dhlim / 약상자 star
임도현 임도현 2022-12-21
model
@958eb0501ed2941220b285a98958002d3b93ca30
 
958eb05
model.py (added)
+++ model.py
@@ -0,0 +1,95 @@
+import pandas as pd
+from datetime import datetime, date
+from sklearn.model_selection import train_test_split
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.metrics import accuracy_score ,confusion_matrix# 정확도 함수
+
+def taking_TF(user_id):
+    df= pd.read_csv(r'D:\takensoft\project7\약상자송신데이터.csv',encoding='euc-kr')
+    df['humantype']=df['type1'].astype(str)+df['type2'].astype(str)+df['type3'].astype(str)+df['type4'].astype(str)+df['type5'].astype(str)
+    df['taketype']=df['take1'].astype(str)+df['take2'].astype(str)+df['take3'].astype(str)+df['take4'].astype(str)+df['take5'].astype(str)
+    li_type = []
+    li_taketype = []
+
+    for i in range(0,len(df)):
+        li_type.append(int(df['humantype'][i],2))
+        li_taketype.append(int(df['taketype'][i],2))
+    humantype=pd.DataFrame({'humantypes':li_type,'taketypes':li_taketype})
+    df=pd.concat([humantype,df],axis=1)
+    
+    df=df[df['user_id']==user_id]
+    if len(df) <30:
+        print("분석불가")
+        return "분석불가"
+    df=df.reset_index()
+
+    df['date_info']=pd.to_datetime(df['date_info'])
+    li_weekday=[]
+    for i in range(0,len(df)):
+        li_weekday.append(df['date_info'][i].weekday())
+    day=pd.DataFrame({'weekday':li_weekday})
+    df=pd.concat([df,day],axis=1)
+    df['error']=df['humantypes']!=df['taketypes']
+    df['prev_humantypes']=df['humantypes'].shift()
+    df['prev_taketypes']=df['taketypes'].shift()
+
+    df=df[1:len(df)-1].reset_index()
+    X= df[['temp1','temp2','temp3','temp4','weekday','prev_humantypes','prev_taketypes']]
+    y = df['error']
+    train_x, test_x, train_y, test_y = train_test_split(X, y, test_size = 0.2, random_state = 42) # 학습데이터와 평가데이터의 비율을 8:2 로 분할| 
+    clf = RandomForestClassifier(n_estimators=20, max_depth=5,random_state=0)
+    clf.fit(train_x,train_y)
+    #print(len(train_x))
+    predict1 = clf.predict(test_x)
+    print(accuracy_score(test_y,predict1))
+    #print(predict1)
+    #print(confusion_matrix(test_y, predict1))
+    #feature_scores = pd.Series(clf.feature_importances_, index=train_x.columns).sort_values(ascending=False)
+    #print(feature_scores[:10])
+
+def taking_class(user_id):
+    df= pd.read_csv(r'D:\takensoft\project7\약상자송신데이터.csv',encoding='euc-kr')
+    df['humantype']=df['type1'].astype(str)+df['type2'].astype(str)+df['type3'].astype(str)+df['type4'].astype(str)+df['type5'].astype(str)
+    df['taketype']=df['take1'].astype(str)+df['take2'].astype(str)+df['take3'].astype(str)+df['take4'].astype(str)+df['take5'].astype(str)
+    li_type = []
+    li_taketype = []
+
+    for i in range(0,len(df)):
+        li_type.append(int(df['humantype'][i],2))
+        li_taketype.append(int(df['taketype'][i],2))
+    humantype=pd.DataFrame({'humantypes':li_type,'taketypes':li_taketype})
+    df=pd.concat([humantype,df],axis=1)
+    
+    df=df[df['user_id']==user_id]
+    if len(df) <30:
+        print("분석불가")
+        return "분석불가"
+    df=df.reset_index()
+
+    df['date_info']=pd.to_datetime(df['date_info'])
+    li_weekday=[]
+    for i in range(0,len(df)):
+        li_weekday.append(df['date_info'][i].weekday())
+    day=pd.DataFrame({'weekday':li_weekday})
+    df=pd.concat([df,day],axis=1)
+    df['error']=df['humantypes']!=df['taketypes']
+    df['prev_humantypes']=df['humantypes'].shift()
+    df['prev_taketypes']=df['taketypes'].shift()
+
+    df=df[1:len(df)-1].reset_index()
+    X= df[['temp1','temp2','temp3','temp4','weekday','prev_humantypes','prev_taketypes']]
+    y = df['taketypes']
+    train_x, test_x, train_y, test_y = train_test_split(X, y, test_size = 0.2, random_state = 42) # 학습데이터와 평가데이터의 비율을 8:2 로 분할| 
+    clf = RandomForestClassifier(n_estimators=20, max_depth=5,random_state=0)
+    clf.fit(train_x,train_y)
+    #print(len(train_x))
+    predict1 = clf.predict(test_x)
+    print(accuracy_score(test_y,predict1))
+    #print(confusion_matrix(test_y, predict1))
+    #feature_scores = pd.Series(clf.feature_importances_, index=train_x.columns).sort_values(ascending=False)
+    #print(feature_scores[:10])
+
+
+
+taking_TF('MEMBER_000000000000584')
+taking_class('MEMBER_000000000000584')(No newline at end of file)
Add a comment
List