DEVELOPMENT OF ARTIFICIAL INTELLIGENCE MODEL FOR ACCIDENT PREDICTION IN THE WORKPLACE

The essence of this study was to develop a model where occupational accidents could be predicted, hence ensuring workplace safety. This study was carried out in order to meet real-time industrial needs. To achieve this, accident history was collected from OSHEA which comprises data on worker experience, lighting condition, use of PPE, weather condition, age of machine, Economic Loss. The data obtained from OSHEA was preprocessed to ensure data uniformity and thereafter, trained with the algorithm to develop an ANN model for accident prediction. The model was trained and validated in a repetitive pattern, until the model yielded the desired output. The training process was carried out in 50 Epochs for both training and Validation. The model was thereafter tested with unseen data, which was able to give a prediction with an accuracy of 97.78%. The results obtained during training, validation and testing recorded high accuracy. The accuracy, losses, RMSE, MSE for both training and validation were established. The values of Precision, F1-score and Recall were also obtained. However, the system had a training accuracy of 100%, which shows how well the neurons were able to learn the features on the dataset, while the test accuracy of the model was 97.78%. The F1-Score, Recall and Precision values were 0.9767, 0.9545 and 1.0000 respectively. This means that the model was able to correctly predict 97.78% of the accidents cases correctly, during Testing. The evaluation of the results indicates an exceptional performance, with near perfect scores across all metrics. From the evaluation of the model, the model can be seen as a great addition to the body of knowledge. The ANN model developed in this study can therefore assist safety engineers, and facility managers in identifying hazards before they escalate into full accidents. Ultimately, the study demonstrates that integrating AI into safety management systems has the potential to significantly reduce accident rates, improve human and equipment safety, and enhance overall operational efficiency. This model is recommended for commercialization, as it will meet real-time accident prediction needs of industries.