4 項特征：花萼長度、花萼寬度、花瓣長度、花瓣寬度，可以通過這4個特征預測鳶尾花卉屬於（iris-setosa, iris-versicolour, iris-virginica）中的哪一品種。
據說在現實中，這三種花的基本判別依據其實是種子（因為花瓣非常容易枯萎）。
0 准備數據
下載數據 http://archive.ics.uci.edu/ml/machine-learning-databases/iris/iris.data
下面對 iris 進行探索性分析，首先導入相關包和數據集：
# 導入相關包
import numpy as np
import pandas as pd
from pandas import plotting
%matplotlib inline
import matplotlib.pyplot as plt
plt.style.use('seaborn')
import seaborn as sns
sns.set_style("whitegrid")
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
from sklearn.neighbors import KNeighborsClassifier
from sklearn import svm
from sklearn import metrics
from sklearn.tree import DecisionTreeClassifier
# 導入數據集
iris = pd.read_csv('F:\pydata\dataset\kaggle\iris.csv', usecols=[1, 2, 3, 4, 5])
查看數據集信息：
iris.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 150 entries, 0 to 149
Data columns (total 5 columns):
SepalLengthCm 150 non-null float64
SepalWidthCm 150 non-null float64
PetalLengthCm 150 non-null float64
PetalWidthCm 150 non-null float64
Species 150 non-null object
dtypes: float64(4), object(1)
memory usage: 5.9+ KB
查看數據集的頭 5 條記錄：
iris.head()
1 探索性分析
先查看數據集各特征列的摘要統計信息：
iris.describe()
通過Violinplot 和 Pointplot，分別從數據分布和斜率，觀察各特征與品種之間的關系：
# 設置顏色主題
antV = ['#1890FF', '#2FC25B', '#FACC14', '#223273', '#8543E0', '#13C2C2', '#3436c7', '#F04864']
# 繪制 Violinplot
f, axes = plt.subplots(2, 2, figsize=(8, 8), sharex=True)
sns.despine(left=True)
sns.violinplot(x='Species', y='SepalLengthCm', data=iris, palette=antV, ax=axes[0, 0])
sns.violinplot(x='Species', y='SepalWidthCm', data=iris, palette=antV, ax=axes[0, 1])
sns.violinplot(x='Species', y='PetalLengthCm', data=iris, palette=antV, ax=axes[1, 0])
sns.violinplot(x='Species', y='PetalWidthCm', data=iris, palette=antV, ax=axes[1, 1])
plt.show()
# 繪制 pointplot
f, axes = plt.subplots(2, 2, figsize=(8, 8), sharex=True)
sns.despine(left=True)
sns.pointplot(x='Species', y='SepalLengthCm', data=iris, color=antV[0], ax=axes[0, 0])
sns.pointplot(x='Species', y='SepalWidthCm', data=iris, color=antV[0], ax=axes[0, 1])
sns.pointplot(x='Species', y='PetalLengthCm', data=iris, color=antV[0], ax=axes[1, 0])
sns.pointplot(x='Species', y='PetalWidthCm', data=iris, color=antV[0], ax=axes[1, 1])
plt.show()
生成各特征之間關系的矩陣圖：
g = sns.pairplot(data=iris, palette=antV, hue= 'Species')
使用 Andrews Curves 將每個多變量觀測值轉換為曲線並表示傅立葉級數的系數，這對於檢測時間序列數據中的異常值很有用。
Andrews Curves 是一種通過將每個觀察映射到函數來可視化多維數據的方法。
plt.subplots(figsize = (10,8))
plotting.andrews_curves(iris, 'Species', colormap='cool')
plt.show()
下面分別基於花萼和花瓣做線性回歸的可視化：
g = sns.lmplot(data=iris, x='SepalWidthCm', y='SepalLengthCm', palette=antV, hue='Species')
g = sns.lmplot(data=iris, x='PetalWidthCm', y='PetalLengthCm', palette=antV, hue='Species')
最后，通過熱圖找出數據集中不同特征之間的相關性，高正值或負值表明特征具有高度相關性：
fig=plt.gcf()
fig.set_size_inches(12, 8)
fig=sns.heatmap(iris.corr(), annot=True, cmap='GnBu', linewidths=1, linecolor='k', square=True, mask=False, vmin=-1, vmax=1, cbar_kws={"orientation": "vertical"}, cbar=True)
從熱圖可看出，花萼的寬度和長度不相關，而花瓣的寬度和長度則高度相關。
2 機器學習
接下來，通過機器學習，以花萼和花瓣的尺寸為根據，預測其品種。
在進行機器學習之前，將數據集拆分為訓練和測試數據集。首先，使用標簽編碼將 3 種鳶尾花的品種名稱轉換為分類值（0, 1, 2）。
# 載入特征和標簽集
X = iris[['SepalLengthCm', 'SepalWidthCm', 'PetalLengthCm', 'PetalWidthCm']]
y = iris['Species']
# 對標簽集進行編碼
encoder = LabelEncoder()
y = encoder.fit_transform(y)
print(y)
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
2 2]
接着，將數據集以 7: 3 的比例，拆分為訓練數據和測試數據：
train_X, test_X, train_y, test_y = train_test_split(X, y, test_size = 0.3, random_state = 101)
print(train_X.shape, train_y.shape, test_X.shape, test_y.shape)
(105, 4) (105,) (45, 4) (45,)
檢查不同模型的准確性：
# Support Vector Machine
model = svm.SVC()
model.fit(train_X, train_y)
prediction = model.predict(test_X)
print('The accuracy of the SVM is: {0}'.format(metrics.accuracy_score(prediction,test_y)))
The accuracy of the SVM is: 1.0
# Logistic Regression
model = LogisticRegression()
model.fit(train_X, train_y)
prediction = model.predict(test_X)
print('The accuracy of the Logistic Regression is: {0}'.format(metrics.accuracy_score(prediction,test_y)))
The accuracy of the Logistic Regression is: 0.9555555555555556
# Decision Tree
model=DecisionTreeClassifier()
model.fit(train_X, train_y)
prediction = model.predict(test_X)
print('The accuracy of the Decision Tree is: {0}'.format(metrics.accuracy_score(prediction,test_y)))
The accuracy of the Decision Tree is: 0.9555555555555556
# K-Nearest Neighbours
model=KNeighborsClassifier(n_neighbors=3)
model.fit(train_X, train_y)
prediction = model.predict(test_X)
print('The accuracy of the KNN is: {0}'.format(metrics.accuracy_score(prediction,test_y)))
The accuracy of the KNN is: 1.0
上面使用了數據集的所有特征，下面將分別使用花瓣和花萼的尺寸：
petal = iris[['PetalLengthCm', 'PetalWidthCm', 'Species']]
train_p,test_p=train_test_split(petal,test_size=0.3,random_state=0)
train_x_p=train_p[['PetalWidthCm','PetalLengthCm']]
train_y_p=train_p.Species
test_x_p=test_p[['PetalWidthCm','PetalLengthCm']]
test_y_p=test_p.Species
sepal = iris[['SepalLengthCm', 'SepalWidthCm', 'Species']]
train_s,test_s=train_test_split(sepal,test_size=0.3,random_state=0)
train_x_s=train_s[['SepalWidthCm','SepalLengthCm']]
train_y_s=train_s.Species
test_x_s=test_s[['SepalWidthCm','SepalLengthCm']]
test_y_s=test_s.Species
model=svm.SVC()
model.fit(train_x_p,train_y_p)
prediction=model.predict(test_x_p)
print('The accuracy of the SVM using Petals is: {0}'.format(metrics.accuracy_score(prediction,test_y_p)))
model.fit(train_x_s,train_y_s)
prediction=model.predict(test_x_s)
print('The accuracy of the SVM using Sepal is: {0}'.format(metrics.accuracy_score(prediction,test_y_s)))
The accuracy of the SVM using Petals is: 0.9777777777777777
The accuracy of the SVM using Sepal is: 0.8
model = LogisticRegression()
model.fit(train_x_p, train_y_p)
prediction = model.predict(test_x_p)
print('The accuracy of the Logistic Regression using Petals is: {0}'.format(metrics.accuracy_score(prediction,test_y_p)))
model.fit(train_x_s, train_y_s)
prediction = model.predict(test_x_s)
print('The accuracy of the Logistic Regression using Sepals is: {0}'.format(metrics.accuracy_score(prediction,test_y_s)))
The accuracy of the Logistic Regression using Petals is: 0.6888888888888889
The accuracy of the Logistic Regression using Sepals is: 0.6444444444444445
model=DecisionTreeClassifier()
model.fit(train_x_p, train_y_p)
prediction = model.predict(test_x_p)
print('The accuracy of the Decision Tree using Petals is: {0}'.format(metrics.accuracy_score(prediction,test_y_p)))
model.fit(train_x_s, train_y_s)
prediction = model.predict(test_x_s)
print('The accuracy of the Decision Tree using Sepals is: {0}'.format(metrics.accuracy_score(prediction,test_y_s)))
The accuracy of the Decision Tree using Petals is: 0.9555555555555556
The accuracy of the Decision Tree using Sepals is: 0.6666666666666666
model=KNeighborsClassifier(n_neighbors=3)
model.fit(train_x_p, train_y_p)
prediction = model.predict(test_x_p)
print('The accuracy of the KNN using Petals is: {0}'.format(metrics.accuracy_score(prediction,test_y_p)))
model.fit(train_x_s, train_y_s)
prediction = model.predict(test_x_s)
print('The accuracy of the KNN using Sepals is: {0}'.format(metrics.accuracy_score(prediction,test_y_s)))
The accuracy of the KNN using Petals is: 0.9777777777777777
The accuracy of the KNN using Sepals is: 0.7333333333333333
從中不難看出，使用花瓣的尺寸來訓練數據較花萼更准確。正如在探索性分析的熱圖中所看到的那樣，花萼的寬度和長度之間的相關性非常低，而花瓣的寬度和長度之間的相關性非常高。