Predictability of the Physical Shipping Market by Freight Derivatives

Duru O., GÜLAY E., Bekiroglu K.

IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, cilt.70, ss.267-279, 2023 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 70
  • Basım Tarihi: 2023
  • Doi Numarası: 10.1109/tem.2020.3046930
  • Dergi Adı: IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Social Sciences Citation Index (SSCI), Scopus, Academic Search Premier, ABI/INFORM, Aerospace Database, Applied Science & Technology Source, Business Source Elite, Business Source Premier, Communication Abstracts, Compendex, Computer & Applied Sciences, INSPEC, Metadex, Public Affairs Index, Civil Engineering Abstracts
  • Sayfa Sayıları: ss.267-279
  • Anahtar Kelimeler: Predictive models, Forecasting, Time series analysis, Training, Optimization, Search engines, Raw materials, Commodity market, freight derivatives, grid search algorithm, lead-lag structure, predictive analytics, TIME-SERIES, PREDICTION, PRICES, MODELS, SPOT
  • Dokuz Eylül Üniversitesi Adresli: Evet

Özet

This article investigates the predictability of dry bulk shipments' physical shipping costs while testing the predictive significance of derivative products. Accordingly, a comprehensive grid search procedure is needed to simulate combinations of model structures subject to a cross-validation process. In this regard, the intelligent model search engine (IMSE) is implemented as the search algorithm. The IMSE algorithm is selected due to its broad model space naturally consisting most of the traditional econometric model structures and advanced features, such as sample size optimization (also detects structural breaks), lag structure optimization (also reflects seasonality), among others. In contrast to the previous studies in the shipping market research, IMSE executes a relaxation of statistical significance criteria to target the out of sample predictive accuracy (instead of the goodness of fit). Sparsification is retained by utilizing a dropout procedure in line with the cross-validation process. The empirical results indicate the fact that predictive accuracy can be achieved or improved with a shorter sample period. More than half of the entire dataset has not been utilized in the final selections for most simulations. One-month and two-month maturity derivative contracts are tested for predictive features, while none of them is selected for the sparse model selections.