二十一世纪是生命科学的

天池精准医疗大赛——人工智能辅助糖尿病遗传风险预测复盘

初赛

赛题

大赛初赛数据共包含两个文件，训练文件d_train.csv和测试文件d_test.csv，每个文件第一行是字段名，之后每一行代表一个个体。文件共包含42个字段，包含数值型、字符型、日期型等众多数据类型，部分字段内容在部分人群中有缺失，其中第一列为个体ID号。训练文件的最后一列为标签列，既需要预测的目标血糖值。

数据


分析

对于缺失严重的数据（乙肝抗原抗体相关数据，乙肝病毒与糖尿病的相关性不大）直接删掉，缺少性别的用男性补全，删掉年龄小于16以及血糖大于18的异常数据

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
def make_feat(train, test):
    train_id = train.id.values.copy()
    test_id = test.id.values.copy()

## 去除异常值
    train = train[train['old']>=16]
    train = train[train['xuetang']<=18]

    data = pd.concat([train, test])
## 补全性别缺失值
    data['male'] = data['male'].map({'男': 1, '女': 0,'??':1})
## 删除数据严重缺失的列，以及与血糖值无关的体检日期
    data.drop(['yigan_biaomian_kangyuan', 'yigan_biaomian_kangti', 'yigan_e_kangyuan', 'yigan_e_kangti', 'yigan_hexin_kangti','date'], axis=1, inplace=True)

    train_feat = data[data.id.isin(train_id)]
    test_feat = data[data.id.isin(test_id)]
    del train_feat['id']
    del test_feat['id']
    return train_feat, test_feat

1
2
3
4
def evalerror(pred, df):
    label = df.get_label().values.copy()
    score = mean_squared_error(label, pred) * 0.5
    return ('mse', score, False)
1
2
3
4
5
6
7
8
def xiaoshu(array):
    new_list = []
    for i in array:
        j = sum(i)/5
        new_list.append(round(j, 3))
    return np.array(new_list)
def to_csv(DF,file_name):
    pd.DataFrame(DF).to_csv(file_name, index = False, header = None)
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
# 导入数据
train = pd.read_csv('d_train_20180102_b.csv')
test = pd.read_csv('d_test_B_20180128.csv')
# 清理数据
train_feat, test_feat = make_feat(train, test)
predictors = [f for f in test_feat.columns if f not in ['xuetang']]

params = {'learning_rate': 0.01,'boosting_type': 'gbdt','objective': 'regression','metric': 'mse','sub_feature': 0.9,'num_leaves': 16,'colsample_bytree': 0.9,'verbose': -1}
cores = []
train_preds = np.zeros(train_feat.shape[0])
test_preds = np.zeros((test_feat.shape[0], 5))
# 开始5折训练
kf = KFold(len(train_feat), n_folds=5, shuffle=True, random_state=520)

for i, (train_index, test_index) in enumerate(kf):
    train_feat1 = train_feat.iloc[train_index]
    train_feat2 = train_feat.iloc[test_index]
    lgb_train1 = lgb.Dataset(train_feat1[predictors], train_feat1['xuetang'], categorical_feature=['male'])
    lgb_train2 = lgb.Dataset(train_feat2[predictors], train_feat2['xuetang'])
    gbm = lgb.train(params,lgb_train1,num_boost_round=3000,valid_sets=lgb_train2,verbose_eval=100,feval=evalerror,early_stopping_rounds=150)
    feat_imp = pd.Series(gbm.feature_importance(), index=predictors).sort_values(ascending=False)
    train_preds[test_index] += gbm.predict(train_feat2[predictors])
    test_preds[:, i] = gbm.predict(test_feat[predictors])
# 导出结果
to_csv(xiaoshu(test_preds), 'result_B.csv')