The Industrial Internet of Things (IIoT) is transforming modern manufacturing environments by providing real-time data from interconnected devices, significantly enhancing information visibility and automation. Through a network of intelligent sensors integrated across the production system and equipment, it enables real-time monitoring and data collection. This continuous stream of sensor data contributes to industrial big data, facilitating more intelligent decision-making by providing significant insights into operational conditions. One of the major applications of this data is predictive maintenance, which uses machine learning models to identify early signs of equipment failure. This helps manufacturers to proactively schedule maintenance, reduce unexpected breakdowns and operational downtime, reduce maintenance costs, and extend the service life of equipment. In this work, a novel predictive maintenance framework is proposed using an enhanced version of Genghis Khan Shark Optimizer (CGKSO), a nature-inspired metaheuristic algorithm. Each CGKSO candidate represents a set of feature weights for a logistic classifier, and these weights evolve over successive iterations to maximize prediction accuracy. The model uses a sigmoid activation function to classify failure conditions, providing an efficient and accurate solution for equipment health monitoring in modern industries.