Covid-19 is a new viral disease that spread in 2019 and turned into a pandemic over time. Due to its unknown nature, this disease caused a lot of human and financial losses in the current world. Several approaches were proposed to diagnose and apply medical care; One of the approaches that were more welcomed according to the results was the methods based on artificial intelligence. In the methods of artificial intelligence, various methods such as image processing, machine learning, etc. are used. Considering that in pandemic diseases, the number of patients is high and medical imaging is not without complications; Therefore, the use of clinical data and data mining techniques can be a suitable method in this field. But it should be kept in mind that in using data mining approaches, the lack of proper selection of features can hurt the analysis. Feature selection can be defined as the process of identifying relevant features and removing irrelevant and repetitive features to observe a subset of features that describe the problem well and with minimal loss of efficiency, and its purpose is to optimally select a subset of features with minimal redundancy and the maximum resolution ability. Therefore, the data mining process should pay attention to feature selection. In this article, a new method based on feature selection was proposed to increase the precision of the diagnosis of covid-19 disease. In the proposed method, we have used the fuzzy-chaotic forest optimization algorithm to select features that are effective on the covid-19 disease. This article has used four data sets collected by researchers of other scientific articles to evaluate the proposed method of diagnosing the disease of covid-19. The results of the evaluation in the important indicators of machine learning (precision, accuracy, recall and F1) show that in Comparing with similar algorithms, the proposed method provides better results and by increasing the accuracy of diagnosis, it can better diagnose the disease of Covid-19 In such a way, it improves the precision of diagnosis by 2% and reduces the dimensions of the diagnosis problem by 46%.