Date Fruit Datasets

https://www.kaggle.com/datasets/muratkokludataset/date-fruit-datasets

Date Fruit Datasets

 

이번 데이터는 Kaggle 에 있는 7개의 과일 종류가 존재합니다.

각 데이터를 구분할 수있는 특성들이 주어졌을때, 분류하는 문제입니다

---

 

Import Modules

import numpy as np
import pandas as pd 
import os 
from sklearn.preprocessing import LabelEncoder,StandardScaler
from sklearn.model_selection import RandomizedSearchCV,train_test_split

!pip install openpyxl


# 데이터 불러오기

DATA_PATH ="/input/Date_Fruit_Datasets.xlsx"

data = pd.read_excel(DATA_PATH)
data.head()

 

데이터의 Null값과 분포 확인

 

data.isnull().sum()

#outputs 
AREA             0
PERIMETER        0
MAJOR_AXIS       0
MINOR_AXIS       0
ECCENTRICITY     0
EQDIASQ          0
SOLIDITY         0
CONVEX_AREA      0
EXTENT           0
ASPECT_RATIO     0
ROUNDNESS        0
COMPACTNESS      0
SHAPEFACTOR_1    0
SHAPEFACTOR_2    0
SHAPEFACTOR_3    0
SHAPEFACTOR_4    0
MeanRR           0
MeanRG           0
MeanRB           0
StdDevRR         0
StdDevRG         0
StdDevRB         0
SkewRR           0
SkewRG           0
SkewRB           0
KurtosisRR       0
KurtosisRG       0
KurtosisRB       0
EntropyRR        0
EntropyRG        0
EntropyRB        0
ALLdaub4RR       0
ALLdaub4RG       0
ALLdaub4RB       0
Class            0
dtype: int64

data.describe()

 

 

 

Features 와 Label로 분할 및 StandardScaler()

raw_features = data.iloc[:,:-1]
label = data.iloc[:,-1]
raw_features

scaler = StandardScaler()

features = pd.DataFrame(scaler.fit_transform(raw_features), columns= raw_features.columns)
features.head()

 

 

 

PCA를 통한 차원 축소 

 

 

Features 수가 많기 때문에, 적절한 차원 축소를 통해 핵심 특성만 찾기

 

from sklearn.decomposition import PCA

n_components= 15 
pca = PCA(n_components= n_components)

reduced_features = pd.DataFrame(pca.fit_transform(features), columns = [f"PC{i+1}" for i in range(n_components)])
reduced_features

 

차원을 축소한 DataFrame에 Class Features를 concatenate를 통해 붙여줍니다.

 

full_data = pd.concat([reduced_features,label],axis=1)
full_data

 

 

Split Train, Test Data

from sklearn.model_selection import StratifiedShuffleSplit

split = StratifiedShuffleSplit(n_splits=1, test_size=0.2)

for train_index, test_index in split.split(full_data,full_data["Class"]):
  train = full_data.loc[train_index]
  test = full_data.loc[test_index]
  
X_train = train.drop("Class",axis=1)
y_train = train["Class"]
X_test = test.drop("Class",axis=1)
y_test =test["Class"]

 

 

최적의 하이퍼 파라미터 찾기(RandomizedSearchCV)

분류기 Classifier 는 CatBoostClassifer를 사용

 

from catboost import CatBoostClassifier

cb_clf = CatBoostClassifier(task_type ="GPU",silent=True)

param_dict = {
    "early_stopping_rounds":np.arange(5,15),
    "learning_rate":np.linspace(0.05,0.15,10),
    "n_estimators":np.arange(100,200,10),
    "max_depth":np.arange(1,10),
}

rscv = RandomizedSearchCV(cb_clf,param_dict,scoring="accuracy",cv=10)

rscv.fit(X_train,y_train)

print(rscv.best_params_)
print(rscv.best_score_)


#outputs 

{'n_estimators': 170, 'max_depth': 6, 'learning_rate': 0.07222222222222222, 'early_stopping_rounds': 13}
0.9108176838810642

 

최적의 파라미터를 일일히 작성하기 보단 **rscv.best_params_ 를 통해 전달

 

from sklearn.metrics import classification_report

cb_clf = CatBoostClassifier(task_type="GPU", silent=True, **rscv.best_params_)

cb_clf.fit(X_train,y_train)

y_pred = cb_clf.predict(X_test)

print(classification_report(y_pred,y_test))

 

 

LogisticRegression을 통한 예측

 

from sklearn.linear_model import LogisticRegressionCV

lrcv = LogisticRegressionCV(cv=10, n_jobs=-1)

lrcv.fit(X_train,y_train)

y_pred= lrcv.predict(X_test)
print(classification_report(y_test,y_pred))

 

LogisticRegression 하이퍼 파라미터 튜닝

lrcv = LogisticRegressionCV(cv=10, n_jobs=-1, max_iter=10000)

param_dict = dict(
    solver=['newton-cg', 'lbfgs', 'liblinear'],
    penalty=["l1", "l2", "elasticnet"],
    Cs=[100, 10, 1.0, 0.1, 0.01]
)

rscv = RandomizedSearchCV(lrcv , param_dict, scoring='accuracy', cv=10)

rscv.fit(X_train, y_train)

print(rscv.best_params_)
print(rscv.best_score_)

#outputs 
{'solver': 'liblinear', 'penalty': 'l2', 'Cs': 10}
0.8969679186228483

 

최적의 파라미터로 로지스틱 회귀 예측

lrcv = LogisticRegressionCV(**rscv.best_params_)

lrcv.fit(X_train,y_train)

y_pred = lrcv.predict(X_test)
print(classification_report(y_pred,y_test))