Scribd
Upload
9 views

Untitled5.ipynb - Colab

The document outlines a data preprocessing and modeling workflow for a stroke prediction task using Python libraries such as pandas, scikit-learn, and imbalanced-learn. It includes steps for handling missing values, fixing outliers, encoding categorical variables, standardizing numerical features, and training a logistic regression model. The final model evaluation metrics indicate a strong performance, with an AUC score of approximately 0.89 and an F-beta score of 0.99.

Uploaded by

gacia der
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
9 views

Untitled5.ipynb - Colab

The document outlines a data preprocessing and modeling workflow for a stroke prediction task using Python libraries such as pandas, scikit-learn, and imbalanced-learn. It includes steps for handling missing values, fixing outliers, encoding categorical variables, standardizing numerical features, and training a logistic regression model. The final model evaluation metrics indicate a strong performance, with an AUC score of approximately 0.89 and an F-beta score of 0.99.

Uploaded by

gacia der
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

2/24/25, 9:55 PM Untitled5.

ipynb - Colab

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import os
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder, StandardScaler
from sklearn.impute import SimpleImputer
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import fbeta_score, roc_auc_score, classification_report, confusion_matrix
from imblearn.over_sampling import SMOTE

from google.colab import files


uploaded = files.upload()

Choose Files 2 files


test.csv(text/csv) - 152301 bytes, last modified: 2/21/2025 - 100% done
train.csv(text/csv) - 146010 bytes, last modified: 2/21/2025 - 100% done
Saving test.csv to test.csv
Saving train.csv to train.csv

train_path = "/content/train.csv"
test_path = "/content/test.csv"
import os
print(os.path.exists(train_path))
print(os.path.exists(test_path))
train_path = "train.csv"
test_path = "test.csv"
train_df = pd.read_csv(train_path)
test_df = pd.read_csv(test_path)

True
True

# 1️⃣ Handling Missing Values


num_features = ["age", "avg_glucose_level", "bmi"]
num_imputer = SimpleImputer(strategy="median")
train_df[num_features] = num_imputer.fit_transform(train_df[num_features])
test_df[num_features] = num_imputer.transform(test_df[num_features])

# Handling categorical missing values separately


cat_features = ["gender", "ever_married", "work_type", "Residence_type", "smoking_status"]
cat_imputer = SimpleImputer(strategy="most_frequent")
train_df[cat_features] = cat_imputer.fit_transform(train_df[cat_features])
test_df[cat_features] = cat_imputer.transform(test_df[cat_features])

# Dropping missing target values (after checking distribution)


train_df = train_df.dropna(subset=["stroke"])

# 2️⃣ Fixing Outlier Issues


# Fixing Unrealistic Age Values (Remove Ages > 120)
train_df = train_df[train_df["age"] <= 120]

def cap_outliers(df, feature, lower_quantile=0.01, upper_quantile=0.99):


lower_cap = df[feature].quantile(lower_quantile)
upper_cap = df[feature].quantile(upper_quantile)
df[feature] = np.clip(df[feature], lower_cap, upper_cap)
return df

# Apply capping to BMI separately for train and test


test_df = cap_outliers(test_df, "bmi")
https://colab.research.google.com/drive/1KLU1JQ7o79yiRs_6lvFPQ1VSR6Q-44d_#scrollTo=THyU-c79XIFw&printMode=true 1/4
2/24/25, 9:55 PM Untitled5.ipynb - Colab
train_df = cap_outliers(train_df, "bmi")

# Log transform avg_glucose_level separately per set


train_df['avg_glucose_level'] = np.log1p(train_df['avg_glucose_level'])
test_df['avg_glucose_level'] = np.log1p(test_df['avg_glucose_level'])

# 3️⃣ Handling "Unknown" in smoking_status


train_df['smoking_status_unknown'] = train_df['smoking_status'].eq('Unknown').astype(int) if 'smoking_status' in train
test_df['smoking_status_unknown'] = test_df['smoking_status'].eq('Unknown').astype(int) if 'smoking_status' in test_df
train_df.drop(columns=["smoking_status"], errors='ignore', inplace=True)
test_df.drop(columns=["smoking_status"], errors='ignore', inplace=True)

# 3️⃣ Handling "Unknown" in smoking_status


train_df['smoking_status_unknown'] = train_df['smoking_status'].eq('Unknown').astype(int) if 'smoking_status' in train
test_df['smoking_status_unknown'] = test_df['smoking_status'].eq('Unknown').astype(int) if 'smoking_status' in test_df
train_df.drop(columns=["smoking_status"], errors='ignore', inplace=True)
test_df.drop(columns=["smoking_status"], errors='ignore', inplace=True)

# Update cat_features after dropping smoking_status


cat_features = ["gender", "ever_married", "work_type", "Residence_type"] # Removed smoking_status

# 4️⃣ Encoding Categorical Variables


encoder = OneHotEncoder(drop='first', sparse_output=False, handle_unknown='ignore')
encoded_train = pd.DataFrame(encoder.fit_transform(train_df[cat_features]), columns=encoder.get_feature_names_out())
encoded_test = pd.DataFrame(encoder.transform(test_df[cat_features]), columns=encoder.get_feature_names_out())

# Transform test data


encoded_test = pd.DataFrame(encoder.transform(test_df[cat_features]))
encoded_test.columns = encoder.get_feature_names_out(cat_features) # Get column names

# Reset index
encoded_train.reset_index(drop=True, inplace=True)
encoded_test.reset_index(drop=True, inplace=True)

# 5️⃣ Standardizing Numerical Features


scaler = StandardScaler()
scaled_train = pd.DataFrame(scaler.fit_transform(train_df[num_features]), columns=num_features)
scaled_test = pd.DataFrame(scaler.transform(test_df[num_features]), columns=num_features)

# 6️⃣ Combining Processed Features


X_train_final = pd.concat([scaled_train, encoded_train, train_df[["hypertension", "heart_disease", "smoking_status_unk
X_test_final = pd.concat([scaled_test, encoded_test, test_df[["hypertension", "heart_disease", "smoking_status_unknown

# Handle potential NaN values in 'stroke' column before conversion


y_train_final = train_df["stroke"].map({'Yes': 1, 'No': 0}).fillna(0).astype(int) # Fill NaN with -1 or another suitab

# 7️⃣ Handling Class Imbalance Using SMOTE


smote = SMOTE(random_state=42, k_neighbors=min(5, y_train_final[y_train_final != -1].value_counts().min() - 1))
# Set k_neighbors to the minimum between 5 and the number of samples in the smallest class(excluding -1 if present), m
X_train_balanced, y_train_balanced = smote.fit_resample(X_train_final, y_train_final)

# 8️⃣ Train-Test Split (Only Once!)


X_train, X_val, y_train, y_val = train_test_split(X_train_balanced, y_train_balanced, test_size=0.2, random_state=42,

# 9️⃣ Train Logistic Regression Model


model = LogisticRegression(class_weight='balanced', max_iter=1000)
model.fit(X_train, y_train)

https://colab.research.google.com/drive/1KLU1JQ7o79yiRs_6lvFPQ1VSR6Q-44d_#scrollTo=THyU-c79XIFw&printMode=true 2/4
2/24/25, 9:55 PM Untitled5.ipynb - Colab

▾ LogisticRegression i ?

LogisticRegression(class_weight='balanced', max_iter=1000)

# Predictions
y_pred = model.predict(X_val)
y_probs = model.predict_proba(X_val)[:, 1]

# Evaluation Metrics
auc_score = roc_auc_score(y_val, y_probs)
f_beta = fbeta_score(y_val, y_pred, beta=10)
class_report = classification_report(y_val, y_pred)
conf_matrix = confusion_matrix(y_val, y_pred)

# Display Metrics
print(f"AUC Score: {auc_score}")
print(f"F-beta Score (β=10): {f_beta}")
print("Classification Report:")
print(class_report)
print("Confusion Matrix:")
print(conf_matrix)

AUC Score: 0.8933009746138141


F-beta Score (β=10): 0.9974195801482286
Classification Report:
precision recall f1-score support

0 1.00 0.74 0.85 510


1 0.79 1.00 0.88 509

accuracy 0.87 1019


macro avg 0.90 0.87 0.87 1019
weighted avg 0.90 0.87 0.87 1019

Confusion Matrix:
[[377 133]
[ 0 509]]

# 1️⃣0️⃣ Prepare Test Data for Submission


test_probs = model.predict_proba(X_test_final)[:, 1]
test_preds = model.predict(X_test_final)

# Save Processed Data


train_df.to_csv("train_cleaned.csv", index=False)
test_df.to_csv("test_cleaned.csv", index=False)
submission_df = pd.DataFrame({"id": test_df["id"], "stroke": test_preds})
submission_df.to_csv("submission.csv", index=False)

print("Preprocessing and training complete! 🚀")


Preprocessing and training complete! 🚀

https://colab.research.google.com/drive/1KLU1JQ7o79yiRs_6lvFPQ1VSR6Q-44d_#scrollTo=THyU-c79XIFw&printMode=true 3/4
2/24/25, 9:55 PM Untitled5.ipynb - Colab

https://colab.research.google.com/drive/1KLU1JQ7o79yiRs_6lvFPQ1VSR6Q-44d_#scrollTo=THyU-c79XIFw&printMode=true 4/4

You might also like