tree algorithm

 Bipolar disorder is a type of psychiatric disease characterized by periodic mood swings that include periods of depression and mania.

 First, three datasets related to bipolar disorder, including GSE53987, GSE35977, and GSE12679, were extracted from the PubMed database, which included 218 human samples and 9 888 458 genes. Then, genes directly related to bipolar disorder were extracted using R programming language. The shared genes were obtained from the database and extracted for 12 states with Cytoscape 3.7.1. The obtained gene expression data were trained by artificial neural network and decision tree method to identify the best models. Four parameters of sensitivity, specificity, accuracy, and area under the curve (AUC) were used to check the optimality of the model resulting from the training of machine learning algorithms.

 After R language preprocessing, 201 common genes were obtained. Then, 12 modes of 20 genes and 10 genes were extracted using the Cytohubba plugin in Cytoscape 3.7.1. The best model of 20 genes in the artificial neural network showed an AUC of 72% and the best model of 10 genes in the decision tree model showed an AUC of 78%.

 We presented two models to diagnose bipolar disorder. One model was developed using artificial neural network and tanh functions and the other model was developed using decision tree algorithm.

 The model developed by artificial neural network and the decision tree can be used in the diagnosis of bipolar disorder in order to screen conscripts who have this disorder with a high risk of relapse and exacerbation of symptoms.