import pandas as pd
import numpy as np


from sklearn.preprocessing import MinMaxScaler, Normalizer, LabelEncoder
from sklearn.model_selection import train_test_split, KFold
from sklearn.svm import SVC
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, precision_score, recall_score
import matplotlib.pyplot as plt
from scipy.fftpack import dct, idct


df = pd.read_csv("Wine.csv")
class_col = df.columns[-1]

matran = df.values
a1 = matran[:, :4]
dieuchinh1 = MinMaxScaler(feature_range=(0, 10))
a1 = dieuchinh1.fit_transform(a1)
print(a1[:5, :])
a2 = matran[:, 4:-1]
dieuchinh2 = Normalizer()
a2 = dieuchinh2.fit_transform(a2)
print(a2[:5, :])

encode = LabelEncoder()
label = matran[:, -1]
label = encode.fit_transform(label)
print(label[:5])

a3 = matran[:5, :-1]
print("5 dong dau")
print(a3)
thuan = dct(dct(a3.T, norm="ortho").T, norm="ortho")
print(thuan)

nghich = idct(idct(a3.T, norm="ortho").T, norm="ortho")
print(nghich)

X = df.drop(matran, axis=1).values
y = df[matran].values

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.03, random_state=42
)


# d. Tạo KFold 10

kf = KFold(n_splits=10, shuffle=True, random_state=42)


# e. SVM 10-fold

svm_acc = []
svm_pre = []
svm_rec = []

for train_idx, test_idx in kf.split(X_train):
    Xtr, Xte = X_train[train_idx], X_train[test_idx]
    ytr, yte = y_train[train_idx], y_train[test_idx]

    model_svm = SVC()
    model_svm.fit(Xtr, ytr)
    pred = model_svm.predict(Xte)

    svm_acc.append(accuracy_score(yte, pred))
    svm_pre.append(precision_score(yte, pred, average='macro'))
    svm_rec.append(recall_score(yte, pred, average='macro'))

log_acc = []
log_pre = []
log_rec = []

for train_idx, test_idx in kf.split(X_train):
    Xtr, Xte = X_train[train_idx], X_train[test_idx]
    ytr, yte = y_train[train_idx], y_train[test_idx]

    model_log = LogisticRegression(max_iter=1000)
    model_log.fit(Xtr, ytr)
    pred = model_log.predict(Xte)

    log_acc.append(accuracy_score(yte, pred))
    log_pre.append(precision_score(yte, pred, average='macro'))
    log_rec.append(recall_score(yte, pred, average='macro'))


# g. Vẽ đồ thị


plt.figure(figsize=(10,5))
plt.plot(svm_acc, label="SVM Accuracy")
plt.plot(log_acc, label="Logistic Accuracy")
plt.title("Accuracy qua 10-fold")
plt.legend()
plt.show()

plt.figure(figsize=(10,5))
plt.plot(svm_pre, label="SVM Precision")
plt.plot(log_pre, label="Logistic Precision")
plt.title("Precision qua 10-fold")
plt.legend()
plt.show()

plt.figure(figsize=(10,5))
plt.plot(svm_rec, label="SVM Recall")
plt.plot(log_rec, label="Logistic Recall")
plt.title("Recall qua 10-fold")
plt.legend()
plt.show()

print("\n===== KẾT QUẢ LOGISTIC REGRESSION THEO TỪNG LẦN TRAIN =====")
for i in range(10):
    print(f"Fold {i + 1}: Accuracy={log_acc[i]:.4f}, Precision={log_pre[i]:.4f}, Recall={log_rec[i]:.4f}")

print("\n ===== KẾT QUẢ SVM THEO TỪNG LẦN TRAIN =====")
for i in range(10):
    print(f"Fold {i + 1}: Accuracy={svm_acc[i]:.4f}, Precision={svm_pre[i]:.4f}, Recall={svm_rec[i]:.4f}")

print("\n===== KẾT QUẢ TRUNG BÌNH =====")
print("SVM:", np.mean(svm_acc), np.mean(svm_pre), np.mean(svm_rec))
print("Logistic:", np.mean(log_acc), np.mean(log_pre), np.mean(log_rec))