Electroencephalogram (EEG) in the healthcare Internet of Things (HIoT) plays a crucial role in collecting and analyzing human brain signals. In the last decade, near-edge computing has emerged as a solution to the time-critical nature of EEG healthcare systems. However, EEG sensors generate large amounts of continuous signals from the human brain, leading delays in transmitting data to the edge layer. To address this issue, the authors propose a machine learning-based efficient lossless data compression technique in a HIoT network. The proposed technique comprises two stages: In stage 1, EEG data files are clustered using k-means clustering, and in stage 2, run-length encoding is applied to each cluster to compress the data. The compressed data is then transmitted to the edge layer for processing. The proposed technique is evaluated using three distinct datasets (Bonn University, Physionet Motor Movement/Imagery, and Physionet Sleep Telemetry) and compared with various existing techniques. The results show that this work achieves a maximum of \(97 \%\) compression power for dataset-1 across various subjects and outperforms all other techniques with more than the doubled compression ratio for dataset-2. Additionally, the work is tested with dataset-3 containing larger files, achieving a good compression power of maximum \(95 \%\) . The proposed technique is efficient for healthcare centers with multiple EEG systems in the HIoT network.