Paper: arXiv 2307.08650
Authors: Supawich Puengdang, Worawate Ausawalaithong, Phiratath Nopratanawong, Narongdech Keeratipranon, Chayut Wongkamthong
Abstract
Real estate is a critical sector in Thailand’s economy, which has led to a growing demand for a more accurate land price prediction approach. Traditional methods of land price prediction, such as the weighted quality score (WQS), are limited due to their reliance on subjective criteria and their lack of consideration for spatial variables. In this study, we utilize aerial or satellite imageries from Google Map API to enhance land price prediction models from the dataset provided by Kasikorn Business Technology Group (KBTG). We propose a similarity-based asset valuation model that uses a Siamese-inspired Neural Network with pretrained EfficientNet architecture to assess the similarity between pairs of lands. By ensembling deep learning and tree-based models, we achieve an area under the ROC curve (AUC) of approximately 0.81, outperforming the baseline model that used only tabular data. The appraisal prices of nearby lands with similarity scores higher than a predefined threshold were used for weighted averaging to predict the reasonable price of the land in question. At 20% mean absolute percentage error (MAPE), we improve the recall from 59.26% to 69.55%, indicating a more accurate and reliable approach to predicting land prices. Our model, which is empowered by a more comprehensive view of land use and environmental factors from aerial or satellite imageries, provides a more precise, data-driven, and adaptive approach for land valuation in Thailand.
Complexity vs Empirical Score
- Math Complexity: 7.5/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced deep learning techniques such as Siamese Networks with EfficientNet and ensemble methods, requiring a solid understanding of neural network architectures and similarity metrics. It demonstrates strong empirical rigor by using real-world datasets from Google Map API and Kasikorn Business Technology Group, reporting specific performance metrics (AUC and MAPE) and practical outcomes like improved recall for land valuation in Thailand.
Research Flowchart
flowchart TD
A["Research Goal: <br>Improve Land Price Prediction <br>Using Aerial/Satellite Imagery"] --> B["Data Collection"]
B --> C["Tabular Data: KBTG Dataset"]
B --> D["Visual Data: Google Map API"]
C & D --> E["Model Architecture"]
subgraph E ["Siamese Neural Network"]
E1["Pretrained EfficientNet"] --> E2["Feature Extraction"]
E2 --> E3["Similarity Scoring"]
end
E --> F["Ensemble with Tree-Based Models"]
F --> G["Key Outcomes"]
subgraph G ["Results"]
G1["AUC: 0.81<br>vs Baseline"]
G2["Recall: 69.55%<br>vs 59.26%"]
G3["MAPE: 20%"]
end
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a similarity-based asset valuation model that uses a Siamese-inspired Neural Network with pretrained EfficientNet to assess land similarity and combines it with tree‑based models.
Source sentence
We propose a similarity-based asset valuation model that uses a Siamese-inspired Neural Network with pretrained EfficientNet architecture to assess the similarity between pairs of lands.
Assumes 1
- The model assumes that sufficient neighboring land data is available to compute reliable similarity scores. data
Source sentence
Since our approach heavily depends on utilizing the similarity score of the neighboring regions, the absence of data about the surrounding areas can adversely affect the accuracy of our land price prediction.
Data
- Land appraisal dataset alternative commercial
Source sentence
We would like to express our gratitude to Kasikorn Business Technology Group (KBTG) for providing us with the valuable land data used in our study
- Satellite imagery alternative public
Source sentence
we utilize aerial or satellite imageries from Google Map API to enhance land price prediction models
Costs
Not applicable
Finds 3
- The best model achieved an AUC of approximately 0.81, outperforming the baseline model. performance
Source sentence
The AUC of our best model was approximately 0.81, indicating a higher level of classification performance compared to the baseline model.
- Coverage Percentage improved from 59.26% to 69.55% using the similarity‑based approach. predictive-accuracy
Source sentence
the model was able to improve the Coverage Percentage from 59.26% (baseline model) to 69.55%, indicating a higher accuracy in predicting the reasonable prices of land.
- At 20% mean absolute percentage error (MAPE), recall increased from 59.26% to 69.55%. predictive-accuracy
Source sentence
At 20% mean absolute percentage error (MAPE), we improve the recall from 59.26% to 69.55%, indicating a more accurate and reliable approach to predicting land prices.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 22 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.