Symptom NLP Classifier

NLP pipeline for classifying medical symptom descriptions into disease categories. Built as an extension of healthcare analytics work at Ayurnidaan (BITS Pilani incubated startup).

Pipeline

Raw symptom text → Preprocessing → TF-IDF (bigrams) → Classifier → Disease category

Models compared

Model	Accuracy	F1 (weighted)
Logistic Regression	~0.98	~0.98
Random Forest	~0.97	~0.97
Gradient Boosting	~0.96	~0.96

How to run

pip install numpy scikit-learn matplotlib
python symptom_classifier.py

Usage as API

from symptom_classifier import predict_symptom, build_pipeline

pipeline = build_pipeline('LogisticRegression')
pipeline.fit(X_train, y_train)

result = predict_symptom("fever chills body ache", pipeline)
print(result['predicted'])     # e.g. "flu"
print(result['confidence'])    # e.g. 0.94

Real datasets to plug in

• MIMIC-III clinical notes (requires credentialing)
• Symptom-Disease dataset on Kaggle
• MedQuAD medical Q&A dataset

Skills demonstrated

Python, NLP, TF-IDF, Scikit-learn Pipeline, text classification, prediction API