CLABSI Prediction Project

Overview

This project focuses on predicting Central Line-Associated Bloodstream Infections (CLABSI) using machine learning models. The goal was to analyze patterns in clinical data, address challenges like class imbalance, and develop robust predictive models to support healthcare decision-making.

---

Key Features

1. Data Preprocessing

• Imputation: Handled missing values using mean and median imputation.
• Scaling: Standardized numerical features to enhance model performance.
• Feature Selection: Applied SelectKBest to identify top predictors.
• Encoding: Used one-hot encoding for categorical variables.

2. Handling Class Imbalance

• Implemented SMOTE (Synthetic Minority Oversampling Technique) to balance the dataset.
• Evaluated performance improvements with balanced training data.

3. Model Development

Developed and compared the following models:

• Logistic Regression: Baseline model, optimized with hyperparameter tuning.
• K-Nearest Neighbors (KNN): Improved performance with SMOTE and hyperparameter tuning.
• Decision Trees: Enhanced with bootstrapping to improve reliability.
• Neural Networks: Achieved high recall for detecting CLABSI cases.
• XGBoost: Applied hyperparameter tuning but struggled with class imbalance.

4. Evaluation Metrics

• Used metrics such as accuracy, precision, recall, F1-score, and AUC to assess model performance.
• Prioritized recall to minimize false negatives and avoid missed diagnoses.

---

Results

Model	Accuracy	FPR (Type I)	FNR (Type II)	TPR	TNR	AUC
KNN	95.54%	4.4%	8.3%	91.7%	95.6%	0.94
Decision Trees	99.98%	0%	0.1%	100%	99.9%	0.99
XGBoost	97.27%	2.4%	90%	10%	97.6%	0.66
Neural Networks	97.55%	4.87%	0%	100%	95.1%	0.97

---

Technologies Used

• Programming Language: Python
• Libraries:
  • Data Preprocessing: Pandas, NumPy, Scikit-learn
  • Oversampling: Imbalanced-learn (SMOTE)
  • Machine Learning: TensorFlow, Keras, XGBoost
  • Visualization: Matplotlib, Seaborn

---

How to Run

1. Clone the repository:

   git clone <[repository-link](https://github.com/SanthoshBotcha/CLABSI_Modeling_and_Prediction.git)>

2. Install dependencies:

   pip install -r requirements.txt

3. Run the Jupyter Notebook:

   jupyter notebook clabsi_prediction.ipynb

---

Key Learnings

• Addressing Class Imbalance: SMOTE significantly improved model sensitivity for minority class detection.
• Model Selection: Neural Networks and KNN (with SMOTE) performed best for this dataset.
• Evaluation Trade-offs: Prioritized recall to reduce false negatives, critical in medical diagnostics.

---

Next Steps

• Use a holdout dataset to validate generalization performance.
• Experiment with cost-sensitive learning to reduce false positives.
• Incorporate explainability tools like SHAP or LIME.

---

Author

Santhosh Botcha
Academic - Business Analysis Project