Scholarlink

<p>In this study, the application of various machine learning (ML) models in forecasting the Revenue Next Year of Vietnamese construction firms is investigated. Using a large dataset spanning 2011–2022, the research compares the performance of Gradient Boosting Regression (GBR) and various other ML models, including Lasso, Ridge Regression, k-Nearest Neighbors Regressor, Support Vector Regression, Random Forest, Extreme Gradient Boosting, and Multilayer Perceptron, with traditional forecasting methods. 16 independent variables were taken into consideration at first, and upon utilizing the Pearson correlation matrix analysis along with rigorous testing, five significant variables were selected to be included within the models to achieve greater predictive power. The results show that the GBR model outperforms the other models, achieving R<sup>2</sup> = 0.669, RMSE = 1168, and MAE = 561. This demonstrates its ability to handle data outliers and achieve high predictive accuracy. Application of the GBR model in forecasting 2023 revenues of five major Vietnamese construction firms achieved an accuracy enhancement of 6–21% compared to traditional methods. The central hypotheses tested in this study are: (1) Advanced machine learning algorithms, particularly ensemble methods like GBR, provide superior revenue forecasting performance compared to traditional statistical methods, and (2) Financial variables such as firm size, financial leverage, and EBIT are important drivers of the accuracy of revenue forecasts. However, the study has several limitations. Using annual data may have overlooked detailed trends, suggesting that analyzing quarterly data in future work could capture more nuanced patterns. Computational limitations restricted the extent of hyperparameter tuning and cross-validation, which could influence the model’s performance. By offering empirical evidence of the usefulness of ML models for financial forecasting and supplying actionable implications for decision-making processes, this study contributes significantly to the existing state of knowledge in the field.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Advanced revenue forecasting techniques for Vietnamese construction enterprises using machine learning

  • 期刊论文

  • 发表日期：

    2025年10月24日

查看全文
  • Pham Vu Hong Son,
  • Le Tung Duong

摘要

In this study, the application of various machine learning (ML) models in forecasting the Revenue Next Year of Vietnamese construction firms is investigated. Using a large dataset spanning 2011–2022, the research compares the performance of Gradient Boosting Regression (GBR) and various other ML models, including Lasso, Ridge Regression, k-Nearest Neighbors Regressor, Support Vector Regression, Random Forest, Extreme Gradient Boosting, and Multilayer Perceptron, with traditional forecasting methods. 16 independent variables were taken into consideration at first, and upon utilizing the Pearson correlation matrix analysis along with rigorous testing, five significant variables were selected to be included within the models to achieve greater predictive power. The results show that the GBR model outperforms the other models, achieving R2 = 0.669, RMSE = 1168, and MAE = 561. This demonstrates its ability to handle data outliers and achieve high predictive accuracy. Application of the GBR model in forecasting 2023 revenues of five major Vietnamese construction firms achieved an accuracy enhancement of 6–21% compared to traditional methods. The central hypotheses tested in this study are: (1) Advanced machine learning algorithms, particularly ensemble methods like GBR, provide superior revenue forecasting performance compared to traditional statistical methods, and (2) Financial variables such as firm size, financial leverage, and EBIT are important drivers of the accuracy of revenue forecasts. However, the study has several limitations. Using annual data may have overlooked detailed trends, suggesting that analyzing quarterly data in future work could capture more nuanced patterns. Computational limitations restricted the extent of hyperparameter tuning and cross-validation, which could influence the model’s performance. By offering empirical evidence of the usefulness of ML models for financial forecasting and supplying actionable implications for decision-making processes, this study contributes significantly to the existing state of knowledge in the field.