Application of the Minkowski Distance Similarity Method in Case-Based Reasoning for Stroke Diagnosis

Abstract

A Stroke is a cerebrovascular disease characterized by impaired brain function due to damage or death of brain tissue caused by reduced or blocked blood and oxygen flow to the brain. Expert systems can be used as learning aids for medical students to diagnose stroke. Medical records of stroke cases can be reused as a reference for diagnosing stroke when there are new cases, known as the case-based reasoning (CBR) method. This study implements the Minkowski distance similarity method in CBR to calculate the similarity value between cases, where each similar case has the same solution. This study uses the Minkowski distance similarity method in CBR to obtain the most optimal value of r and the most appropriate threshold value in the expert system for stroke diagnosis. The diagnosis process is carried out by inputting the patient's condition, symptoms, and risk factors. Then the system will calculate the similarity value and take the case with the highest similarity value as the solution, providing that the similarity value must be greater than or equal to the threshold value. Based on system testing, the best accuracy value was achieved by applying a threshold value of 75 with an r value of 3 or 4, with an accuracy rate of 88,89%, a recall value of 88%, and a precision of 100%.

Comparison of the Application of Weighted Cosine Similarity and Minkowski Distance Similarity Methods in Stroke Diagnostic Systems

“…The system demonstrated a sensitivity level of 89.88% and accuracy of 81.67% with indexing and 84.44% without indexing. Further research was carried out using the same dataset with the Minkowski Distance similarity calculation method [27]. The research results show that Minkowski Distance provides a better accuracy rate of 88.89% compared to the Jaccard Coefficient method.…”