PythonGCS 2018-12-18
使用python语言,实现求特征选择的信息增益,可以同时满足特征中有连续型和二值离散型属性的情况。
师兄让我做一个特征选择的代码,我在网上找了一下,大部分都是用来求离散型属性的信息益益,但是我的数据是同时包含二值离散型和连续型属性的,所以这里实现了一下。
代码块
import numpy as np import math class IG(): def __init__(self,X,y): X = np.array(X) n_feature = np.shape(X)[1] n_y = len(y) orig_H = 0 for i in set(y): orig_H += -(y.count(i)/n_y)*math.log(y.count(i)/n_y) condi_H_list = [] for i in range(n_feature): feature = X[:,i] sourted_feature = sorted(feature) threshold = [(sourted_feature[inde-1]+sourted_feature[inde])/2 for inde in range(len(feature)) if inde != 0 ] thre_set = set(threshold) if float(max(feature)) in thre_set: thre_set.remove(float(max(feature))) if min(feature) in thre_set: thre_set.remove(min(feature)) pre_H = 0 for thre in thre_set: lower = [y[s] for s in range(len(feature)) if feature[s] < thre] highter = [y[s] for s in range(len(feature)) if feature[s] > thre] H_l = 0 for l in set(lower): H_l += -(lower.count(l) / len(lower))*math.log(lower.count(l) / len(lower)) H_h = 0 for h in set(highter): H_h += -(highter.count(h) / len(highter))*math.log(highter.count(h) / len(highter)) temp_condi_H = len(lower)/n_y *H_l+ len(highter)/n_y * H_h condi_H = orig_H - temp_condi_H pre_H = max(pre_H,condi_H) condi_H_list.append(pre_H) self.IG = condi_H_list def getIG(self): return self.IG if __name__ == "__main__": X = [[1, 0, 0, 1], [0, 1, 1, 1], [0, 0, 1, 0]] y = [0, 0, 1] print(IG(X,y).getIG())
输出结果为:
[0.17441604792151594, 0.17441604792151594, 0.17441604792151594, 0.6365141682948128]