Scholarlink

Using PHP and IBM's SPSS Modeler for data mining and analytics, the authors collected and processed housing data from Rumah123 in Surabaya. A total of 10,336 data points were divided into two clusters. Three machine learning (ML) models were developed using SPSS Modeler for 2,460 data points. The results indicate that the Artificial Neural Network (ANN) model provided the most consistent correlation across all scenarios, achieving an R-squared value of 0.787 and a Mean Absolute Error (MAE) of Rp 634,118,868 in the largest cluster. Meanwhile, the Support Vector Machine (SVM) model performed the worst, with an R-squared of only 0.218 and a significantly higher MAE of Rp 1,327,826,130. The Classification and Regression Tree (CART) model showed good performance in both training and testing for the larger cluster, with an R-squared of 0.524 and MAE of Rp 804,371,767, but did not perform as well in the smaller cluster. Overall, the ANN and CART models demonstrated strong potential for predicting housing prices, with ANN offering the highest accuracy and lowest error rates across both clusters.

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

Developing Machine Learning Algorithms for Predicting House Prices in Surabaya Using IBM SPSS Modeler

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hansel Davin Sugiarto,
  • Doddy Prayogo,
  • Njo Anastasia

摘要

Using PHP and IBM's SPSS Modeler for data mining and analytics, the authors collected and processed housing data from Rumah123 in Surabaya. A total of 10,336 data points were divided into two clusters. Three machine learning (ML) models were developed using SPSS Modeler for 2,460 data points. The results indicate that the Artificial Neural Network (ANN) model provided the most consistent correlation across all scenarios, achieving an R-squared value of 0.787 and a Mean Absolute Error (MAE) of Rp 634,118,868 in the largest cluster. Meanwhile, the Support Vector Machine (SVM) model performed the worst, with an R-squared of only 0.218 and a significantly higher MAE of Rp 1,327,826,130. The Classification and Regression Tree (CART) model showed good performance in both training and testing for the larger cluster, with an R-squared of 0.524 and MAE of Rp 804,371,767, but did not perform as well in the smaller cluster. Overall, the ANN and CART models demonstrated strong potential for predicting housing prices, with ANN offering the highest accuracy and lowest error rates across both clusters.