Predicting Next-Month Passenger Traffic at Taiwan Railway Stations to Optimize Restaurant Food Preparation

Project Details

Term: 

Fall 2023

Students: 

Hsiang-Jung Cheng, Yu-Kang Lai, Shu-Ting Chen, Li-En Tsai

University: 

NTHU

Report: 

Recorded Presentation

Small and medium-sized restaurants near train stations face challenges in accurately predicting daily customer numbers, leading to food waste and operational inefficiencies. Unpredictable fluctuations in customer demand and the need to balance food freshness and waste prevention create a dilemma for these establishments. To address this, our goal is to enhance food preparation processes by providing strategic recommendations that minimize waste and optimize readiness for varying customer flows targeting small restaurants near the train stations.

We utilized Daily Passenger Traffic data from Taiwan Railway stations available at https://data.gov.tw/dataset/8792, focusing on four stations in the Hsinchu area, Hsinchu, Zhunan, Zhudong, and Neiwan. We reframed the data with a new metric, totalPass, representing the sum of entry and exit counts. Data from 2017 to most recent data Oct 2023, was prioritized to align with our business goal. Time plots illustrate daily passenger counts for selected stations. We decompose time series, analyze the impact for COVID-19, employ training models with data spans Nov 2017 to Oct 2022 and data spans Nov 2022 to Oct 2023 as validation. Our performance evaluation emphasizes minimizing over-forecasting in response to our stakeholders’ needs. Our evaluation metrics and forecasting plot using data in Nov 2023 revealed that the forecasting can provide fine prediction for validation.

To aid restaurant decision-making, forecasts will be provided by 2 pm each Sunday. Moreover, visual representation of forecasted passenger numbers, similar to Google Maps' "Popular Times graph" will enhance accessibility. Besides, original value and percentage for better readability will be provided to enhance restaurant’s intuitiveness to make decisions. Future improvements include real-time updates, collaboration with the Taiwan Railway Administration for weekly data, and incorporating sales data to enhance forecast precision. The ultimate goal is to extend the forecasting system nationwide to benefit restaurants surrounding all train stations in Taiwan, improving operational efficiency on a broader scale.

Application Area: