PCA
PCA主要是用来数据降维，将高纬度的特征映射到低维度，具体可学习线性代数。
这里，我们使用sklearn中的PCA.
from sklearn.decomposition import PCA X = np.array([[-1, -1, 1, -3], [-2, -1, 1, -3], [-3, -2, 1, -3], [1, 1, 1, -3], [2, 1, 1, -3], [3, 2, -1, -3]]) pca = PCA(n_components=4) pca.fit(X) print(pca.explained_variance_ratio_) #各成分百分比 print(pca.explained_variance_) #各成分值 pca = PCA(n_components=1) #原来是4维，现在降至1维 XX = pca.fit_transform(X) print(XX)
结果：
[0.94789175 0.04522847 0.00687978 0. ] [8.21506183 0.39198011 0.05962472 0. ] [[-1.42149543] [-2.2448796 ] [-3.60382274] [ 1.29639085] [ 2.11977502] [ 3.85403189]]
其实，直接看数据也能发现。例如，最后一维没变化所以百分比为0，倒数第二维只有一点点变化所以百分比也很小，它们对结果的影响很小，在降维时可以去掉。
KNN
所谓K最近邻，就是k个最近的邻居的意思，说的是每个样本都可以用它最接近的k个邻居来代表。
kNN算法的核心思想是如果一个样本在特征空间中的k个最相邻的样本中的大多数属于某一个类别，则该样本也属于这个类别，并具有这个类别上样本的特性。
流程如下：
- 计算出样本数据和待分类数据的距离；
- 为待分类数据选择K个与其距离最小的样本；
- 统计出K个样本中大多数样本所属的分类；
- 这个分类就是待分类数据所属的分类。
classifier = KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski', metric_params=None, n_jobs=1, n_neighbors=10, p=2, weights='uniform')
超参数需要自己尝试。
其他
from sklearn.neighbors import KNeighborsClassifier from sklearn.linear_model import LogisticRegression from sklearn.ensemble import RandomForestClassifier from sklearn.tree import DecisionTreeClassifier from sklearn.svm import SVC from sklearn.naive_bayes import GaussianNB from sklearn import metrics classifier = LogisticRegression(random_state = 0) #0.78 #classifier = KNeighborsClassifier(algorithm='kd_tree',n_neighbors = 5, metric = 'minkowski', p = 2, weights='uniform') #0.839 #classifier = SVC(kernel = 'linear', random_state = 0) #0.81 #classifier = SVC(kernel = 'rbf', random_state = 0) #0.77 #classifier = GaussianNB() #0.77 #classifier = DecisionTreeClassifier(criterion = 'entropy', random_state = 0) #0.64 #classifier = RandomForestClassifier(n_estimators = 10, criterion = 'entropy', random_state = 0) #0.83 classifier.fit(X_stard, Y_stard) YY_pred = classifier.predict(X_pred) result_NMI=metrics.normalized_mutual_info_score(YY_pred, Y_pred) print("result_NMI:",result_NMI) #3,1,minkowski 3,1,manhattan
GridSearchCV寻找超参数
sklearn调参有一个工具gridsearchcv，它存在的意义就是自动调参，只要把参数输进去，就可以对算法进行相应的调优，找到合适的参数。
### KNN from sklearn.neighbors import KNeighborsClassifier from sklearn.model_selection import GridSearchCV clf = KNeighborsClassifier() n_neighbors = list(range(1,10)) weights = ['uniform','distance'] algorithm_options = ['auto','ball_tree','kd_tree','brute'] leaf_range = list(range(1,10)) p = list(range(1,10)) param_grid = [{'n_neighbors': n_neighbors, 'weights': weights, 'algorithm': algorithm_options, 'leaf_size': leaf_range, 'p':p}] grid_search = GridSearchCV(clf, param_grid=param_grid, cv=10) grid_search.fit(X_pred, Y_pred) grid_search.best_score_, grid_search.best_estimator_, grid_search.best_params_
结果：
(0.9675572519083969, KNeighborsClassifier(algorithm='auto', leaf_size=1, metric='minkowski', metric_params=None, n_jobs=None, n_neighbors=7, p=2, weights='uniform'), {'algorithm': 'auto', 'leaf_size': 1, 'n_neighbors': 7, 'p': 2, 'weights': 'uniform'})
参考链接：
1. https://blog.csdn.net/puredreammer/article/details/52255025
2. https://www.makcyun.top/2019/06/15/Machine_learning08.html