more Chinese comments on chapter 2

Author: dashjim

.gitignore

@@ -7,3 +7,4 @@ ch03/data/379/raw/
 ch03/data/379/
 ch03/charts/
 playGround/user-basket/data/raw_user.csv
+ch01/charts/

ch02/chapter.py

@@ -1,3 +1,4 @@
+# coding=utf-8
 # This code is supporting material for the book
 # Building Machine Learning Systems with Python
 # by Willi Richert and Luis Pedro Coelho
@@ -29,11 +30,14 @@
 elift==2:
 c='b'
 marker='x'
+# target数组中=0的是一类，作为features的下标，来选出同类的，再把列选出来。
+# feature一共有四种（四列）【0，1,2,3】，一共可以组合出6种图，这里只显示一种
 plt.scatter(features[target==t, 0],
-features[target==t, 1],
+features[target==t, 2],
 marker=marker,
 c=c)
 # We use NumPy fancy indexing to get an array of strings:
+# 可以理解为新的数组中的元素是不断的用每一个下标（target）选出来的。
 labels=target_names[target]
 
 # The petal length is the feature at position 2
@@ -49,23 +53,26 @@
 
 print('Minimum of others:{0}.'.format(min_non_setosa))
 
+# setosa 是最好区分的，可以直接选出来，后面都是要区分另外两种
 # ~ is the boolean negation operator
 features=features[~is_setosa]
 labels=labels[~is_setosa]
 # Build a new target variable, is_virigina
+# 这是后面用来检测预测效果用的。
 is_virginica= (labels=='virginica')
 
 # Initialize best_acc to impossibly low value
 best_acc=-1.0
-forfiinrange(features.shape[1]):
+forfiinrange(features.shape[1]):# features 是一个N * 4的二维数组
 # We are going to test all possible thresholds
 thresh=features[:,fi]
-fortinthresh:
+fortinthresh:# 分别取第一个feature类别（一列数据）中的每一个数据。
 
 # Get the vector for feature `fi`
 feature_i=features[:, fi]
 # apply threshold `t`
 pred= (feature_i>t)
+# 预测和真实值一致则为True也就是1，其他为零，然后求平均数
 acc= (pred==is_virginica).mean()
 rev_acc= (pred==~is_virginica).mean()
 ifrev_acc>acc:
@@ -81,13 +88,18 @@
 best_reverse=reverse
 
 print(best_fi, best_t, best_reverse, best_acc)
-
-defis_virginica_test(fi, t, reverse, example):
-'Apply threshold model to a new example'
-test=example[fi] >t
-ifreverse:
-test=nottest
-returntest
+# result:
+# (3, 1.6000000000000001, False, 0.93999999999999995)
+
+# The following code has error and thus disable for now - Jim
+# def is_virginica_test(fi, t, reverse, example):
+# 'Apply threshold model to a new example'
+# test = example[fi] > t
+# if reverse:
+# test = not test
+# return test
+
+# 交叉验证：1 - 极端情况是每次取出来一个作为测试数据，其他的作为训练数据。
 fromthresholdimportfit_model, predict
 
 # ning accuracy was 96.0%.
@@ -96,9 +108,9 @@ def is_virginica_test(fi, t, reverse, example):
 
 foreiinrange(len(features)):
 # select all but the one at position `ei`:
-training=np.ones(len(features), bool)
-training[ei] =False
-testing=~training
+training=np.ones(len(features), bool)# 全部为True
+training[ei] =False# 一个为false
+testing=~training# 全部为false, 一个为True
 model=fit_model(features[training], is_virginica[training])
 predictions=predict(model, features[testing])
 correct+=np.sum(predictions==is_virginica[testing])
@@ -124,17 +136,18 @@ def is_virginica_test(fi, t, reverse, example):
 features, labels=load_dataset('seeds')
 
 
-
+# 交叉验证：2 - k fold
 fromsklearn.neighborsimportKNeighborsClassifier
 classifier=KNeighborsClassifier(n_neighbors=1)
 fromsklearn.cross_validationimportKFold
 
 kf=KFold(len(features), n_folds=5, shuffle=True)
 means= []
+# kf只是下标，还是要事先准备好训练数据和标记。
 fortraining,testinginkf:
 # We learn a model for this fold with `fit` and then apply it to the
 # testing data with `predict`:
-classifier.fit(features[training], labels[training])
+classifier.fit(features[training], labels[training])# features包含的多列数据
 prediction=classifier.predict(features[testing])
 
 # np.mean on an array of booleans returns fraction
@@ -143,7 +156,7 @@ def is_virginica_test(fi, t, reverse, example):
 means.append(curmean)
 print('Mean accuracy:{:.1%}'.format(np.mean(means)))
 
-
+# 同样是KNN，对比参数标准化后的效果
 fromsklearn.pipelineimportPipeline
 fromsklearn.preprocessingimportStandardScaler
 
@@ -161,4 +174,5 @@ def is_virginica_test(fi, t, reverse, example):
 # of correct decisions for this fold:
 curmean=np.mean(prediction==labels[testing])
 means.append(curmean)
+# 尼玛，每次结果都不一样！什么鬼？
 print('Mean accuracy:{:.1%}'.format(np.mean(means)))