Traffic Flow Prediction Based on Hybrid Deep Learning Models Considering Missing Data and Multiple Factors

In the case of missing data, traffic forecasting becomes challenging. Many existing studies on flow with data often overlook relationship between imputation and external factors. To address this gap, study proposes two hybrid models that incorporate multiple factors for predicting in scenarios involving loss. Temperature, rainfall intensity whether it is a weekday will be introduced as forecasting. Predictive mean matching (PMM) K-nearest neighbor (KNN) can find are most similar to values interpolation value. module, bidirectional long short-term memory (BiLSTM) network extract time series features, which improve accuracy. Therefore, PMM KNN were combined BiLSTM P-BiLSTM K-BiLSTM forecast flow, respectively. Experiments conducted using dataset from expressway S6 Poland, considering various rates. The experimental results showed proposed outperform other traditional terms prediction Furthermore, consideration working day further improves predictive performance models.