理论
自闭者主要受到遗传和环境因素的共同影响。由于自闭症是一种谱系障碍,因此每个自闭症患者都有独特的优势和挑战。自闭症患者学习、思考和解决问题的方式可以是高技能的,也可以是严峻的挑战。研究表明,高质量的早期干预可以改善学习、沟通和社交技能,以及潜在的大脑发育。然而诊断过程可能需要数年时间。本项目主要实现自闭者的早期检测(正常vs非正常),为早期筛查和干预提供及时的预警。
工具
自闭者脑电数据集
方法实现
数据加载
python
from sklearn.metrics import roc_auc_score
from sklearn.model_selection import train_test_split
from xgboost import XGBClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.feature_selection import SelectKBest
from sklearn.feature_selection import mutual_info_classif,f_classif
from sklearn.pipeline import Pipeline
from sklearn.model_selection import cross_val_score,StratifiedKFold
from sklearn.feature_selection import RFE
from sklearn.feature_selection import RFECV
from sklearn.neural_network import MLPClassifier
from category_encoders.target_encoder import TargetEncoder
from sklearn.model_selection import GridSearchCV,RandomizedSearchCV
from sklearn.tree import DecisionTreeClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import StandardScaler,RobustScaler
from category_encoders import MEstimateEncoder
from sklearn.preprocessing import LabelEncoder
from imblearn.over_sampling import RandomOverSampler
from sklearn.inspection import permutation_importance
from imblearn.over_sampling import SMOTE
from sklearn.svm import SVC
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import confusion_matrix, accuracy_score
from sklearn.model_selection import train_test_split
from sklearn.naive_bayes import GaussianNB
from sklearn.ensemble import StackingClassifier,VotingClassifier
from sklearn.metrics import roc_auc_score, roc_curve, confusion_matrix, ConfusionMatrixDisplay
train=pd.read_csv('/Autism_Prediction/train.csv')
test=pd.read_csv('/Autism_Prediction/test.csv')
k-折交叉验证数据划分
python
np.random.seed(1) #I'm using this because there's some
#randomness in how the selectors work, without this, in each run we get different results
kf = StratifiedKFold(n_splits=2, random_state=None,shuffle=False) #for cross validation/ random_state
# is None because shuffle is False
score=[]
for train_index, val_index in kf.split(train_set,y):
#indices for train and validation sets
X_train, X_val =train_set.iloc[train_index,:], train_set.iloc[val_index,:]
y_train, y_val = y[train_index], y[val_index]
#******************************* CLEANING ***********************************
#for train set
X_train.ethnicity=X_train.ethnicity.str.replace('others','Others',regex=False)
X_train.ethnicity=X_train.ethnicity.str.replace('?','Others',regex=False)
X_train.relation=X_train.relation.str.replace('?','Others',regex=False)
X_train.relation=X_train.relation.str.replace('Health care professional','Others',regex=False)
#for validation set:
X_val.ethnicity=X_val.ethnicity.str.replace('others','Others',regex=False)
X_val.ethnicity=X_val.ethnicity.str.replace('?','Others',regex=False)
X_val.relation=X_val.relation.str.replace('?','Others',regex=False)
X_val.relation=X_val.relation.str.replace('Health care professional','Others',regex=False)
#***************************************ENCODING******************************************
#FOR ENCODING USE THE TRAINING VALUES, DO NOT CALCULATE THEM AGAIN FOR THE TEST SET!
le=LabelEncoder()
for col in ['jaundice','austim']:
#for the training set:
X_train[col]=le.fit_transform(X_train[col])
#for the validation set:
X_val[col]=le.transform(X_val[col])
#*********************Encoding Relation Column***************************
#create an encoding map, using the training set, then implementing it on val and test sets
rel=X_train.relation.value_counts()
rel=dict(zip(rel.index,range(len(rel))))
#for the training set:
X_train.relation=X_train.relation.map(rel)
#for the validation set: if there's a category not present in the map, we'll assign sth. to it
X_val.relation=X_val.relation.map(rel)
X_val.relation[X_val.relation.isna()]=len(rel)
#*********************Encoding Ethnicity Column***************************
#create an encoding map, using the training set, then implementing it on val and test sets
eth=X_train.ethnicity.value_counts()
eth=dict(zip(eth.index,range(len(eth))))
#for the training set:
X_train.ethnicity=X_train.ethnicity.map(eth)
#for the validation set: if there's a category not present in the map, we'll assign sth. to it
X_val.ethnicity=X_val.ethnicity.map(eth)
X_val.ethnicity[X_val.ethnicity.isna()]=len(eth)
#*****************************Encoding Country Of Res******************************
#create an encoding map, using the training set, then implementing it on val and test sets
cont=X_train.contry_of_res.value_counts()
cont=dict(zip(cont.index,range(len(cont))))
#for the training set:
X_train.contry_of_res=X_train.contry_of_res.map(cont)
#for the validation set: if there's a category not present in the map, we'll assign sth. to it
X_val.contry_of_res=X_val.contry_of_res.map(cont)
X_val.contry_of_res[X_val.contry_of_res.isna()]=len(cont)
#***************************Age Grouping***********************************
# age_grouper(X_train)
# age_grouper(X_val)
#*******************************Standardization*************************
ss=StandardScaler()
rs=RobustScaler()
X_train[['result','age']]=rs.fit_transform(X_train[['result','age']])
X_val[['result','age']]=rs.transform(X_val[['result','age']])
使用不同模型进行数据分类
python
model_list = ['KNearestNeighbours', 'DecisionTree', 'LGBM','XGBRF','CatBoostClassifier','RandomForest','Logistic Regression', 'SVC' ]
k近邻模型
python
# K Neighbors Classifier
kn_clf = KNeighborsClassifier(n_neighbors=6)
kn_clf.fit(X_train,y_train)
y_pred=pd.DataFrame(kn_clf.predict_proba(X_val))[1].values
score.append(roc_auc_score(y_val,y_pred))
np.array(score)
cm = confusion_matrix(y_val, kn_clf.predict(X_val))
cmd = ConfusionMatrixDisplay(cm)
cmd.plot();
决策树模型
python
#DecissionTree
dt_clf = DecisionTreeClassifier(max_leaf_nodes=10, random_state=0, criterion='entropy')
dt_clf.fit(X_train, y_train)
y_pred=pd.DataFrame(dt_clf.predict_proba(X_val))[1].values
score.append(roc_auc_score(y_val,y_pred))
np.array(score)
cm = confusion_matrix(y_val, dt_clf.predict(X_val))
cmd = ConfusionMatrixDisplay(cm)
cmd.plot();
lightgbm模型
python
# lightgbm
import lightgbm
lgb_clf = lightgbm.LGBMClassifier(max_depth=2, random_state=4)
lgb_clf.fit(X_train, y_train)
y_pred=pd.DataFrame(lgb_clf.predict_proba(X_val))[1].values
score.append(roc_auc_score(y_val,y_pred))
np.array(score)
cm = confusion_matrix(y_val, lgb_clf.predict(X_val))
cmd = ConfusionMatrixDisplay(cm)
cmd.plot();
代码获取
相关问题和项目开发,欢迎交流沟通。