#General Purpose
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

#Data transform
from sklearn.utils import resample
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
from sklearn.preprocessing import LabelEncoder
from sklearn.preprocessing import MinMaxScaler
from sklearn.preprocessing import Normalizer
from sklearn.manifold import TSNE

#Model
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.tree import export_text

#Metrics
from sklearn.metrics import accuracy_score, classification_report
from sklearn.metrics import f1_score, confusion_matrix, matthews_corrcoef

# Sample Datasets
from sklearn.datasets import load_iris

sns.set_theme()

iris         = load_iris()
iris_data    = iris.data
iris_target  = iris.target
iris_targ_n  = iris['target_names']
iris_targ_n  = iris_targ_n[iris.target]
iris_feat_n  = iris['feature_names']

df = pd.DataFrame(iris_data, columns=iris_feat_n)
df['target'] = iris_targ_n
df.head()

axes = df.boxplot(column=[iris_feat_n[0], iris_feat_n[2], iris_feat_n[1], iris_feat_n[3]],
                  by='target', figsize=(10, 10), fontsize=15)

np.random.seed(2021)

tsne = TSNE(n_components=2, verbose=0, perplexity=40, n_iter=300)
tsne_results = tsne.fit_transform(iris_data)

df['tsne0'] = tsne_results[:, 0]
df['tsne1'] = tsne_results[:, 1]

plt.figure(figsize=(10,10))
sns.scatterplot( x="tsne0", y="tsne1", hue="target",
    palette=sns.color_palette("hls", 3), data=df,
    legend="full", alpha=1);

decision_tree = DecisionTreeClassifier(random_state=2021, max_depth=3)
decision_tree = decision_tree.fit(iris_data, iris_target)
r = export_text(decision_tree, feature_names=iris['feature_names'])

np.random.seed(2021)
has_cancer = (np.random.random(size=1000) > 0.9).astype(int)
predicted  = (np.random.random(size=1000) > 0.99).astype(int)#np.zeros(shape=1000, dtype=int)

cm = confusion_matrix(has_cancer, predicted)

df_cm = pd.DataFrame(cm, index = ["Actual Positive", "Actual Negative"],
                      columns = ["Predict Positive", "Predict Negative"])
plt.figure(figsize = (8,8))
axes = sns.heatmap(df_cm, annot=True, fmt='g', annot_kws={"size": 25})
plt.show()

accuracy = accuracy_score(has_cancer, predicted)
f1       = f1_score(has_cancer, predicted)
mcc      = matthews_corrcoef(has_cancer, predicted)
print(f"Results: Accuracy {accuracy:0.3f}\n         F1       {f1:0.3f}",
      f"\n         MCC     { mcc:0.3f}")

    Results: Accuracy 0.906
             F1       0.021
             MCC     -0.000

data = pd.read_csv('../data/kaggle-titanic-train.csv')
data.head(10)

data_filter_col = data[['Pclass', 'Sex', 'Age', 'SibSp', 'Parch', 'Fare', 'Survived']]
data_filter_col.head()

data_filter_col['Sex'].replace(['female','male'],[1,0],inplace=True)
data_filter_col.isnull().sum()

    Pclass        0
    Sex           0
    Age         177
    SibSp         0
    Parch         0
    Fare          0
    Survived      0
    dtype: int64

data_filter_col = data_filter_col.dropna()

data_filter_col.shape
(714, 7)

X = data_filter_col[['Pclass', 'Sex', 'Age', 'SibSp', 'Parch', 'Fare']]
y = data_filter_col['Survived']

X.head(2)

corr = data_filter_col.corr()
corr.style.background_gradient(cmap='coolwarm')

X.describe()

X_train, X_val, y_train, y_val = train_test_split(
                    X, y, test_size=0.30, random_state=2021)

tree = DecisionTreeClassifier()

tree.fit(X_train, y_train)

predicted_val = tree.predict(X_val)
predicted_train = tree.predict(X_train)

acc_train = accuracy_score(y_train, predicted_train)
acc_val = accuracy_score(y_val, predicted_val)
print(f"Accuracy: Train {acc_train:0.3f} | Val {acc_val:0.3f}")

report = classification_report(y_val, predicted_val)
print(report)

    Accuracy: Train 0.990 | Val 0.744

                  precision    recall  f1-score   support

               0       0.78      0.79      0.78       126
               1       0.69      0.69      0.69        89

        accuracy                           0.74       215
       macro avg       0.74      0.74      0.74       215
    weighted avg       0.74      0.74      0.74       215

knn = KNeighborsClassifier(n_neighbors=5)

knn.fit(X_train, y_train)

predicted_val = knn.predict(X_val)
predicted_train = knn.predict(X_train)

acc_train = accuracy_score(y_train, predicted_train)
acc_val = accuracy_score(y_val, predicted_val)
print(f"Accuracy: Train {acc_train:0.3f} | Val {acc_val:0.3f}")

report = classification_report(y_val, predicted_val)
print(report)

    Accuracy: Train 0.796 | Val 0.656
                  precision    recall  f1-score   support

               0       0.69      0.75      0.72       126
               1       0.59      0.53      0.56        89

        accuracy                           0.66       215
       macro avg       0.64      0.64      0.64       215
    weighted avg       0.65      0.66      0.65       215

norm = Normalizer()
X_train = norm.fit_transform(X_train)
X_val   = norm.transform(X_val)

knn = KNeighborsClassifier(n_neighbors=5)
knn.fit(X_train, y_train)

predicted_val = knn.predict(X_val)
predicted_train = knn.predict(X_train)

acc_train = accuracy_score(y_train, predicted_train)
acc_val = accuracy_score(y_val, predicted_val)
print(f"Accuracy: Train {acc_train:0.3f} | Val {acc_val:0.3f}")

report = classification_report(y_val, predicted_val)
print(report)

    Accuracy: Train 0.820 | Val 0.721
                  precision    recall  f1-score   support

               0       0.74      0.82      0.77       126
               1       0.69      0.58      0.63        89

        accuracy                           0.72       215
       macro avg       0.71      0.70      0.70       215
    weighted avg       0.72      0.72      0.72       215