特征决定了最优效果的上限，算法与模型只是让效果更逼近这个上限，所以特征工程与选择什么样的特征很重要！
以下是一些特征筛选与降维技巧
# -*- coding:utf-8 -*- import scipy as sc import libsvm_file_process as data_process import numpy as np from minepy import MINE from sklearn.feature_selection import VarianceThreshold from sklearn.feature_selection import SelectKBest from sklearn.feature_selection import chi2 from sklearn.feature_selection import f_regression from sklearn.feature_selection import RFE from sklearn.svm import SVR from sklearn.linear_model import LogisticRegression from sklearn.decomposition import PCA from sklearn.discriminant_analysis import LinearDiscriminantAnalysis class feature_select: """ 特征筛选方式： 相关链接：http://scikit-learn.org/stable/modules/classes.html#module-sklearn.feature_selection 皮尔逊相关性 互信息 单因素 - 卡方判断，F值，假正率 方差过滤 递归特征消除法 - 每次消除一个特征，依据是特征前面的系数 基于模型（LR/GBDT等）的特征选择 SelectFromModel 模型（LR/GBDT）必须有feature_importances_ 或 coef_这个属性 降维： PCA（unsurperised）：一般用于无监督情况下的降维，有监督的时候，也可以小幅降维 去除噪音，然后再使用LDA 降维 LDA（surperised）:本质上是一个分类器,在使用上，要求降低的维度要小于分类的维度 """ def __init__(self): self.data_path = "/trainData/libsvm2/" self.trainData = ["20180101"] # 计算互信息 self.mine = MINE(alpha=0.6, c=15, est="mic_approx") # 方差过滤 一般用于无监督学习 self.variance_filter = VarianceThreshold(threshold=0.1) # chi2 - 卡方检验; f_regression - f值; SelectFpr-假正率;等 self.chi_squared = SelectKBest(f_regression, k=2) # 递归特征消除 self.estimator = LogisticRegression() # SVR(kernel="linear") self.selector = RFE(self.estimator, 5, step=1) # PCA 降维 self.pca = PCA(n_components=5) # LDA 降维 self.lda = LinearDiscriminantAnalysis(n_components=2) def select(self): for i in range(len(self.trainData)): generator = data_process.get_data_batch(self.data_path + self.trainData[i] + "/part-00000", 100000) labels, features = generator.next() # 方差过滤 filter1 = self.variance_filter.fit_transform(features) print filter1.shape, features.shape print self.variance_filter.get_support() # 卡方检验 filter2 = self.chi_squared.fit_transform(features, labels) print filter2.shape print self.chi_squared.get_support() # 递归特征消除（比较耗时 暂时先注释掉） # self.selector.fit(features, labels) # print self.selector.support_ # PCA 降维 transform1 = self.pca.fit_transform(features) print 'transform1:', transform1 # LDA降维 self.lda.fit(features, labels) transform2 = self.lda.transform(features) print 'transform2:', transform2 for j in range(int(features.shape[1]) - 870): features_j = features[0:, j + 870: j + 871] self.mine.compute_score(features_j.flatten(), labels.flatten()) # 计算互信息 print self.mine.mic() # 计算皮尔逊系数 print j, sc.stats.pearsonr(features_j.reshape(-1, 1), labels.reshape(-1, 1)) if __name__ == '__main__': feature_util = feature_select() feature_util.select()
