一個好的據點位置對於企業未來業務發展的成功與否影響重大，例如銀行 、金控業者會投入許多人力與實體調查尋找最佳的分行設立據點，但這些傳統方法仍然效益不彰，原因是太多因素必須考量，且很難綜合這些內外部因素去評估一個據點的設立地點是好或壞。近年來由於城市中可取得的資料越來越多樣化，例如：地理、人口、人車流、都市計畫、道路結構等，許多研究開始運用部分資料對於站點位置的影響進行探索，然而大多數的研究忽略了如何綜合考量這些因素以及他們潛在複雜的交互特徵，因此無法得到很好的準確度；又或者是這些研究只能針對單一重要銀行指標進行預測，不符合現實站點位置選擇的多樣需求。因此本研究提出 MATE模型，針對銀行的多個績效進行多任務人工智慧深度學習，使模型能夠同時考量不同的績效表現來產生預測。由於我們的模型受益於決策樹模型的特性，因此可以非常高效的提取複雜交叉特徵，並且提供模型的可解釋性。除此之外，我們額外提出了 Financial Embedding模組，使我們的模型能夠更進一步取得複雜的地區特徵。最終我們的模型相較於其他現有研究所提出之模型，提升了至少 5%~15%的準確度，更提供了 AI模型如何決策的解釋性。本研究建立之站點推薦系統目前已實際應用於輔助國內知名銀行內部站點評估，兼具實務與研究之效益。

Selecting the optimal location for a new bank branch is challenging but worth studying, due to its importance to growing a successful business. Bankers have been devoting a lot of effort to this issue. In recent years, the proliferation of multisource data in smart cities has called for more need of the data-driven optimal location recommendation approach. Existing studies usually focus on mining different features and process these data through domain expert knowledge, which fail to discover potential complicated factors interactions. Besides, most of these studies only use a single indicator to evaluate the candidate locations, which overlooks a lot of information and resulting in an inaccurate evaluation. This paper proposes MATE, a Multi-task Attentive Tree-Enhanced model, to predict multiple performance indicators of a bank branch's location simultaneously. Our model takes advantage of the tree-based model, which can effectively extract cross features and provide interpretability according to inferred decision rules. In addition, we designed the Financial Embedding module and Attentive Interaction block, allowing us to obtain more complex and diverse features. Finally, extensive experiments on real-world bank branches datasets demonstrate the effectiveness of our method outperform baseline and state-of-art methods from 5% to 15%. MATE has been deployed for assisting one of the largest bank companies in Taiwan to deploy new branch locations in all cities in 2020.

[1] Baviera-Puig, J. Buitrago-Vera, C. Escriba-Perez, "Geomarketing models in supermarket location strategies", J. Bus. Econ. Manag, 2016, 17 (6) 1205–1221.
[2] A. Sevtsuk, "Location and agglomeration: The distribution of retail and food businesses in dense urban environments", J. Plan. Educ. Res., 2014, vol. 34, no. 4, pp. 374–393.
[3] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin, "Attention is all you need", In NIPS, 2017, pp. 5998–6008.
[4] B. Cuo, J. Li, V. W. Zheng, Z. Wang, and Z. Yu, "Citytransfer: Transferring inter- and intra-city knowledge for chain store site recommendation based on multi-source urban data", In Proc. ACM Interact. Mobile Wearable Ubiquitous Technol, 2018, vol. 1, no. 4, 135.
[5] Chaoxiong Chen, Chao Chen, Chaocan Xiang, Songtao Guo, Zhu Wang, Bin Guo, "ToiletBuilder: A PU Learning based Model for Selecting New Public Toilet Locations", IEEE Internet of Things Journal, 2020.
[6] D. Karamshuk, A. Noulas, S. Scellato, V. Nicosia, and C. Mascolo, "Geo-spotting: mining online location-based services for optimal retail store placement", In SIGKDD, 2013, 793-801.
[7] D. Misra, "Mish: A self regularized non-monotonic neural activation function", arXiv preprint arXiv:1908.08681, 2019.
[8] F. Wang, L. Chen, and W. Pan, "Where to place your next restaurant? Optimal restaurant placement via leveraging user-generated reviews", In Proc. 25th ACM Int. Conf. Inf. Knowl. Manag, 2016, pp. 2371–2376.
[9] Guolin Ke, Zhenhui Xu, Jia Zhang, Jiang Bian, and Tie-Yan Liu, "DeepGBM: A Deep Learning Framework Distilled by GBDT for Online Prediction Tasks", In SIGKDD, 2019, 384–394
[10] Jiaqi Ma, Zhe Zhao, Xinyang Yi, Jilin Chen, Lichan Hong, and Ed H. Chi, "Modeling Task Relationships in Multi-task Learning with Multi-gate Mixture-of- Experts", In SIGKDD, 2018, 1930–1939.
[11] Jiejie Zhao, Bowen Du, Leilei Sun, Weifeng Lv, Yanchi Liu, Hui Xiong, "Deep multi-task learning with relational attention for business success prediction", Pattern Recognition, 2021, 110 (2021) 107469.
[12] J. Li, B. Guo, Z. Wang, M. Li, and Z. Yu, "Where to place the next outlet? Harnessing cross-space urban data for multi-scale chain store recommendation", In Proc. ACM Joint Conf. Pervasive Ubiquitous Computing, 2016, 149–152.
[13] J. Lian, X. Zhou, F. Zhang, Z. Chen, X. Xie, and G. Sun, "xDeepFM: Combining Explicit and Implicit Feature Interactions for Recommender Systems", In SIGKDD, 2018, 1754–1763.
[14] Jonathan Baxter et al, "A model of inductive bias learning", J. Artif. Intell. Res.(JAIR), 12, 149-198 (2000), 3.
[15] J. Zhao, B. Du , L. Sun , F. Zhuang , W. Lv , H. Xiong, "Multiple Relational Attention Network for Multi-task Learning", In SIGKDD, 2019, 1123–1131.
[16] L. Chen, D. Zhang, G. Pan, X. Ma, D. Yang, K. Kushlev, W. Zhang, and S. Li, "Bike sharing station placement leveraging heterogeneous urban open data", In Proc. of ACM UbiComp’15, 2015, pp. 571-575.
[17] Liyuan Liu, Haoming Jiang, Pengcheng He, Weizhu Chen, Xiaodong Liu, Jianfeng Gao, and Jiawei Han, "On the Variance of the Adaptive Learning Rate and Beyond", In ICLR, 2020.
[18] P. Jensen, "Network-based predictions of retail store commercial categories and optimal locations", Physical Review E, 2006, vol. 74, no. 3.
[19] R Caruna, "Multitask learning: A knowledge-based source of inductive bias", In Machine Learning: Proceedings of the Tenth International Conference, 1993, 41–48.
[20] Rich Caruana, "Multitask learning", In Learning to learn, Springer, 1998, 95–133.
[21] Ruoxi Wang, Bin Fu, Gang Fu, and Mingliang Wang, "Deep & Cross Network for Ad Click Predictions", In AdKDD, 2017, 12:1-12:7.
[22] Sebastian Ruder, "An overview of multi-task learning in deep neural networks", arXiv preprint arXiv:1706.05098, 2017.
[23] Sundong Kim, Yu-Che Tsai, Karandeep Singh , Yeonsoo Choi , Etim Ibok, Cheng-Te Li , Meeyoung Cha, "DATE: Dual Attentive Tree-aware Embedding for Customs Fraud Detection", In SIGKDD, 2020, 2880-2890.
[24] Tianqi Chen and Carlos Guestrin, "XGBoost: A Scalable Tree Boosting System", In SIGKDD, 2016 785–794.
[25] XiangWang, Xiangnan He, Fuli Feng, Liqiang Nie, and Tat-Seng Chua, "TEM: Tree-enhanced Embedding Model for Explainable Recommendation", In WWW, 2018 1543–1552.
[26] Xiangyu Zhang, Xinyu Zhou, Mengxiao Lin, and Jian Sun, "ShuffleNet: An extremely efficient convolutional neural network for mobile devices", In CVPR, 2018, 6848–6856.
[27] Xinran He, Junfeng Pan, Ou Jin, Tianbing Xu, Bo Liu, Tao Xu, Yanxin Shi, Antoine Atallah, Ralf Herbrich, Stuart Bowers, and Joaquin Quiñonero Candela, "Practical Lessons from Predicting Clicks on Ads at Facebook", In AdKDD, 2014, 5:1–5:9.
[28] Yan Liu, Bin Guo, Nuo Li, Jing Zhang, Jingmin Chen, Daqing Zhang, Yinxiao Liu, Zhiwen Yu, Sizhe Zhang, and Lina Yao, "DeepStore: An Interaction-Aware Wide&Deep Model for Store Site Recommendation With Attentional Spatial Embeddings", IEEE Internet of Things Journal 6, 4 (2019), 7319–7333.
[29] Z. Yu, M. Tian, Z. Wang, B. Guo, and T. Mei, "Shop-Type Recommendation Leveraging the Data from Social Media and Location- Based Services", In TKDD, 2016, vol. 11, no. 1, 1-21.