基于贝叶斯优化支持向量回归的老旧小区改造成本预测研究

Research on Cost Prediction of Old Residential Community Renovation Based on Bayesian Optimization Support Vector Regression

摘要: 近年来，随着我国推进城市更新的进程不断提速，老旧小区的改造工程持续增长。为了保障投资效益与项目质量，项目初期对工程造价进行科学、准确的评估已成为关键一环。针对传统预测方法在参数调优方面存在经验性强、效率低等问题，本文提出结合贝叶斯优化算法与支持向量回归(SVR)模型的方法。该模型可自动优化SVR的关键超参数，从而提升模型对复杂非线性数据的适应性与预测准确性。研究选取华东某省会城市200个老旧小区改造项目作为样本，提取了九项影响因子，并利用均方根误差(RMSE)、平均绝对误差(MAE)和决定系数(R²)等指标对比分析了标准SVR与BO-SVR模型的性能。结果表明，BO-SVR模型在预测精度和稳定性方面均优于传统模型。该方法为城市更新工程提供了有效的成本估算工具，具有良好的实际应用前景。

Abstract: In recent years, as China’s process of promoting urban renewal continues to accelerate, the renovation of old neighbourhoods continues to grow. In order to guarantee the investment benefits and project quality, scientific and accurate assessment of project cost at the early stage of the project has become a key part. Aiming at the problems of empirical and low efficiency in parameter tuning of traditional prediction methods, this paper proposes a method combining Bayesian optimisation algorithm and support vector regression (SVR) model. The model can automatically optimise the key hyperparameters of SVR, thus improving the model’s adaptability to complex nonlinear data and prediction accuracy. The study selected 200 old district renovation projects in a capital city in East China as samples, extracted nine influencing factors, and compared and analysed the performance of the standard SVR and BO-SVR models using the root mean square error (RMSE), mean absolute error (MAE) and coefficient of determination (R²). The results show that the BO-SVR model is superior to the traditional model in terms of prediction accuracy and stability. The method provides an effective cost estimation tool for urban renewal projects and has good practical application prospects.