[ML] 결정트리(Decision Trees) 이용 H1N1 모델링 캐글 첫 제출

EDA

1 2 3 4 5 6 7 8 9 10 11 12 13

!pip install category_encoders # 카테고리 인코더스 설치 !pip install pandas-profiling==2.11.0 # 프로파일링 설치 import pandas as pd from sklearn.model_selection import train_test_split target = 'vacc_h1n1_f' # 타겟 설정 train = pd.merge(pd.read_csv('train.csv'), pd.read_csv('train_labels.csv')[target], left_index=True, right_index=True) test = pd.read_csv('test.csv') train.head().T

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

train.dtypes ''' h1n1_concern float64 h1n1_knowledge float64 behavioral_antiviral_meds float64 behavioral_avoidance float64 behavioral_face_mask float64 behavioral_wash_hands float64 behavioral_large_gatherings float64 behavioral_outside_home float64 behavioral_touch_face float64 doctor_recc_h1n1 float64 doctor_recc_seasonal float64 chronic_med_condition float64 child_under_6_months float64 health_insurance float64 health_worker float64 opinion_h1n1_vacc_effective object opinion_h1n1_risk object opinion_h1n1_sick_from_vacc object opinion_seas_vacc_effective object opinion_seas_risk object opinion_seas_sick_from_vacc object agegrp object education_comp float64 raceeth4_i int64 sex_i int64 inc_pov int64 marital float64 rent_own_r float64 employment_status object census_region int64 census_msa object n_adult_r float64 household_children float64 n_people_r float64 employment_industry object employment_occupation object hhs_region int64 state object vacc_h1n1_f int64 dtype: object ''' # 8:2 비율로 나눠줌 train, val = train_test_split(train, train_size=0.80, test_size=0.20, stratify=train[target], random_state=2) # stratify : 데이터가 계층화된 방식으로 분할되어 클래스 레이블로 사용 train.shape, val.shape, test.shape ''' ((33723, 39), (8431, 39), (28104, 38)) ''' train[target].value_counts(normalize=True) # 기준모델 : 0.76 ''' 0 0.760935 1 0.239065 Name: vacc_h1n1_f, dtype: float64 ''' # Profiling으로 데이터 리포트 생성 from pandas_profiling import ProfileReport profile = ProfileReport(train, minimal=True).to_notebook_iframe()

1 2	# 실수형 타입 확인 train.select_dtypes('float').head(20).T

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

# 중복된 특성 확인 train.T.duplicated() ''' h1n1_concern False h1n1_knowledge False behavioral_antiviral_meds False behavioral_avoidance False behavioral_face_mask False behavioral_wash_hands False behavioral_large_gatherings False behavioral_outside_home False behavioral_touch_face False doctor_recc_h1n1 False doctor_recc_seasonal False chronic_med_condition False child_under_6_months False health_insurance False health_worker False opinion_h1n1_vacc_effective False opinion_h1n1_risk False opinion_h1n1_sick_from_vacc False opinion_seas_vacc_effective False opinion_seas_risk False opinion_seas_sick_from_vacc False agegrp False education_comp False raceeth4_i False sex_i False inc_pov False marital False rent_own_r False employment_status False census_region False census_msa False n_adult_r False household_children False n_people_r False employment_industry False employment_occupation False hhs_region False state False vacc_h1n1_f False dtype: bool ''' # 카디널리티 확인 train.describe(exclude='number').T.sort_values(by='unique')

Feature Engineering

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

# 특성공학 import numpy as np def engineer(df): # 높은 카디널리티를 가지는 특성을 제거 selected_cols = df.select_dtypes(include=['number', 'object']) labels = selected_cols.nunique() # 특성별 카디널리티 리스트 selected_features = labels[labels <= 30].index.tolist() # 카디널리티가 30보다 작은 특성만 선택 df = df[selected_features] # 새로운 특성을 생성 behaviorals = [col for col in df.columns if 'behavioral' in col] df['behaviorals'] = df[behaviorals].sum(axis=1) dels = [col for col in df.columns if ('employment' in col or 'seas' in col)] df.drop(columns=dels, inplace=True) return df train = engineer(train) val = engineer(val) test = engineer(test) # 피쳐에서 타겟 드롭 features = train.drop(columns=[target]).columns # 훈련/검증/테스트 데이터를 특성과 타겟으로 분리 X_train = train[features] y_train = train[target] X_val = val[features] y_val = val[target] X_test = test[features]

Modelling

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

from category_encoders import OneHotEncoder from sklearn.impute import SimpleImputer from sklearn.preprocessing import StandardScaler from sklearn.linear_model import LogisticRegression from sklearn.pipeline import make_pipeline # 파이프라인 활용 로지스틱 회귀 모델링 pipe = make_pipeline( OneHotEncoder(), SimpleImputer(), StandardScaler(), LogisticRegression(n_jobs=-1) ) pipe.fit(X_train, y_train) print('검증세트 정확도', pipe.score(X_val, y_val)) y_pred = pipe.predict(X_test) ''' 검증세트 정확도 0.8185268651405527 ''' pipe.named_steps ''' {'logisticregression': LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True, intercept_scaling=1, l1_ratio=None, max_iter=100, multi_class='auto', n_jobs=-1, penalty='l2', random_state=None, solver='lbfgs', tol=0.0001, verbose=0, warm_start=False), 'onehotencoder': OneHotEncoder(cols=['opinion_h1n1_vacc_effective', 'opinion_h1n1_risk', 'opinion_h1n1_sick_from_vacc', 'agegrp', 'census_msa'], drop_invariant=False, handle_missing='value', handle_unknown='value', return_df=True, use_cat_names=False, verbose=0), 'simpleimputer': SimpleImputer(add_indicator=False, copy=True, fill_value=None, missing_values=nan, strategy='mean', verbose=0), 'standardscaler': StandardScaler(copy=True, with_mean=True, with_std=True)} ''' # 시각화 import matplotlib.pyplot as plt model_lr = pipe.named_steps['logisticregression'] enc = pipe.named_steps['onehotencoder'] encoded_columns = enc.transform(X_val).columns coefficients = pd.Series(model_lr.coef_[0], encoded_columns) plt.figure(figsize=(10,30)) coefficients.sort_values().plot.barh();

모델 개발

DecisionTree

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

from sklearn.tree import DecisionTreeClassifier pipe = make_pipeline( OneHotEncoder(use_cat_names=True), SimpleImputer(), DecisionTreeClassifier(random_state=1, criterion='entropy') ) pipe.fit(X_train, y_train) print('훈련 정확도: ', pipe.score(X_train, y_train)) print('검증 정확도: ', pipe.score(X_val, y_val)) ''' 훈련 정확도: 0.9908667674880646 검증 정확도: 0.7572055509429486 과적합 상태라고 볼 수 있다. ''' # 시각화 import graphviz from sklearn.tree import export_graphviz model_dt = pipe.named_steps['decisiontreeclassifier'] enc = pipe.named_steps['onehotencoder'] encoded_columns = enc.transform(X_val).columns dot_data = export_graphviz(model_dt , max_depth=3 , feature_names=encoded_columns , class_names=['no', 'yes'] , filled=True , proportion=True) display(graphviz.Source(dot_data))

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

from sklearn.metrics import f1_score pipe = make_pipeline( OneHotEncoder(use_cat_names=True), SimpleImputer(), DecisionTreeClassifier(max_depth=7, random_state=2) # depth 변경 ) pipe.fit(X_train, y_train) print('훈련 정확도', pipe.score(X_train, y_train)) print('검증 정확도', pipe.score(X_val, y_val)) # f1 score 계산 from sklearn.metrics import f1_score pred = pipe.predict(X_val) print('f1 스코어',f1_score(y_val, pred)) ''' 훈련 정확도 0.8317468789846693 검증 정확도 0.8254062388803226 f1 스코어 0.551219512195122 ''' # imputer median 사용 pipe = make_pipeline( OneHotEncoder(use_cat_names=True), SimpleImputer(strategy = 'median'), DecisionTreeClassifier(max_depth=3, random_state=2) # depth 값 조절 ) pipe.fit(X_train, y_train) print('훈련 정확도', pipe.score(X_train, y_train)) print('검증 정확도', pipe.score(X_val, y_val)) pred = pipe.predict(X_val) print('f1 스코어',f1_score(y_val, pred)) ''' 훈련 정확도 0.7994247249651573 검증 정확도 0.8054797770134029 f1 스코어 0.5866935483870968 ''' # 시각화 model_dt = pipe.named_steps['decisiontreeclassifier'] importances = pd.Series(model_dt.feature_importances_, encoded_columns) plt.figure(figsize=(10,30)) importances.sort_values().plot.barh();

Kaggle Submit

1 2 3 4 5 6 7 8 9 10

# 테스트 학습 tpred = pipe.predict(X_test) # 제출 양식 생성 submission = pd.read_csv('submission.csv') submission['vacc_h1n1_f'] = tpred submission # file export submission.to_csv('submission.csv', index= False)

• 5번에 걸쳐 튜닝해 그래도 조금이나마 F1 점수를 올려서 기분이 좋았다.
• 처음 해본 캐글 제출인데… 정말이지… 너무 재밌다…
• 앞으로 배울 모델링 기법들이 기대된다.
• 하지만 지금까지 배운것들을 소홀히 하지말자!