<![CDATA[Penilaian tanah massal berbasis bidang tanah dapat dilakukan dengan menggunakan model berdasarkan variabel-variabel bebas pembentuk nilai tanah. Salah satu model yang disarankan untuk model nilai tanah yaitu algoritma random forest. Agar random forest memiliki kemampuan prediksi yang baik maka perlu dilakukan pengaturan hyperparameter antara lain mtry, jumlah pohon yang dibangun (num.trees), dan jumlah minimal node (min.node.size). Penelitian ini bertujuan untuk melihat kinerja model penilaian tanah menggunakan random forest dengan analisis pengaturan hyperparameter agar mendapatkan hasil akurasi terbaik. Hasil penelitian menunjukkan model nilai tanah memiliki nilai R2 sebesar 82,96% dan nilai mean absolute percent error (MAPE) sebesar 26,83 sehingga model memiliki kemampuan prediksi yang layak. Berdasarkan standar rasio yang dikeluarkan oleh IAAO, uji kualitas model penilaian tanah menghasilkan nilai coefficient of variation (COV) sebesar 22,62 dan nilai coefficient of dispersion (COD) sebesar 25,29 serta nilai price related differential (PRD) sebesar 1,12. Nilai COV, COD, dan PRD masih di luar batas toleransi yang ditetapkan. Parcel-based mass valuation can be carried out using a model based on the land value independent variables. One model that is recommended for modeling land value is the random forest algorithm. In order for random forest to have good predictive capabilities, it is necessary to tune hyperparameters including mtry, number of trees built (num.trees) and minimum number of nodes (min.node.size). This research aims to show the performance of a land valuation model using random forest with hyperparameter tuning analysis to get the best accuracy. The research results show that the land valuation model has an R2 of 82.96% and a mean absolute percent error (MAPE) of 26.83, so the model has reasonable predictive ability. Based on the standard ratio issued by the IAAO, the land valuation model quality test gives a coefficient of variation (COV) value of 22.62, a coefficient of dispersion (COD) value of 25.29, and a price-related differential (PRD) value of 1.12. COV, COD, and PRD are still outside the specified tolerance limits]]>
Copyrights © 2024
