Infection Classification and Event Detection System

This repository implements an end-to-end machine learning pipeline for classifying infections using structured patient data, time-series vitals, and event-based text sequences. It integrates LSTM-based deep learning with a domain-aware event generator to support predictive modeling in medical scenarios, such as sepsis detection.

---

📌 Project Highlights

• 🔍 Event-Driven Data Modeling
  Converts time-series patient vitals into structured event chains based on clinically significant changes.

• 🧠 LSTM-Based Classification Model
  A deep learning model (with Embedding + Bidirectional LSTM) trained on textual event sequences and demographic features.

• ⚙️ TPU/GPU Compatible
  Automatically detects and utilizes TPU for faster training if available.

• 🧪 Real-Time Testing & Evaluation
  Includes confusion matrix, classification report, and prediction on new sample cases.

---

🔧 Execution Flow

1. Infection Classification with Deep Learning

• Load and preprocess the dataset (infection_dataset_final_balanced.csv)
• Tokenize the Sequence feature representing clinical events
• Combine tokenized sequences with numerical data (Age, Gender, Health Condition)
• Encode infection labels and apply class weighting
• Train a deep learning model with the following architecture:
  • Embedding → BiLSTM → BatchNorm → Dropout → BiLSTM → Dense Layers
• Evaluate on test data and visualize results

2. Synthetic Event Sequence Generator for Sepsis Simulation

• generate_sepsis_dataset(n=20) simulates vitals over time (e.g., HRV, BP, Temp, SpO2)
• detect_events(patient_records) detects significant clinical changes across time intervals
• Transforms sequential vital changes into a structured Event Chain, e.g.:
  "High HR Variability (Short) + Low BP (Short) → High Body Temp (Long) + Low SpO2 (Short)"

This text sequence is compatible with the model’s tokenizer and can be used for inference or further training.

---

🚀 How to Run

1. Place your dataset (infection_dataset_final_balanced.csv) in the project directory.
2. Run the infection classification model script to train and evaluate:

   python infection_model.py