Machine Learning Methods for Predicting the Outcome of Safety Culture Attributes

Measuring organisational safety culture is resource intensive. Safety culture measurement tools must be simplified and cost-effective to enable SMEs to measure their safety culture. This can be achieved by reducing the data required by such devices. Predictive analytics suits such a situation as they are designed to predict outcomes with limited input data. This paper investigated the effectiveness and reliability of predictive analytics in reducing the data requirement of safety culture measurement tools. The paper utilised only the perception survey component of a culturally safe tool conducted by over 100 companies to generate a neural network, a predictive algorithm. The predicted result of the neural network, when compared to the original culture-safe score, yielded an r2 greater than 0.7. This outcome demonstrates the feasibility of using machine learning algorithms to reduce the amount of input data required while retaining its accuracy.