This repository contains code for a stroke prediction model based on the "Stroke Prediction" Dataset, it includes various features related to individuals health and lifestyle; whether they have experienced a stroke or not. our aim is to develop a model that can predict the likelihood of a stroke based on these features. It contains a number of features such as pertaining to people's health, lifestyle choices, and stroke history. The objective is to create a model that uses these characteristics to forecast the chance of a stroke.
https://www.kaggle.com/datasets/fedesoriano/stroke-prediction-dataset
According to the World Health Organization stroke is the second leading cause of death in the world, responsible for approximately 11% of deaths. This dataset is used to predict whether a patient is likely to get stroke based on the input parameters like gender, age, various diseases, and smoking status.
The following dependencies are required to run the code in this repository:
- Python
- Pandas
- Seaborn
- Scipy
- Numpy
- Matplotlib
- Scikit-learn
- Imbalanced-learn
- Xgboost
- Lightgbm
- Catboost
The code is organized into the following sections:
- Data Preprocessing: In this step, the code reads and combines the training and testing data, handles missing values, encodes categorical variables, and adds additional features.
- Feature Selection: The importance of features is determined using a Random Forest Classifier. The code selects the top-k features for further analysis.
- Model Training: Two models, Logistic Regression and XGBoost, are trained on the selected features. The models are evaluated on the validation set using various performance metrics, including accuracy, precision, recall, F1-score, ROC AUC score, and PR AUC score.
- Model Evaluation: The trained models are evaluated on the validation set, and the code displays their performance metrics.
- Model Testing: The final trained models, Logistic Regression and XGBoost, are used to make predictions on the test data. The code stores the predictions in a CSV file named "results_ronaldinho.csv".
- Ensemble Model: An ensemble model is implemented using a voting classifier. It combines the trained Logistic Regression, XGBoost, and LightGBM models. The code evaluates and displays the performance of the ensemble model.
- CatBoost Classifier: The code trains and evaluates a CatBoost Classifier on the validation set.
| Metric | Score |
|---|---|
| Accuracy | 0.8222 |
| Precision | 0.9562 |
| Recall | 0.7628 |
| F1-score | 0.8745 |
| ROC AUC score | 0.8769 |
| PR AUC score | 0.1875 |
| Kaggle Submission | 0.814 |
- Fatih Emir Güler
- Mert Erbak
- Mohammed Ammar Salahie