Harvesting ambient energy has enabled the development of energy-harvesting wireless sensor networks (EH-WSNs). However, in these networks, the uncertainty in harvesting rate due to dynamic weather conditions raises new challenges. Therefore, this drives the development of energy harvesting-aware solutions. Formerly, many MAC protocols have been developed for EH-WSNs, which offer various features based on available harvested energy to support different applications. Nevertheless, optimizing MAC performance by incorporating predicted future energy intake is relatively new in EH-WSNs. Therefore, this thesis presents a machine learning prediction based adaptive duty cycle medium access control (MAC) protocol for solar energy harvesting wireless sensor networks WSNs. The developed protocol incorporates information about the current and future harvested energy using mathematical formulations to improve network performance. By doing so, the proposed MAC protocol effectively addresses the primary goals of solar energy harvesting WSNs: ensuring long-term network sustainability and efficient utilization of harvested energy to enhance the application performance under dynamically changing energy harvesting conditions.