<p>This paper introduces an approach to reduce the adaptation space in self-adaptive systems, with a focus on Internet of Things (IoT) environments. Self-adaptive systems dynamically observe and adapt their behavior to changing environmental conditions. IoT systems need to make decisions autonomously and adapt to their surroundings. Therefore, they can be considered self-adaptive systems. A major challenge in self-adaptive systems lies in efficiently searching the knowledge base, extracting necessary information, and making quick adaptation decisions. To address this, the proposed approach incorporates a neuro-fuzzy network to reduce the adaptation options, ultimately improving decision-making speed. The approach consists of a two-part framework: an enhanced self-adaptive system architecture that integrates a machine learning component to optimize the adaptation options, and a neuro-fuzzy system for continuous learning with minimal human intervention. The neuro-fuzzy network includes five layers, utilizing Gaussian membership functions and adaptive fuzzy rules for efficient decision-making. The proposed approach was evaluated using the DeltaIoT simulator. Three scenarios were tested: baseline, artificial neural network, and neuro-fuzzy network implementations. Results demonstrated that the neuro-fuzzy approach outperformed other methods in terms of packet loss rate, delay, and energy consumption, thus validating its potential for improved adaptation in IoT-based self-adaptive systems.</p>