Artificial intelligence algorithms for passenger forecasting at Changi Airport

• 主要作者: Chan, Li Long
• 其他作者: Sameer Alam
• 格式: Final Year Project
• 語言: English
• 出版: Nanyang Technological University 2020
• 主題: Business::Management::Forecasting; Engineering::Aeronautical engineering
• 在線閱讀: https://hdl.handle.net/10356/141138

Explanatory Variables can be provided to best forecast Changi Airport Arrival Passenger Frequencies. These explanatory variables can be Econometric: (GDP/Oil Price/ CPI etc) or they can be other causal time series data like Google Trend Query Frequencies. Explanatory variables that are identified as causal can then be used in a Regression Model for forecasting future Arrival Passenger Frequencies. In this Final Year Project, 2 Datasets (Econometric Variables and Google Trend Queries) are used for forecasting Changi Airport Arrival Passenger Frequencies. 27 Econometric Variables and 688 Google Trend Queries were found to be Linearly Granger Causal to Arrival Passenger Frequencies. A further new type of Granger Causality test named Neural Granger Causality, shows consistency in identifying possible non-linear causal explanatory variables (Econometric and Google Trends). 3 types of regression models were compared against each other: Linear Regression, SARIMAX and Neural Networks. The best forecasting model is the Neural Network Forecasting model with Google Trend Queries as explanatory variables. It achieved an R^2 value of 0.89. Neural Network models were also found to possess a “variance-accuracy trade-off” characteristic in the forecasting results and this is highly likely due to the randomised weights initialization at the start of the training.