Context: Chagas disease (CD) is one of the 17 neglected tropical diseases according to the World Health Organization, affecting approximately 8 million people worldwide. This disease has been identified as one of the main causes of death due to cardiovascular problems, manifested through cardiac arrhythmias, heart failure and thromboembolisms. Echocardiography (ECO) provides important diagnostic tools, mainly considering left venticular systolic dysfunction (LVSD). However, ECO is more expensive, complex and difficult to access than electrocardiography (ECG). Objective: Investigate strategies for extracting ECG signal parameters for predicting LVSD defined as a left ventricular ejection fraction (LVEF) determined by ECO ≤ 40%. Data description: We process a dataset containing ECG holter signals from 219 CD patients obtained from University Hospital Clementino Fraga Filho –Federal University of Rio de Janeiro, Rio de Janeiro, Brazil. The local ethics committee approved the research (number 45360915.1.1001.5262). Data Processing: Initially, we combine Wavelet and Hilbert transforms for detecting and delineating each QRS complex, P-wave and T-wave. Then, we consider the segmentation of the original holter signals in intervals during: (a) 10 minutes; (b) 15 minutes and (c) 30 minutes. For each specific scenario, we obtain statiscal measures related to intervals between fiducial points: R-R, Q-Tend, Q-Tpeak, Tend-Q, Tpeak-Q. Then, we apply a set of Machine Learning (ML) algorithms for each scenario for discriminating between LVSD patients and non-LVSD patients. Results: As results, we obtain the highest performance for 30-minute ECG intervals. Among the applied ML techniques, we achieve the highest results accuracy 80,01% +- 4,55% and Area under ROC curve 0,76 +- 0,06 for Logistic Regression, and accuracy 79,14% +- 5% and Area under ROC curve 0,77 +- 0,06 for Multilayer Perceptron. Conclusion: The results indicate the feasibility of using short-term one-channel ECG signals to predict reduced LVEF.