The Baltic Dry Index (BDI) is a commonly utilized indicator of global shipping and trade activity. It influences stakeholders’ and ship-owners’ decisions respecting investments, chartering, operational plans, and export and import activities. Accurate prediction of the BDI is very challenging due to its volatility, non-stationarity, and complexity. To help stakeholders and ship-owners make sound short- and long-term maritime business decisions and avoid market risk, we performed short- and long-term predictions of BDI using an ensemble deep-learning approach. In this study, we propose to apply recurrent neural network models for BDI prediction. The state-of-the-art of sequential deep-learning models such as RNN, LSTM, and GRU are employed to predict one- and multi-step-ahead BDI values. In order to increase the accuracy, we assemble the models. In experiments, we compared our results with those of traditional methods such as ARIMA and MLP. The results showed that our proposed method outperforms ARIMA, MLP, RNN, LSTM, and GRU in both short- and long-term prediction of BDI.

Event logs are records of events that are generally used in process mining to determine the manner in which various processes are practically implemented. Previous studies on process mining attempted to combine the results based on different perspectives such as control flow, data, performance, and resources (organizational) to create a simulation model. This study focuses on the resource perspective. A prior study from the resource perspective focused on clustering the resources into organizational units. Implementing the results of the above study in a simulation model will yield inaccurate results because the resources are assumed to always be available if no task is performed. In a practical scenario, resources (particularly humans) tend to work based on shifts. Thus, we propose mining the shift work operation of resources from event logs to tackle this issue. We utilized a self-organizing map and k-means clustering to incorporate the shift work information from the event logs into the simulation model. Moreover, we introduce a distance function and weight-centroid updating rule in the clustering technique to realize our objective. We conducted extensive experiments with artificial data sets to assess the effectiveness of the proposed method. The simulation shows that introducing the shift work operation time of resources can yield more accurate results. Furthermore, the proposed distance function can capture the shift work operation of the resources more precisely compared with the general distance function.

No abstract

No abstract

No abstract