Application of the Logistic Regression Algorithm for the Diagnosis of Stomach Disease

Logistic regression is a widely utilized machine learning algorithm for binary classification tasks. In this study, the logistic regression algorithm is used to classify whether a disorder can be diagnosed as a stomach disease. The levels of four test indicators, ceruloplasmin, blue reaction, indole-acetic acid, and neutral sulfides, are considered as features. Eleven data sets in the training set and four data sets in the testing set are employed. The data sets have been desensitized. The predicted classifications for the testing set align with the true classifications. Feature importance analysis is also performed. The first three test indicators significantly influence classification, each contributing an average of over 30%, whereas the neutral sulfides level has minimal impact, accounting for less than 1%. Therefore, the level of neutral sulfides alone is not regarded as a key indicator for diagnosing stomach disease. This study not only offers insights for clinical diagnosis but also highlights the potential of machine learning algorithms in medical applications.