Scholarlink

<p>Against this backdrop of a global economic downturn, the real estate market exhibits characteristics of both a consumer good and an investment asset, with housing prices displaying fluctuating nonlinear trends. Predicting these trends holds significant importance for governments, developers, and individuals. Addressing the vulnerability of BP neural networks (BPNN) to initial weights and biases in housing price forecasting, this study enhances the Artificial Lemming Algorithm (ALA) by incorporating pinhole imaging learning, evolutionary mutation perturbation, and golden sine development strategies. This improved ALA is applied to update the BPNN network structure, establishing the novel IALA-BPNN housing price prediction model. Experiments using four public datasets compare IALA-BPNN against eight comparable hybrid models, validating the model’s success and the effectiveness of the three improvement strategies. Further comparisons with four machine learning models (RF, SVM, LSTM, CNN) demonstrate the model’s strong competitiveness in housing price forecasting. Finally, a Shanghai housing price dataset is constructed. Comparing IALA-BPNN with the top-performing models from the previous experiments reveals that IALA-BPNN achieves significant improvements across evaluation metrics: MAE (13.63%), MAPE (14.18%), RMSE (14.99%), R<sup>2</sup> (9.38%), SMAPE (19.15%), SD (8.46%), and Time (34.30%), demonstrating its practical application potential.</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

IALA-BPNN: superior house price prediction through multi-strategy optimized Artificial Lemming Algorithm and BP neural network

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月20日

查看全文
  • Qifeng Guo,
  • Shengpeng Li,
  • Tianqi Xia,
  • Daren Chen

摘要

Against this backdrop of a global economic downturn, the real estate market exhibits characteristics of both a consumer good and an investment asset, with housing prices displaying fluctuating nonlinear trends. Predicting these trends holds significant importance for governments, developers, and individuals. Addressing the vulnerability of BP neural networks (BPNN) to initial weights and biases in housing price forecasting, this study enhances the Artificial Lemming Algorithm (ALA) by incorporating pinhole imaging learning, evolutionary mutation perturbation, and golden sine development strategies. This improved ALA is applied to update the BPNN network structure, establishing the novel IALA-BPNN housing price prediction model. Experiments using four public datasets compare IALA-BPNN against eight comparable hybrid models, validating the model’s success and the effectiveness of the three improvement strategies. Further comparisons with four machine learning models (RF, SVM, LSTM, CNN) demonstrate the model’s strong competitiveness in housing price forecasting. Finally, a Shanghai housing price dataset is constructed. Comparing IALA-BPNN with the top-performing models from the previous experiments reveals that IALA-BPNN achieves significant improvements across evaluation metrics: MAE (13.63%), MAPE (14.18%), RMSE (14.99%), R2 (9.38%), SMAPE (19.15%), SD (8.46%), and Time (34.30%), demonstrating its practical application potential.