Browsing by Author "Edem, I. E."

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Development of an artificial neural network-fuzzy-Markov model for industrial accidents forecasting
    (Department of Industrial and Production Engineering, University of Ibadan, 2020) Edem, I. E.; Adebimpe, O. A.
    Industrial accidents possess the potential of causing physical, psychological and even fatal consequences when they occur. In that regard, Industrial accidents forecasting aid stakeholders in properly managing and improving workplace safety by anticipating accident occurrences to prevent or minimise their consequences. Due to the high variation, random and fluctuating characteristics of industrial accident occurrences, machine learning and Markov based models methods have become increasingly popular as a tool for understanding their occurrence patterns and detecting their vibrational directions. However, little investigation has been made towards combining the positive characteristics of these methods for industrial accidents forecasting. This study is concerned with the development of a neuro-fuzzy-Markov model for the prediction and forecasting of industrial accident occurrences. The methodology employed essentially involves the implementation of the Artificial Neural Network (ANN) model through the development of structured control methods to enhance improved forecast candidate generation potential. Further, an analysis of the model's residual was undertaken to obtain an ANN forecast correction factor by using a combination of fuzzy and Markov techniques. Based on this, investigations were then carried out to determine the direction of vibration of the ANN predictor model. Subsequently, results were generated from the prediction mechanism. The model was validated by comparing its one window ahead (OWA) forecast potential with those of the ANN model using secondary industrial accident data based on the mean absolute percentage error (MAPE) and the Root Mean Square Error (RMSE). Also, an evaluation was done by comparing the forecast performances of the model with those of two traditional (Autoregressive moving average [ARIMA] and exponential smoothing [EXPSM]) models and two non-traditional (Mao and Sun grey-Markov (MSGM) and Grey-Fuzzy-Markov Pattern Recognition GFMAPR) models. The forecast performance results obtained on the model's application showed that it possessed the capability to correct and improve ANN forecasts. The MAPE and RMSE results obtained for the ANN-fuzzy-Markov model were 15.39 and 26.39, while those produced by the ANN were 20.13 and 30.57 respectively. The model produced more superior forecast when compared with ARIMA and EXPSM, and compared well with the MSGM and GFMAPR. The obtained results indicate that the model possesses the capability of carrying the predictions of industrial fire accident to an acceptable degree of accuracy.