Because of its fast-increasing prevalence, chronic kidney disease (CKD) is soon becoming a serious worry for the general public's health. This study's objective is to determine whether or whether machine learning methods are useful in the process of developing CRFs for chronic kidney disease (CKD) using the restricted number of clinical characteristics that are currently accessible. It has been determined via the use of several statistical procedures, including the analysis of variance, the Pearson correlation, and the Cramer's V test, that some features may be eliminated. For the purposes of training and evaluating logistic regression, support vector machines (SVMs), random forests, and gradient boosting, ten-fold cross-validation was used. When we use the Gradient Boosting classifier, we can get an accuracy of 99.1 percent using the F1-measure. In addition, we concluded that hemoglobin is a more reliable indicator of chronic kidney disease (CKD) than either random forest or gradient boosting. In conclusion, when compared to previous research, our results are among the most significant even though we have only accomplished a smaller number of characteristics so far. Because of this, the total cost of diagnosing CKD with all three tests is just $26.65. The rapidly increasing prevalence of chronic kidney disease (CKD) makes it a significant problem for the public's health. Throughout the course of this investigation, we want to test several machine learning algorithms to determine the extent to which they can diagnose chronic kidney disease based on a restricted number of clinical characteristics. Several statistical tests, including the ANOVA, the Pearson's correlation, and the Cramer's V test, have been carried out to get rid of features that aren't essential.