Chagas disease is caused by the protozoan parasite Trypanosoma cruzi. It is estimated that in the world there are between 6 and 7 million people infected with it, mainly in endemic areas of 21 Latin American countries, where it is transmitted to humans through the feces or urine of triatomine insects. Chagas is slowly becoming a health problem in more urban areas and countries. That is the reason why it is convenient to develop diagnosis methods for it. In this work, we used a multilayer perceptron neural network to classify 292 subjects (volunteers and patients) composed of 83 volunteers (Control group); 102 asymptomatic chagasic patients with a positive Machado-Gerreiro serological test (CH1 group); 107 seropositive chagasic patients with incipient heart disease: first degree atrioventricular block, sinus bradycardia or right bundle branch block of His (CH2 group). The three groups underwent clinical evaluation test, chest X-ray, echocardiogram, electrocardiogram and Holter, not all patients received medication. The Approximation Entropy ApEn was calculated from the tachograms of the circadian profiles of 24 hours every 5 minutes (288 frames). The data of these frames was used to train the proposed architecture. The classification work done by the neural network had 94% of accuracy and 94% of precision indicating a good performance of the classifier, validated with the ROC curve where all AUC values are close to the unit (0.98 for control group, 0.99 for CH1 and 0.98 for CH2). Taking into account the good results we can consider this artificial neural network and approximation entropy as a useful tools to obtain early markers for the diagnosis of the cardiac compromise related to the illness.