A Token-based Self-Supervised Network on Traffic Flow data
The source code for the pre-print paper Masked Token Enabled Pre-training A Task-Agnostic Approach for Understanding Complex Traffic Flow, which is under review of IEEE Transactions on Intelligent Transportation Systems.
The main scripts are in folder main, which can be directly ran in Pytorch after cloning this repository and getting all the requried dependencies ready. However, necessary data pre-processing is required to get the same data shape. The pre-processing scheme can be found in our paper.
The model structure is in folder model. The code for pretrain and finetune phase is listed in pretrain and finetune respectively.
The comparison models of LSTM and Transformer are in LSTM and transformer.
-
A novel network, i.e. TSSN, is proposed for generating an effective task-agnostic model for various downstream tasks on traffic flow data.
Meanwhile, a novel pretext task, i.e. masked token prediction(MTP), is designed to provide strong surrogate supervision signals for the pre-training of TSSN.
-
Three types of downstream tasks, i.e. TF classification, prediction and completion, are solved by using the representations of tokens created in pre-training model.
The datasets for pre-training and TF prediction and completion tasks are collected from PeMS, and those used for TF classidication task is Seattle Inductive Loop Detector Dataset.