The Electroencephalogram (EEG), low cost and a non-invasive biomarker used for detection of neurological diseases such as Alzheimer’s disease. It helps in diagnosing the subject as normal or abnormal by measuring the electrical activity of the brain. Alzheimer’s disease (AD) is one of the most concerned neurodegenerative diseases. It is a difficult task for the physicians to interpret EEG signals by only visual inspection. By designing proper Machine Learning classifier a decision support system can be developed to support the physicians to detect AD with greater accuracy. Feature extraction is one of the most important step of any machine learning pipeline. This work presents the knowledge base creation method using empirical analysis of features extracted from channels of EEG signals as a novel study. Relative power, Hjorth parameter, six statistical features are presented as some selected sample features in this work. Analysis of results show that difference of range is there in the feature values of AD and a normal person in some of these parameters whereas overlapping of range is found in some features. Using visual inspection, approximate range for relative power in delta band of P3 channel can be identified as 0–0.4 for the normal control and 0.4–1.0 for AD. Whereas, the relative power in delta band of Fp1 channel showed no clear range. Additionally, it is observed that not all the channels of EEG signal and features are useful for the extraction of discriminating features for Alzheimer’s like brain disease detection.