Chagas disease is an endemic disease that in recent decades has ceased to be a rural disease to become mainly an urban disease. In this way, it currently constitutes a public health problem since 70 million people are at risk of contagion of this potentially fatal disease. This disease has an acute and a chronic phase, where in the latter it usually has cardiac involvement that can often be silent and asymptomatic at the beginning. As a result, the es- tablishment of early markers in this type of patients is of great interest. To achieve this, the present study proposes the analysis of RR data through permutation entropy and feature extraction. This study analyzes three groups: 83 volunteers (Control), 102 with Chagas but without cardiac involvement (CH1) and 107 with mild to moderate incipient heart failure (CH2). The data used is from the 24-hour ECG record- ing, RR intervals are shown in 288 5-minute frames. The analysis performed using permutation entropy and feature extraction shows significant differences between the 3 groups. These data, after a selection of significant segments and dimension reduction by means of PCA, were used in a densely connected neural network that has shown more than satisfactory results, obtaining 98% total accu- racy and precision greater than 97% when classifying each group, thus constituting a powerful tool for risk stratifica- tion and classification of patients.