This study predicts the cash holdings policy of Turkish firms, given the 20 selected features with machine learning algorithm methods. 211 listed firms in the Borsa Istanbul are analyzed over the period between 2006 and 2019. Multiple linear regression (MLR), k-nearest neighbors (KNN), support vector regression (SVR), decision trees (DT), extreme gradient boosting algorithm (XGBoost) and multi-layer neural networks (MLNN) are used for prediction. Results reveal that MLR, KNN, and SVR provide high root mean square error (RMSE) and low R2 values. Meanwhile, more complex algorithms, such as DT and especially XGBoost, derive higher accuracy with a 0.73 R2 value. Therefore, using advanced machine learning algorithms, we may predict cash holdings considerably.

The Strait of Istanbul is one of the world's busiest, narrowest and most winding waterways. As such, there is a high grounding probability for vessels. Although a number of grounding probability models exist, they have been deemed unsuitable by local maritime experts, due to their insufficient stopping distance criteria for narrow waterways. Thus, there is a need for a new model. This paper proposes a two-component grounding probability model that multiplies the geometric grounding probability (calculated with a kinematic-based model) with the causation probability (calculated with a specially designed Bayesian network). The geometric probability model is improved in terms of stopping distance parameters and the Bayesian network is crafted for narrow waterways. The model is then deployed with pre-determined parameters within the Strait of Istanbul to run risk analysis scenarios. The results, validated with actual grounding records, show that the causation probability is the key component for quantifying the probability of grounding in narrow waterways. If navigated without frequent evasive manoeuvres, grounding would be almost inevitable. Although this study focuses on the Strait of Istanbul, the proposed approach can be applied to research into grounding probability of vessels navigating through other waterways.

The aim of this study is to develop a simulation model which is capable of mimicking actual vessel arrival patterns and vessel entrance decisions (which are made based on expert opinions generally) on congested, narrow waterways. The model is tested on the transit traffic pattern in the Strait of Istanbul. Based on a heuristic scheduling algorithm, this model decides entrance times and vessel types on the strait. The model, with different policies for day and night traffic, is run for a period of seven years with 20 replications for each year. The performance measures of the model are: average interarrival times, number of vessels passed and entrance times for each successive vessel pair in both traffic directions. The model results are congruent with the actual results of performance measures. Therefore, it may be deduced that the proposed algorithm can be a guide for operators regarding scheduling decisions in congested, narrow waterways.

No abstract