Hidayat, Mochamad Arief (2022) Analisis Harga Jual Rumah di Jabodetabek dengan Web Scraping dan Machine Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
09211850096013 - Mochamad Arief Hidayat - Thesis - Final.pdf - Accepted Version Download (6MB) | Preview |
Abstract
Di dalam jaman big data saat ini, internet menyediakan beragam informasi dimana data adalah sebuah kekayaan baru. Dengan perkembangan saat ini, ada disrupsi tentang cara jual beli rumah, marketplace membuat informasi jual beli rumah semakin terbuka, mudah dan gratis didapatkan. Akibatnya, calon pembeli dihadapkan oleh banyaknya pilihan dalam pembelian rumah. Banyaknya pilihan itu bisa diperkecil jika marketplace melakukan tambahan fitur sebagai saran atau rekomendasi untuk calon pembeli. Fungsi marketplace saat ini hanya sebagai search engine dalam pencarian rumah yang akan dijual. Seperti apa perkembangan marketplace juga menarik untuk dipelajari. Selain itu, ketiadaan prediksi harga jual rumah di marketplace yang bisa membantu calon pembeli dalam menyiapkan anggaran berikut renca. Oleh karena itu, tesis ini bertujuan untuk mengetahui tren pasar jual beli rumah berdasarkan data yang didapatkan di marketplace untuk Jakarta, dan daerah pendukung seperti Bogor, Depok, Tangerang dan Bekasi (Jabodetabek) dengan web scraping. Tujuan berikutnya adalah memberikan saran untuk marketplace untuk fitur tambahan, mengetahui perbandingan model prediktif yang paling akurat dari data yang di dapat tersebut. Model prediktif yang dibandingkan adalah metode hedonic price, generalized additive model (GAM) dan artificial neural network (ANN). Hedonic model paling sering digunakan penelitian sebelumnya dalam analisis harga jual. Prediksi semi parametric dengan GAM digunakan untuk data-data non linear. Metode prediksi machine learning yang sering digunakan adalah metode ANN.
Berdasarkan data bersih yang didapat dibulan juli, september, nopember dan desember 21, rata-rata harga jual rumah di jabodetabek didapatkan kenaikan dari 3,4 miliar menjadi 4,6 miliar. Adanya penambahan rumah yang sebagian besar dijual di jakarta selatan dan bernilai lebih dari 3,5 miliar membuat rata-rata harga jual rumah meningkat.
Tambahan fitur di marketplace sehingga marketplace tidak berfungsi seperti search engine. Adanya fitur calon pembeli yang khusus, rekomendasi luas tanah/luas bangunan berdasarkan harga jual rumah, atau fitur prediksi harga rumah dengan variabel yang di inginkan oleh calon pembeli bisa menjadi daya tarik tambahan untuk marketplace tersebut. Selain itu penambahan fitur seperti link untuk mempertemukan dengan agen, performance agen dan sebagainya
Hasil prediksi ANN memberikan prediksi yang paling akurat dengan R2 sebesar 98%. Selanjutnya GAM memberikan hasil prediksi dengan R2 sebesar 88%. Hasil prediksi dengan hedonic model memberikan prediksi yang paling kurang akurat yaitu 79%.
=======================================================================================================
In the big data era, the internet had been swarmed with a lot of
information. Information is a new “oil”. Virtual marketplaces on the internet have
become a new way for trading and disruptive to the offline trading. All
information in the marketplace are opan , and free. Customer is faced with a lot of
information, options and difficulties for making decision. However the
marketplace is behaves like search engine to display which house to be offered. If
the marketplace can provide a feature to reduce options for customer, it will helps
the customer for making decision. Growth of house selling market based on
marketplace data is interesting point to discover. Price prediction based on
customer needs is not proviced, if prediction is available then the customer will
understand the budget to be prepared and its financial plan. where marketplace is
not provide The purpose of this tesis is to study market trends of houseprice in
Jakarta metropolitan area such as Bogor, Depok, Tangerang, and Bekasi
(Jabodetabek) are extracted. Next, to provide a suggestion for the marketplace and
compare each predictive methods. Which method is the most accurate for
predicting the house selling price, Predictice models is consist of hedonic model,
Generalized additive models (GAM) and Artificial Neural Network. Hedonic
models is a common models for house selling price. GAM is a popular model for
semiparametric prediction. ANN is a popular method in machine learning. Data
Visualization based on map are provided using Google data Studio. Model has
been developed for prediction of house selling price. Generalized Additive models
(GAM) and Artificial Neural Network (ANN) had been used to develop the
model. The predicted values of selling house and real price are compared for each
area.
Based on analysis resulted, average of house selling price in nopember is
higher than juli’s data.Average house selling price in desember of 4.6 billion
rupiah is higher than Juli of 3,4 billion. The number of house to be sell is
increased and the most of the additional in jakarta selatan have selling price more
than 3,5 billion rupiah. Hence it creates increasing average of house selling price.
Suggestion for marketplace to create a better market place and more than
“search engine” such as additional feature for special user or summary for
narrowing the searching based on user interest. The special user will know
expected land size / house size based on user’s budget. The special user can have
additional feature for predicting the house sale price and compared with its
budget. Additional feature for the marketplace is connecting with agent and
providing the agent’s performance.Prediction using ANN resulted highest accuracy of 98%. Prediction using
hedonic price resulted minimum accuration of all methods.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Big Data Analytic, Machine Learning, Predictive model, Visualisasi data, Web scrapping |
| Subjects: | Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) Q Science > QA Mathematics > QA76.9.I52 Information visualization |
| Divisions: | Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT) |
| Depositing User: | Mochamad Arief Hidayat |
| Date Deposited: | 12 Mar 2022 07:07 |
| Last Modified: | 07 Oct 2024 06:58 |
| URI: | http://repository.its.ac.id/id/eprint/94818 |
Actions (login required)
 |
View Item |