kaggle入門（一） knn解決mnist

開發平臺 google colab + python3.6

package: panads，sklearn

panads 用來處理csv文件 教程鏈接

sklearn 是python 機器學習中常用的第三方模塊 教程鏈接

knn 講解以及使用 sklearn的教程鏈接

還是和以前一樣 先處理colab的文件夾掛載問題

from google.colab import drive
drive.mount('/content/drive/')

import os
os.chdir("/content/drive/My Drive/kaggle")

再導入要用的package （matplotlib和seaborn沒有用到，習慣性導入這些）

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import seaborn as sns
#將那些用matplotlib繪製的圖顯示在頁面裏而不是彈出一個窗口
%matplotlib inline   

np.random.seed(2)

from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix
from sklearn.neighbors import KNeighborsClassifier
import itertools

導入數據

train_data = pd.read_csv("mnist/train.csv")
test_data = pd.read_csv("mnist/test.csv")

x_train = train_data.values[:,1:]
y_train = train_data.values[:,0]
test_value = test_data.values

定義knn算法

def knnClassfiyer(value,lable):
    knnclf = KNeighborsClassifier()
    knnclf.fit(value,np.ravel(lable))
    return knnclf

訓練和預測

knnclf = knnClassfiyer(x_train,y_train)
test_label = knnclf.predict(test_value)

 

以下是獲取到的test_label(predict時間很長)

保存模型（kaggle需要交csv 包含ImageId,label）

test_label = pd.Series(test_label,name="Label")

submission = pd.concat([pd.Series(range(1,28001),name = "ImageId"),test_label],axis = 1)

submission.to_csv("mnist/Result_sklearn_KNN.csv",index=False)

 

使用kaggle api 提交 (需要先將kaggle.json 放入root下,可以參考colab和kaggle使用) 

!cp /content/drive/'My Drive'/kaggle/kaggle.json /root/.kaggle
!kaggle competitions submit -c digit-recognizer -f mnist/Result_sklearn_KNN.csv -m "forth submit"

運行完會有Successfully submitted to Digit Recognizer，再到kaggle 中My submissions 中查看