Cereberal-Stroke-Analysis

Overview of Dataset Processing Workflow

Data Loading and Import

• Use pandas, numpy, seaborn, matplotlib.pyplot, and other libraries to import and read the CSV file into a DataFrame (df).

Exploratory Data Analysis (EDA)

• Perform basic data exploration with head() and describe().
• Check and count missing values using isnull().sum().

Handling Categorical Variables

• Apply pd.get_dummies() for one‑hot encoding of categorical variables.

Handling Missing Values

• Fill missing values using the KNNImputer algorithm.

Feature Scaling and Train‑Test Split

• Perform feature scaling with MinMaxScaler.
• Split the dataset into training and testing sets.

Model Selection and Evaluation

• Conduct preliminary testing with models such as KNeighborsClassifier, GaussianNB, DecisionTreeClassifier, and RandomForestClassifier.
• Generate a classification report to evaluate model performance on the imbalanced dataset.

Data Resampling

• Apply SMOTE for oversampling.
• Perform random undersampling to balance class distribution.
• Use SMOTEENN to combine oversampling and undersampling.

Post‑Resampling Model Evaluation

• Retrain and evaluate models on the oversampled, undersampled, and combined sampled datasets.

Conclusion

• Various resampling techniques, especially SMOTEENN, substantially improve the model’s ability to identify positive stroke cases.