All Times ET
Keywords: Work-related Injury, Knowledge Graph, Machine Learning, Classification
This research illustrates the statistics of the fatal work-related injury cases in Shenzhen, China and displays them with features using Knowledge Graph with a search function of relevant entities. AdaBoost is the superior model among selected machine learning classification methods to predict the most relevant work-related accidents to a certain type of company under a given circumstance. Advice could be further made upon utilizing the prediction model for enterprises to make precautional plans to minimize the risk of work-related fatal accidents and injuries.
It mainly contains 3 parts: 1. Descriptive statistical analysis. The dataset adopted in this study reflects the structure of fatal work-related injury events in Shenzhen. 2. Knowledge graph. Using the graph construction tool, Neo4j, data are transformed into a clearer entity- relationship diagram to more intuitionally display the associations between entities from the data, based on which search function could be applied. 3. Machine-learning-based prediction model. Features are selected to realize the prediction of the potential type of fatal accident for companies. Machine learning classification algorithms including Decision Tree, SVM, KNN, and the ensemble algorithm AdaBoost are selected and compared.
The statistical analysis of the experimental dataset provides a ground-breaking demonstration of the distribution of indicators for fatal work-related injury events in Shenzhen. Event-related entity-relationship display and search using knowledge graph provide an efficient tool for conveying and queries about associated information. The machine learning classification model of work-related injury events computes the most potential type of accident for specific work types in certain companies. The model could be utilized to provide valid advice for decision-makers with the specific work-related injury prevention for companies and employees.