A Combination of Data Mining Methods for Disease Classification Using Patient-Perceived Symptoms from Medical Records

One of the problems faced by healthcare facilities is the influx of patients every day, resulting in long queues and decreased public satisfaction. Moreover, the growing volume of medical record data remains underutilized, despite its potential to provide valuable insights. This paper proposes the development of a disease classification model that combines classification algorithms with text processing techniques. We analyzed patient medical records from a healthcare facility collected between August and September 2023. Text preprocessing techniques were employed to clean and extract meaningful data from medical records, with bag of words and TF-IDF methods used to convert the text into numerical feature suitable for classification. We compared the performance of two classification algorithms: Naïve Bayes and K-Nearest Neighbor, each combined with bag of words and TF-IDF. The result indicates that the KNN and bag of words achieved the highest accuracy of $75 \%$, followed by Naïve Bayes and bag of words at $\mathbf{7 4 \%}$. Naïve Bayes with TF-IDF and KNN and TF-IDF achieved the accuracy of $73 \%$. This classification aims to assist doctors by providing an overview of patient illnesses, highlighting frequently occurring diseases and those requiring further examination. The findings underscore the potential of combining text processing and classification algorithms to optimize healthcare services and improve patient satisfaction.