Abstract

Medical chatbots have become an important source of patient-generated symptom data, creating opportunities for machine learning-based healthcare analytics. This study evaluates the effectiveness of machine learning algorithms for predicting patient gender from textual symptom descriptions collected through a medical chatbot deployed in Chennai. A dataset containing approximately 7,000 patient symptom reports was used for analysis. The symptom text was preprocessed and transformed into numerical features using text vectorization techniques. The dataset was divided into training and testing subsets using an 80:20 ratio. Three classification algorithms k-Nearest Neighbors (k-NN), Multinomial Naive Bayes, and Multinomial Logistic Regression were implemented and compared. Model performance was assessed using accuracy, precision, recall, and F1-score, while confusion matrices and WordCloud visualization were used to examine classification behavior and symptom prevalence. Among the evaluated models, k-NN achieved the highest performance with an accuracy of 34.9% and a weighted F1-score of 0.34. The overall results indicate that gender prediction based solely on symptom descriptions remains a challenging task because of substantial overlap in symptom patterns across gender groups. The visualization results showed that symptoms such as muscle pain, fatigue, and diarrhea were among the most frequently reported complaints. Although the predictive performance was moderate, the study provides a comparative evaluation of lightweight machine learning models on real-world chatbot-generated symptom data and highlights the need for richer clinical features and advanced natural language processing techniques for improved demographic prediction in chatbot-assisted healthcare systems.

Authors

R. Lakshmi Priya¹, G. Manimannan², R. Arunadevi³, E. Babby⁴
Dr. Ambedkar Government Arts College, Vysarpadi, Chennai, India¹, St. Joseph’s College (Arts and Science), India^2,4, Vidhya Sagar Women’s College, India³

Keywords

Medical Chatbot, Symptom Analysis, Machine Learning, Gender Prediction, Natural Language Processing, Conversational Healthcare Data