Pengembangan Model Appraisal Berbasis Data Citra Satelit Menggunakan Metode Deep Learning Untuk Penilaian Properti Di Kota Surabaya

Babgei, Bassam (2025) Pengembangan Model Appraisal Berbasis Data Citra Satelit Menggunakan Metode Deep Learning Untuk Penilaian Properti Di Kota Surabaya. Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 2043201060-Undergraduate_Thesis.pdf] Text
2043201060-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only

Download (4MB) | Request a copy

Abstract

Pertumbuhan pesat ekonomi di Surabaya telah mendorong perkembangan signifikan dalam sektor properti, dengan peningkatan konsisten sejak awal tahun 2017, terutama dalam sub-sektor tanah dan perumahan. Permintaan tempat tinggal yang meningkat didorong oleh pertumbuhan populasi dan keluarga baru, sementara Pajak Bumi dan Bangunan (PBB) menjadi salah satu sumber utama pendapatan negara. Penilaian properti umumnya dilakukan melalui pendekatan perbandingan pasar, membandingkan properti dengan sejenisnya di sekitar lokasi. Faktor-faktor seperti ukuran, jumlah ruangan, dan keadaan lingkungan memengaruhi nilai properti, dengan citra satelit dan data Point of Interest (POI) dapat menggambarkan keadaan lingkungan dan tingkat kepadatan aktivitas manusia di sekitar properti. Meskipun telah banyak penelitian fokus pada penilaian menggunakan spesifikasi fisik rumah, masih jarang studi yang mempertimbangkan aspek lingkungan di sekitar rumah. Penelitian ini mengusulkan pendekatan berbasis data citra dengan menggabungkan citra satelit, spesifikasi rumah, dan POI untuk memprediksi nilai properti di Kota Surabaya menggunakan metode Convolutional Neural Network (CNN). CNN dipilih karena kemampuannya mengolah data citra serta efektivitasnya dalam mengenali pola spasial dan fitur visual kompleks, membuatnya ideal untuk memanfaatkan informasi visual lingkungan yang tidak terstruktur. Kota Surabaya dipilih sebagai lokasi penelitian karena keunikan dan kompleksitas kota tersebut, menjadikannya konteks relevan untuk menguji metode ini dalam dinamika perkembangan properti. Hasil ekstraksi fitur citra dengan CNN akan dijadikan variabel prediktor untuk memprediksi harga rumah di Kota Surabaya dengan membandingkan metode antara XGBoost dan MLP. Metode terbaik yang terpilih adalah XGBoost menggunakan data spesifikasi dengan nilai R2 sebesar 16,33%, MAPE sebesar 69,03%, dan RMSE sebesar Rp 182.518.279.
==================================================================================================================================
The rapid economic growth in Surabaya has driven significant developments in the property sector, with consistent increases since early 2017, especially in the land and housing sub-sectors. The rising demand for housing is driven by population growth and new families, while Land and Building Tax (PBB) has become one of the primary sources of government revenue. Property valuation is generally conducted through a market comparison approach, comparing the property with similar ones in the surrounding area. Factors such as size, number of rooms, and environmental conditions affect property Values, with satellite imagery and Point of Interest (POI) data providing insights into the property's surroundings and the density of human activity in the area. While much research has focused on physical property valuation, few studies have considered environmental aspects. This study proposes an image-based approach by combining satellite imagery, house specifications, and POI to predict property Values in Surabaya using Convolutional Neural Networks (CNN). CNN is chosen for its ability to process image data and POI simultaneously and for its effectiveness in recognizing spatial patterns and complex visual features, making it ideal for utilizing unstructured environmental visual information. Surabaya was selected as the study location due to its unique and complex urban characteristics, which provide a relevant context for testing this method in the dynamic property development landscape. The feature extraction results from CNN serve as predictor variables to estimate housing prices in Surabaya, with a comparison of methods between XGBoost and Multi-Layer Perceptron (MLP). The best-performing method, XGBoost using specification data, achieved an R2 value of 69,03%, a Mean Absolute Percentage Error (MAPE) of 16,33%, and a Root Mean Square Error (RMSE) of IDR 182.518.279.

Item Type: Thesis (Other)
Uncontrolled Keywords: Pajak Bumi Bangunan, Penilaian Properti, Citra Satelit, Point of Interest, Convolutional Neural Networks
Subjects: Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions: Faculty of Vocational > 49501-Business Statistics
Depositing User: Bassam Babgei
Date Deposited: 23 Jul 2025 07:59
Last Modified: 23 Jul 2025 07:59
URI: http://repository.its.ac.id/id/eprint/120849

Actions (login required)

View Item View Item