Sleep is important part for human health and quality of life in the daily routine basis. However, numerous individuals face sleep problems due to rapid changes occurred in both social and professional lifestyles. These problems can lead to several neurological and physical disorder diseases, and therefore, decrease their overall life quality. Machine learning methods for automated sleep stage classification (ASSC) are a fundamental approach to evaluate and treat this public health challenge. The main objective of this study is to propose a high-effective and high-accuracy based multiple sleep staging classification model based on single-channel electroencephalogram (EEG) signals using machine learning (ML) model. The proposed automated sleep staging system followed four basic stages: signal preprocessing, feature extraction and screening, classification algorithms, and performance evaluation. In this research work, a novel method is applied for signal preprocessing, feature screening and classification models. In signal preprocessing we obtain the wiener filter techniques for removing the different types of artifacts from input sleep recordings. In feature extraction, we obtain a total of 28 features based on both time and frequency domain features and non-linear features. The relevant features are screened through ReliefF weight feature selection algorithm, and eliminating the redundant features using Pearson correlation coefficients. The important contribution of this research work is establishes two layers an ensembling learning stacking model for classifying the multiple sleep stages. Three different subgroups of ISRUC-Sleep (SG-I/SG-II/SG-III) subjects sleep recordings having different health condition obtained for our proposed experimental work. Comparing with the recent contributions on sleep staging performances using single-channel EEG signals, it has found that our proposed ensemble learning stacking model was reported excellent in sleep staging classification accuracy performance for five sleep stages classification task (CT-5). The overall classification accuracy reported as 99.34%, 90.8%, and 98.50% for SG-I, SG-II, and SG-III dataset, respectively.