This thesis presents the development of an ergonomic risk classification model usingInertial Measurement Unit (IMU) sensor data, aimed at improving workplace safety inphysically demanding environments, with a focus on forestry operations. A subset ofpreviously collected IMU data was analyzed to identify bending angle patterns andposture durations. Postures were segmented into ergonomic risk levels low, medium,and high and classified using machine learning models, including Random Forest andneural networks via Edge Impulse.The models demonstrated strong performance, achieving high accuracy in classificationtasks. Results revealed a significant portion of time spent in high risk postures,indicating potential for targeted intervention. This research supports the developmentof intelligent ergonomic monitoring systems and lays groundwork for future applicationsin adaptive exoskeletons and real time posture feedback tools.Keywords: IMU, Ergonomics, Posture Classification, Machine Learning, NeuralNetworks, Edge Impulse, Exoskeleton, Confusion matrix.