douglaspsteen

import numpy as np
import pandas as pd

# Visualization
import matplotlib.pyplot as plt

# Data processing, modeling, and model evaluation
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA
from sklearn.neighbors import KNeighborsClassifier

douglaspsteen

# Load and check data

df = pd.read_csv('data.csv')
display(df.head())

douglaspsteen

# Drop first column of id information

df = df.drop('Unnamed: 0', axis=1)

# Re-cast target variable (y) as either having a seizure (1) and all else (0)

for i in range(len(df)):
    
    if df.iloc[i]['y'] != 1:
        df.at[i, 'y'] = 0

douglaspsteen

# Fill null values with each column mean

df = df.fillna(df.mean())

# Define X and y

X = df.drop('y', axis=1)
y = df.y

# Visualize class distribution

douglaspsteen

# Standardize data

scaler = StandardScaler()
X = scaler.fit_transform(X)

douglaspsteen

# PCA

pca = PCA(n_components=178)
pca.fit(X)
X_pca = pca.transform(X)

douglaspsteen

# Calculate cumulative explained variance across all PCs

cum_exp_var = []
var_exp = 0
for i in pca.explained_variance_ratio_:
    var_exp += i
    cum_exp_var.append(var_exp)

# Plot cumulative explained variance for all PCs

douglaspsteen

train_f1 = []
test_f1 = []

for i in range(20):
    
    X = X_pca[:,0:i+1]
    
    # Train-test-split
    X_train, X_test, y_train, y_test = train_test_split(X, y,
                                                        test_size=0.25,

douglaspsteen

X = X_pca[:,0:4]
    
# Train-test-split
X_train, X_test, y_train, y_test = train_test_split(X, y,
                                                    test_size=0.25,
                                                    random_state=42)

# Perform feature scaling
scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)

douglaspsteen

plot_confusion_matrix(knn, X_test, y_test, display_labels=['No Seizure',
                                                          'Seizure'],
                     normalize='true', cmap='Blues');