Λογότυπο αποθετηρίου

Πλοήγηση ανά Συγγραφέα "Lazaros, Nikolaos-Ioannis"

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω
Φιλτράρετε τα αποτελέσματα πληκτρολογώντας τα πρώτα γράμματα
Τώρα δείχνει 1 - 1 από 1
  • Μικρογραφία εικόνας
    Τεκμήριο
    Modeling and forecasting construction material time series using dynamic panel data models
    (2025-10-08) Lazaros, Nikolaos-Ioannis; Λάζαρος, Νικόλαος-Ιωάννης; Besbeas, Panagiotis; Psarakis, Stelios; Vrontos, Ioannis
    This thesis investigates the modeling and forecasting of U.S. cement import using dynamic econometric methods, focusing on short- and long-term trends over four central states: California, Florida, North Carolina, and Texas. Our interest stems from the fact that construction materials have always played a critical role in infrastructure planning, while their prices have been characterized by considerable volatility. Our approach is based on the construction of a panel dataset covering the period 2008–2022 and the use of time series and panel data techniques. The use of import volume instead of prices is chosen in this case due to address data availability issues. Besides, imports dynamics reflect the fluctuations of construction activity and, therefore, those of domestic demand and prices of materials. After transforming the data to ensure stationarity, Lasso regression is used for the selection of state-specific predictors. These variables are incorporated into ARIMAX models to improve forecasting performance. We evaluate each model’s predictive ability using metrics such as RMSE, MAE, and Mean Error, with ARIMAX generally producing strong results, particularly in California and Texas. In addition, a number of panel modeling approaches, including Pooled OLS, Fixed Effects, and Seemingly Unrelated Regressions (SUR) are explored. While these models offer interpretive insights, their predictive performance varies across states, with the SUR model examined primarily as an exploratory framework for capturing potential cross-sectional dependencies. The results highlight the effectiveness of combining variable selection and ARIMA modeling for state-level forecasting of construction imports. They also emphasize the importance of adapting model structures to regional characteristics and data behavior, offering guidance for both practitioners and policymakers in the construction and economic planning sectors.