1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
|
def read_fa(file, mode='train'):
assert mode in {'train','test'}
labels = []
seqs_info = []
cates_id = []
seq = ''
with open(file,mode='r') as f:
line = f.readline().strip()
while line:
if line[0]=='>':
info = line[1:].split(' ')
cates_id.append(info[0])
if mode == 'train':
labels.append(''.join(info[1].split('.')[:2]))
if seq:
seqs_info.append(seq)
seq = ''
else:
seq += line
line = f.readline().strip()
seqs_info.append(seq)
return cates_id,seqs_info,labels
from utils import *
import pandas as pd
import numpy as np
from sklearn.metrics import f1_score, fbeta_score, precision_score, recall_score, roc_auc_score
from sklearn.model_selection import StratifiedKFold as KFold
import lightgbm as lgb
train_file = './训练集/astral_train.fa'
test_file = './测试集/astral_test.fa'
train_sample_id, train_seqs_info, train_labels = read_fa(train_file, mode='train')
test_sample_id, test_seqs_info, _ = read_fa(test_file, mode='test')
train_data = {
'sample_id': train_sample_id,
'seq_info': train_seqs_info,
'label': train_labels
}
test_data = {
'sample_id': test_sample_id,
'seq_info': test_seqs_info,
}
label_map = {l:idx for idx,l in enumerate(set(train_labels))}
rev_label_map = {v:k for k,v in label_map.items()}
train = pd.DataFrame(data=train_data)
test = pd.DataFrame(data=test_data)
train['label'] = train['label'].map(label_map)
alp = list(set(''.join(train_seqs_info + test_seqs_info)))
train['seq_len'] = train['seq_info'].apply(lambda x:len(x))
test['seq_len'] = test['seq_info'].apply(lambda x:len(x))
for s in alp:
train['count_'+s] = train['seq_info'].apply(lambda x:x.count(s))
train['freq_'+s] = train['seq_info'].apply(lambda x:x.count(s)/len(x))
test['count_'+s] = test['seq_info'].apply(lambda x:x.count(s))
test['freq_'+s] = test['seq_info'].apply(lambda x:x.count(s)/len(x))
feats = [i for i in train.columns if i not in ['label','sample_id','seq_info']]
x_train = train[feats]
y_train = train['label']
x_test = test[feats]
params = {
'boosting_type': 'gbdt',
'objective': 'multiclass',
'num_class': 245,
'metric': 'multi_error',
'num_leaves': 300,
'min_data_in_leaf': 500,
'learning_rate': 0.007,
'max_depth': 8,
'feature_fraction': 0.8,
'bagging_fraction': 0.8,
'bagging_freq': 5,
'lambda_l1': 0.4,
'lambda_l2': 0.5,
'min_gain_to_split': 0.2,
'verbose': -1,
'num_threads':2,
}
folds = KFold(n_splits=5, shuffle=True, random_state=2021)
oof = np.zeros([len(x_train),245])
predictions = np.zeros([len(x_test),245])
for fold_, (trn_idx, val_idx) in enumerate(folds.split(x_train, y_train)):
print("fold n°{}".format(fold_+1))
trn_data = lgb.Dataset(x_train.iloc[trn_idx], y_train.iloc[trn_idx])
val_data = lgb.Dataset(x_train.iloc[val_idx], y_train.iloc[val_idx])
num_round = 1000
clf = lgb.train(params,
trn_data,
num_round,
valid_sets = [trn_data, val_data],
verbose_eval = 100,
early_stopping_rounds = 50)
oof[val_idx] = clf.predict(x_train.iloc[val_idx][feats], num_iteration=clf.best_iteration)
predictions += clf.predict(x_test, num_iteration=clf.best_iteration) / folds.n_splits
x_test['sample_id'] = test['sample_id']
x_test['category_id'] = [rev_label_map[list(x).index(max(x))] for x in predictions]
x_test['category_id'] = x_test['category_id'].apply(lambda x: x[0]+'.'+x[1:])
x_test[['sample_id', 'category_id']].to_csv('base_sub.csv', index=False)
y_pre = oof.argmax(axis=1)
print("F1 score: {}".format(f1_score(y_train, y_pre,average='micro')))
print("Precision score: {}".format(precision_score(y_train, y_pre,average='micro')))
print("Recall score: {}".format(recall_score(y_train, y_pre,average='micro')))
|
