Reducing travel survey cost by utilising cross-sectional data from multiple time points

2018 Fiscal Year Final Research Report

• Project/Area Number: 16K03931
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Commerce
• Research Institution: Kobe University
• Principal Investigator: SANKO nobuhiro 神戸大学, 経営学研究科, 教授 (00403220)
• Project Period (FY): 2016-04-01 – 2019-03-31
• Keywords: 消費者行動 / 交通行動 / 需要予測 / 繰り返し断面データ / 非集計モデル / サンプル数 / 調査頻度 / モデルの更新

Outline of Final Research Achievements

When estimating disaggregate demand models for forecasting, only the most recent data is utilised even when cross-sectional data from multiple time points are available. This is not a good use of the data. This study applied a method, proposed by the author, which improves demand forecasts by utilising data from multiple time points. The study demonstrated that using data from multiple time points, where the number of observations from each time point is smaller, produced statistically significantly better forecasts than using data from the most recent time point. Practical contribution is to improve forecasting performance without any additional cost for the survey.

Free Research Field

交通　需要予測　意思決定　選択モデル　商学　消費者行動　土木計画学

Academic Significance and Societal Importance of the Research Achievements

交通調査には多大な費用を要する．本研究では，複数時点のデータを用いることで予測精度の向上が可能であり，それは調査済みの過去のデータを用いることで追加費用なく，また，各回の調査規模（調査費用）を削減することによっても実現可能であることを示した．また，大規模低頻度調査よりも小規模高頻度調査のほうが望ましい可能性を示した．