assignment 3

1)
import numpy as np
import pandas as pd

q1)a
def answer_one():
    
    data = pd.read_csv('assets/fraud_data.csv')
    result = len(data[data.Class == 1])/len(data)
    
    return result

answer_one()

q1)b

# Use X_train, X_test, y_train, y_test for all of the following questions
from sklearn.model_selection import train_test_split

df = pd.read_csv('assets/fraud_data.csv')

X = df.iloc[:,:-1]
y = df.iloc[:,-1]

X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)


q2)
def answer_two():
    from sklearn.dummy import DummyClassifier
    from sklearn.metrics import recall_score
    
    # Negative class (0) is most frequent
    dummy_majority = DummyClassifier(strategy = 'most_frequent').fit(X_train, y_train)
    # Therefore the dummy 'most_frequent' classifier always predicts class 0
    y_dummy_predictions = dummy_majority.predict(X_test)
    
    acc_score = dummy_majority.score(X_test, y_test)
    rec_score = recall_score(y_test, y_dummy_predictions) 
    
    return acc_score, rec_score
answer_two()

q3)
def answer_three():
    from sklearn.metrics import recall_score, precision_score
    from sklearn.svm import SVC

    svm = SVC().fit(X_train, y_train)
    y_pred = svm.predict(X_test)
    
    acc_score = svm.score(X_test, y_test)
    rec_score = recall_score(y_test, y_pred)
    prec_score = precision_score(y_test, y_pred)
    
    return acc_score, rec_score, prec_score
answer_three()

q4)
def answer_four():
    from sklearn.metrics import confusion_matrix
    from sklearn.svm import SVC

    THRESHOLD = -220.0
    params = {'C': 1e9, 'gamma': 1e-07}
    y_score = SVC(**params).fit(X_train, y_train).decision_function(X_test)
    y_score_adj = np.zeros_like(y_score)
    y_score_adj[np.where(y_score >= THRESHOLD)] = 1
       
    confusion = confusion_matrix(y_test, y_score_adj)
    
    return confusion

answer_four()

q5)
def answer_five():
    from sklearn.linear_model import LogisticRegression
    from sklearn.metrics import precision_recall_curve
    from sklearn.metrics import roc_curve, auc
        
    lr = LogisticRegression()
    y_proba_lr = lr.fit(X_train, y_train).predict_proba(X_test)
    y_scores_lr = lr.fit(X_train, y_train).decision_function(X_test)
    precision, recall, thresholds = precision_recall_curve(y_test, y_scores_lr)
    fpr_lr, tpr_lr, _ = roc_curve(y_test, y_scores_lr)
    
    rec = float(recall[np.where(precision==0.75)])
    tpr = float(tpr_lr[np.where((fpr_lr >= 0.159) & (fpr_lr <= 0.161))][0])
    
    return rec, tpr

answer_five()

q6)a
def answer_six():    
    from sklearn.model_selection import GridSearchCV
    from sklearn.linear_model import LogisticRegression

    grid_values = {'penalty': ['l1', 'l2'], 'C':[0.01, 0.1, 1, 10]}
    
    lr = LogisticRegression()
    grid_lr = GridSearchCV(lr, param_grid = grid_values, scoring = 'recall')
    grid_lr.fit(X_train, y_train)
    result = grid_lr.cv_results_['mean_test_score'].reshape(-1, 2)
    
    return result

answer_six()

q6)b
# Use the following function to help visualize results from the grid search
def GridSearch_Heatmap(scores):
    %matplotlib notebook
    import seaborn as sns
    import matplotlib.pyplot as plt
    plt.figure()
    sns.heatmap(scores.reshape(4,2), xticklabels=['l1','l2'], yticklabels=[0.01, 0.1, 1, 10])
    plt.yticks(rotation=0);

GridSearch_Heatmap(answer_six())