assessment of a hybrid machine learning algorithm in healthcare management for predicting diabetes disease

Diabetes mellitus is one of the most chronic diseases in all over the world. every year, many people die due to this disease in all countries. therefore, identifying early detection methods for this disease can reduce its mortality. today, many diseases can be diagnosed and prevented from progressing by using data mining techniques and machine learning algorithms. in this paper, diabetes prediction has been aimed by comparing the efficiency of several classical machine-learning techniques. for this reason, for the sake of diabetes prediction algorithms such as naïve bayes, logistic regression (lr), multi-layer perceptron (mlp), sequential minimal optimization (smo), j48, random forest (rf), regression tree (rt) algorithms and a new hybrid algorithm based on multi-verse optimizer (mvo) and multi-layer perceptron (mlp) algorithms are employed for this evaluation based on accuracy (acc) indicator and area under curve (auc) criteria. numerous and diverse methods and algorithms have been used to predict diabetes. each of these algorithms has been effective in predicting diabetes with a different level of accuracy. our goal in this research is to introduce a new combined algorithm that has the highest level of accuracy in predicting diabetes compared to the old frequent algorithms so that it can help people in the timely treatment of this disease. in the structure of the mlp algorithm, the backpropagation algorithm is used for training. this article uses the mvo algorithm to train the mlp instead of the backpropagation algorithm, which built the hybrid algorithm called mvo-mlp. the accuracy results and the area under the roc diagram indicated that the proposed hybrid algorithm increases the accuracy by 107% compared to the mlp algorithm with the default structure. the outcomes of the accuracy of the new model are also higher than other algorithms used in this article