def calculate_table_train_tree(str_list,feature_list, trainx, trainy, predictx, predicty):
positive_table=[]
nagative_table=[]
length=len(str_list[0])
for i in range(0,length):
positive_table.append(len(str_list))
nagative_table.append(len(str_list))
for num in str_list:
if num_all_zero(num):
num[0]=1
feature=num_to_feature(num,feature_list)
train_sample=read_data_feature(feature,trainx)
predict_sample=read_data_feature(feature,predictx)
acc_original = train_tree(train_sample, trainy, predict_sample, predicty)
for i in range(0,length):
new_num=reverse_index(num,i)
feature = num_to_feature(new_num, feature_list)
train_sample = read_data_feature(feature, trainx)
predict_sample = read_data_feature(feature, predictx)
acc_new= train_tree(train_sample, trainy, predict_sample, predicty)
if acc_new>acc_original:
if num[i]==0:
nagative_table[i]+=1
else:
positive_table[i]+=1
else:
if num[i]==1:
nagative_table[i]+=1
else:
positive_table[i]+=1
return positive_table,nagative_table
def one_point_hybridization_knn(forest_list,feature_list,trainx,trainy,predictx,predicty,neighbor):
'''
一次单点杂交
:param forest_list: 森林列表，双层列表
:param feature_list:特征集合索引,特征集合的角标
:param trainx:训练集
:param trainy:训练集对应的分类
:param predictx:预测集合
:param predicty:预测集对应的分类
:return:森林字典:包括单点杂交后每棵树的准确率和森林的01串
森林列表:单点杂交后新的森林
'''
forest = {} # 记录森林里的准确率
forest_list=one_point_hybridization(forest_list)
for num in forest_list:
feature=num_to_feature(num,feature_list)
train_sample=read_data_feature(feature,trainx)
predict_sample=read_data_feature(feature,predictx)
acc = train_knn(train_sample, trainy, predict_sample, predicty,neighbor)
num_string = num_to_string(num)
forest[num_string] = acc
return forest,forest_list
feature_list = [] # 特征集合索引,特征集合的角标
for i in range(0, len(trainx[0])):
feature_list.append(i)
forest = {} #记录森林里的准确率
init_forest = random_init(50, len(trainx[0]))
# for i in init_forest:
# print i
# print 'trainx',trainx
# print 'trainy',trainy
for num in init_forest:
feature = num_to_feature(num, feature_list)
train_sample = read_data_feature(feature, trainx)
predict_sample = read_data_feature(feature, predictx)
# print 'train_sample',train_sample
# print 'train_y',trainy
# print 'predict_sample',predict_sample
# print 'predict_y',predicty
acc = train_knn(train_sample, trainy, predict_sample, predicty, 1)
num_string = num_to_string(num)
forest[num_string] = acc
# forest_area = []
forest_old = init_forest
res = {}
for i in range(0, 5):