臺灣博碩士論文加值系統

房價預測在目前是非常熱門的探討議題，且有愈來愈多研究團隊採取深度學習及機器學習的方法進行相關的議題研究。但是有些研究只探討了房子本身的特徵，並沒有考慮更多可能會影響房價的資訊，導致預測的結果不夠準確，所以我們認為需要添加更多外部資料來輔助房價預測。受到注意力機制的啟發，此機制可以應用於圖像、語音辨識，以及翻譯的議題上，所以我們試圖將注意力機制引入到我們的問題上，來協助模型找出購買房子的重要特徵。
我們認為，透過引入更多與房價相關的外部資訊可以幫助提升預測的效果，於是，我們取得可以代表周圍生活機能的公共設施，像是公園、學校和捷運站等，同時也使用衛星地圖以顯示周圍環境。受到注意力機制概念的啟發，給定不同的資料情況下，模型能夠自動學習模型的權重，使模型在訓練時能夠更加靈活，因此在這篇論文，我們提出共同注意機制模型，此模型與以往的自注意機制不同的地方，在於它不僅考慮了不同特徵各自的關係，同時還考慮到兩兩不同特徵的交互關係，透過此機制可以使模型可以預測的更加準確。
本研究以2017和2018年內政部提供的實價登錄中的不動產交易資料，以及台北市政府和新北市政府提供的公開資料，以及使用Google Map API 抓取的衛星地圖，做為我們實驗的研究標的，以用於評估我們所提出的模型成效。我們取得這些資料後，接著以我們所提出的模型進行訓練，並且與其他的機器學習方法，例如XGBoost 和LightGBM，以及深度學習方法和不同的注意力模型做比較。實驗結果說明本研究提出的模型與基準模型的23.8% 相比，達到最低的MAPE 13.31%，並且優於其他對照組。

House price prediction has been a popular for a few years, and more and more research teams are doing the related studies with deep learning or machine learning methods. However, some of the studies do not consider more information that will impact house price, so the prediction results are always not precise enough, therefore we think that import more information can help support house information. We also inspired by attention mechanism which is widely used on image, speech, and translation tasks, and we adopt it on our issue to find out the important features when buying a house. We import public facilities such as park, school, and MRT stations can represent the level of surrounding living function, and satellite map can demonstrate the surrounding environment.
Inspired by the concept of attention mechanism, we propose a joint self-attention model which combines attention mechanism to highlight those important features that buyers would mind, and also make model more flexible when training due to the characteristics of automatically assign model weight when given different transaction data. Our proposed model is different from the self-attention model, it considers the interaction between two different features to make model learn the complicated relationship in each feature to make prediction results more precise.
In our experiments, actual selling price of real estate transaction data in 2017 and 2018, public facilities data acquired from Taipei and New Taipei Government, and satellite maps crawled with Google Map API are experimental dataset to evaluate our proposed model. We utilize these datasets and trained with our proposed model with other machine learning-based methods such as XGBoost and LightGBM, and deep learning methods and several attention models. The experimental results show that the proposed model achieves lowest prediction MAPE 13.31% compared to baseline 23.82%, and outperforms other approaches.

摘要 i
Abstract ii
Acknowledgement iii
Table of Contents iv
List of Figures vi
List of Tables vii
1 Introduction 1
1.1 Motivation 1
1.2 Background and Problem Settings 3
1.3 Research Goal 4
1.4 Contributions and Value 6
1.5 Organization 6
2 Literature Review 7
2.1 House Price Prediction. 7
2.1.1 Machine Learning Based Approaches 9
2.1.2 Deep Learning Based Approaches 10
2.2 House Price Prediction with Heterogeneous Data 11
2.3 Attention mechanism 13
2.3.1 Machine Translation Application 13
2.3.2 Speech Recognition Application 14
2.3.3 Computer Vision Application 15
2.4 Summary 16
3 Proposed Methods 17
3.1 System Overview 17
3.2 Data Description and Preprocessing 18
3.2.1 Data Description 18
3.2.2 Data Preprocessing 20
3.3 Public Facilities Data Representation Method 24
3.4 Image Feature Extractor 26
3.4.1 Localisation Network 27
3.4.2 Grid Generator 28
3.4.3 Differentiable Image Sampling 29
3.4.4 Street View Image 30
3.5 Attention Mechanism 30
3.5.1 Gated Neural Network 31
3.5.2 Bahadanau Attention 34
3.5.3 Joint Self-Attention Mechanism 37
3.6 Prediction Model 40
3.7 Summary 41
4 Experiments 42
4.1 Evaluation Metrics 42
4.2 House Transaction Data with Heterogeneous Data 43
4.3 Joint Self-Attention Compared with Other Models 48
4.4 Applied for House Price Analysis in Metropolitan Region 50
4.5 System Demonstration 55
4.5.1 Demonstration of Suite 55
4.5.2 Demonstration of Residential Building 58
4.5.3 Demonstration of Townhouse 61
5 Conclusions and Future Works 64
5.1 Conclusions 64
5.2 Future Works 65
References 67

[1] R. Gupta, A. Kabundi, and S. M. Miller, “Forecasting the us real house price index: Structural and nonstructural models with and without fundamentals,” Economic Modelling, vol. 28, no. 4, pp. 2013–2021, 2011.
[2] J. Mu, F. Wu, and A. Zhang, “Housing value forecasting based on machine learning methods,” in Abstract and Applied Analysis, vol. 2014. Hindawi, 2014.
[3] L. Bork and S. V. Møller, “Forecasting house prices in the 50 states using dynamic model averaging and dynamic model selection,” International Journal of Forecasting, vol. 31, no. 1, pp. 63–78, 2015.
[4] A. Ng and M. Deisenroth, “Machine learning for a london housing price prediction mobile application,” in Imperial College London, 2015.
[5] M. Risse and M. Kern, “Forecasting house price growth in the euro area with dynamic model averaging,” The North American Journal of Economics and Finance, vol. 38, pp. 70–85, 2016.
[6] B. Afonso, L. Melo, W. Oliveira, S. Sousa, and L. Berton, “Housing prices prediction with a deep learning and random forest ensemble,” in Anais do XVI Encontro Nacional de Inteligência Artificial e Computacional. SBC, 2019, pp. 389–400.
[7] Y. LeCun, L. Bottou, Y. Bengio, P. Haffner et al., “Gradient-based learning applied to document recognition,” Proceedings of the IEEE, vol. 86, no. 11, pp. 2278–2324, 1998.
[8] B. Park and J. K. Bae, “Using machine learning algorithms for housing price prediction: The case of fairfax county, virginia housing data,” Expert Systems with Applications, vol. 42, no. 6, pp. 2928–2934, 2015.
[9] V. Plakandaras, R. Gupta, P. Gogas, and T. Papadimitriou, “Forecasting the us real house price index,” Economic Modelling, vol. 45, pp. 259–267, 2015.
[10] C. Rahal, House Price Forecasts with Factor Combinations. Department of Economics, University of Birmingham, 2015.
[11] J. Tu and J. Zhang, “What pushes up the house price: Evidence from shanghai,” World Econ, vol. 5, pp. 28–37, 2005.
[12] X. Changrong, “Volatility clustering and short-term forecast of china house price [j],” Chinese Journal of Management, vol. 6, no. 024, 2010.
[13] J. Gu, M. Zhu, and L. Jiang, “Housing price forecasting based on genetic algorithm and support vector machine,” Expert Systems with Applications, vol. 38, no. 4, pp. 3383–3386, 2011.
[14] Y. Wei and Y. Cao, “Forecasting house prices using dynamic model averaging approach: Evidence from china,” Economic Modelling, vol. 61, pp. 147–155, 2017.
[15] J.H. Chen, C. F. Ong, L. Zheng, and S.C. Hsu, “Forecasting spatial dynamics of the housing market using support vector machine,” International Journal of Strategic Property Management, vol. 21, no. 3, pp. 273–283, 2017.
[16] P.F. Chen, M.S. Chien, and C.C. Lee, “Dynamic modeling of regional house price diffusion in taiwan,” Journal of Housing Economics, vol. 20, no. 4, pp. 315–332, 2011.
[17] Z. Peng, Q. Huang, and Y. Han, “Model research on forecast of second-hand house price in chengdu based on xgboost algorithm,” in 2019 IEEE 11th International Conference on Advanced Infocomm Technology (ICAIT). IEEE, 2019, pp. 168–172.
[18] C. R. Madhuri, G. Anuradha, and M. V. Pujitha, “House price prediction using regression techniques: A comparative study,” in 2019 International Conference on Smart Structures and Systems (ICSSS). IEEE, 2019, pp. 1–5.
[19] V. Limsombunchai, “House price prediction: hedonic price model vs. artificial neural network,” in New Zealand Agricultural and Resource Economics Society Conference, 2004, pp. 25–26.
[20] T. D. Phan, “Housing price prediction using machine learning algorithms: The case of melbourne city, australia,” in 2018 International Conference on Machine Learning and Data Engineering (iCMLDE). IEEE, 2018, pp. 35–42.
[21] X. Chen, L. Wei, and J. Xu, “House price prediction using lstm,” arXiv preprint arXiv:1709.08432, 2017.
[22] J. Rostami and F. Hansson, “Time series forecasting of house prices: An evaluation of a support vector machine and a recurrent neural network with lstm cells,” 2019.
[23] R. Liu and L. Liu, “Predicting housing price in china based on long short-term memory incorporating modified genetic algorithm,” Soft Computing, vol. 23, no. 22, pp. 11 829–11 838, 2019.
[24] A. S. Temür, M. Akgün, and G. Temür, “Predicting housing sales in turkey using arima, lstm and hybrid models,” Journal of Business Economics and Management, vol. 20, no. 5, pp. 920–938, 2019.
[25] L. Yu, C. Jiao, H. Xin, Y. Wang, and K. Wang, “Prediction on housing price based on deep learning,” International Journal of Computer and Information Engineerin, vol. 12, no. 2, pp. 90–99, 2018.
[26] C. Ge, “A lstm and graph cnn combined network for community house price forecasting,” in 2019 20th IEEE International Conference on Mobile Data Management (MDM). IEEE, 2019, pp. 393–394.
[27] Y. Piao, A. Chen, and Z. Shang, “Housing price prediction based on cnn,” in 2019 9th International Conference on Information Science and Technology (ICIST). IEEE, 2019, pp. 491–495.
[28] S. Law, B. Paige, and C. Russell, “Take a look around: using street view and satellite images to estimate house prices,” ACM Transactions on Intelligent Systems and Technology (TIST), vol. 10, no. 5, p. 54, 2019.
[29] X. Fu, T. Jia, X. Zhang, S. Li, and Y. Zhang, “Do street-level scene perceptions affect housing prices in chinese megacities? an analysis using open access datasets and deep learning,” PloS one, vol. 14, no. 5, p. e0217505, 2019.
[30] E. Ahmed and M. Moustafa, “House price estimation from visual and textual features,” arXiv preprint arXiv:1609.08399, 2016.
[31] Q. You, R. Pang, L. Cao, and J. Luo, “Image-based appraisal of real estate properties,” IEEE transactions on multimedia, vol. 19, no. 12, pp. 2751–2759, 2017.
[32] O. Poursaeed, T. Matera, and S. Belongie, “Vision-based real estate price estimation,” Machine Vision and Applications, vol. 29, no. 4, pp. 667–676, 2018.
[33] J. Bin, B. Gardiner, Z. Liu, and E. Li, “Attention-based multimodal fusion for improved real estate appraisal: a case study in los angeles,” Multimedia Tools and Applications, vol. 78, no. 22, pp. 31 163–31 184, 2019.
[34] M.T. Luong, H. Pham, and C. D. Manning, “Effective approaches to attention-based neural machine translation,” arXiv preprint arXiv:1508.04025, 2015.
[35] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” in Advances in neural information processing systems, 2017, pp. 5998–6008.
[36] O. Firat, K. Cho, and Y. Bengio, “Multiway, multilingual neural machine translation with a shared attention mechanism,” arXiv preprint arXiv:1601.01073, 2016.
[37] P.Y. Huang, F. Liu, S.R. Shiang, J. Oh, and C. Dyer, “Attention-based multimodal neural machine translation,” in Proceedings of the First Conference on Machine Translation: Volume 2, Shared Task Papers, 2016, pp. 639–645.
[38] J. K. Chorowski, D. Bahdanau, D. Serdyuk, K. Cho, and Y. Bengio, “Attention-based models for speech recognition,” in Advances in neural information processing systems, 2015, pp. 577–585.
[39] J. Chorowski, D. Bahdanau, K. Cho, and Y. Bengio, “End-to-end continuous speech recognition using attention-based recurrent nn: First results,” arXiv preprint arXiv:1412.1602, 2014.
[40] N. Moritz, T. Hori, and J. Le Roux, “Triggered attention for end-to-end speech recognition,” in ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2019, pp. 5666–5670.
[41] S. Watanabe, T. Hori, S. Kim, J. R. Hershey, and T. Hayashi, “Hybrid ctc/attention architecture for end-to-end speech recognition,” IEEE Journal of Selected Topics in Signal Processing, vol. 11, no. 8, pp. 1240–1253, 2017.
[42] Q. You, H. Jin, Z. Wang, C. Fang, and J. Luo, “Image captioning with semantic attention,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 4651–4659.
[43] X. Zhou, Y. Shen, Y. Zhu, and L. Huang, “Predicting multistep citywide passenger demands using attention-based neural networks,” in Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining. ACM, 2018, pp. 736–744.
[44] L. Gao, Z. Guo, H. Zhang, X. Xu, and H. T. Shen, “Video captioning with attention-based lstm and semantic consistency,” IEEE Transactions on Multimedia, vol. 19, no. 9, pp. 2045–2055, 2017.
[45] Z. Yang, X. He, J. Gao, L. Deng, and A. Smola, “Stacked attention networks for image question answering,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 21–29.
[46] M. Jaderberg, K. Simonyan, A. Zisserman et al., “Spatial transformer networks,” in Advances in neural information processing systems, 2015, pp. 2017–2025.
[47] D. Bahdanau, K. Cho, and Y. Bengio, “Neural machine translation by jointly learning to align and translate,” arXiv preprint arXiv:1409.0473, 2014.