x Metro • Travel time • Price Car • Travel time • Price Individual characteristics z • Age • Gender • Commuter pass • Car license • Purpose Utility 10 20 Choice probabality 0.3 0.7 Choice Moldeing =「効⽤関数 f の定義」と「選好パラメータ β の推定」 Choice result y
the Age of Machine Learning.” arXiv. モデル︓ • ASU-DNN [Wang+, 2020] “Deep Neural Networks for Choice Analysis: Architecture Design with Alternative-Specific Utility Functions.” Transportation Research Part C • L-MNL [Sifringer+, 2020] “Enhancing Discrete Choice Models with Representation Learning.” Transportation Research Part B • TasteNet-MNL [Han+, 2020] “A Neural-Embedded Choice Model: TasteNet-MNL Modeling Taste Heterogeneity with Flexibility and Interpretability.” arXiv. http://arxiv.org/abs/2002.00922 • ResLogit [Wong+, 2021] “A Residual Neural Network Logit Model for Data-Driven Choice Modelling.” Transportation Research Part C • Neural-Embedded LCCM [Han, 2019] “Neural-Embedded Discrete Choice Models.” Massachusetts Institute of Technology. https://dspace.mit.edu/handle/1721.1/124207?show=full?show=full • Amortized-MXL [Rodrigues, 2020] “Scaling Bayesian inference of mixed multinomial logit models to large datasets.” Transportation Research Part B
と同じ著者) [Van Cranenburgh and Kouwenhoven, 2021] 学習したNNモデルからVOTを計算して 分布を出している [Alwosheel+, 2020] CV系の解釈⽅法を転⽤ どの変数が予測にどれだけ寄与しているか [Wang+, 2020] Deep neural networks for choice analysis: Extracting complete economic information for interpretation. Transp. Res. Part C [Van Cranenburgh and Kouwenhoven, 2021] Using artificial neural networks for recovering the value-of-travel-time distribution, in: Advances in Computational Intelligence, International Work-Conference on Artificial Neural Networks, pp. 88–102. [Alwosheel+, 2020] Why did you predict that? Towards explainable artificial neural networks for travel demand analysis. Transp. Res. Part C 補⾜︓選択モデルの予測性能・解釈性能向上の3アプローチ ① 効⽤関数をブラックボックス化 ② 効⽤関数の⼀部をブラックボックス化 ③ 効⽤関数をホワイトボックスのまま改良 効⽤関数の特定(変数の組合せや⾮線形関数の特定) [Ortelli+, 2021] どの変数を使うか,変数の対数をとる・⼆乗するか,などを最適化 変数の数と対数尤度にはトレードオフの関係があるので多⽬的最適化として解く [Ortelli+, 2021] Assisted specification of discrete choice models. Journal of Choice Modelling