Applied Machine Learning in Python 2 - Lab: KNN

我們用癌症統計資料來玩玩看KNN

讀入資料

import numpy as np
import pandas as pd
from sklearn.datasets import load_breast_cancer
 
cancer = load_breast_cancer()

不過這是一個dict，還不是一個pandas dataframe。雖然Scikit-learn不需要一定使用dataframe，但這對清理資料有幫助，嘗試將之轉換成dataframe。

def to_dataframe():
    df = pd.DataFrame(data=cancer['data'], columns=cancer['feature_names'])
    df['target'] = cancer['target']
    return df

class distribution

惡性與良性的分類各有多少呢?

def class_distribution():
    df = to_dataframe()
    result = pd.Series({'malignant':len(df[df['target']==0]),
                        'benign':len(df[df['target']==1])})
    return result

區分label與data

def split_data_label():
    df = to_dataframe()
    X = df[df.columns[:-1]]
    y = df['target']
 
    return X,y

製作75% : 25% training set vs test set

def training_set():
    X, y = split_data_label()
 
    from sklearn.model_selection import train_test_split
    X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)
 
    return X_train, X_test, y_train, y_test

train 1-NN classifier

def _1NN_classifier():
    X_train, X_test, y_train, y_test = training_set()
 
    knn = KNeighborsClassifier(n_neighbors=1)
    knn.fit(X_train, y_train)
 
    return  knn

使用classifier來預測

要製作input，這邊使用一個假的data，就是每個feature在dataframe的mean組成的feature vector。

def predict():
    cancerdf = to_dataframe()
    means = cancerdf.mean()[:-1].values.reshape(1, -1)
 
    knn = _1NN_classifier()
    result = knn.predict(means)
 
    return  result

predict test set

當然我們要來evaluate estimator的好壞的話，還是要來測試test set:

def predict_test_set():
    X_train, X_test, y_train, y_test = training_set()
    knn = _1NN_classifier()
 
    result = knn.predict(X_test)
 
    return  result

Accuracy

最後可以檢驗此次training結果的accuracy:

def answer_eight():
    X_train, X_test, y_train, y_test = training_set()
    knn = _1NN_classifier()
 
    result = knn.score(X_test, y_test)
 
    return  result

這次training結果對training set以及test set中的兩種classes的accuracy如下: