Impact of feature cross in hybrid optimization based convolutional neural networks for train delay prediction

Indian Railway is considered the biggest railway network among other networks; it operates passenger and freight trains. The government of India manages Indian Railways, and it is also a critical component of the country's transportation infrastructure. The inconvenience of people may cause a delay, or any issue will affect the schedule of a railway system. Latency is a main issue in the life cycle of every worker used to working on the train. A novel methodology that performs feature cross and prediction is introduced to predict the delay from any situation. The proposed model is trained using a hybridized adaptive Bidirectional long-short memory (BiLSTM) combined with a Convolutional neural network (CNN). These models are comparatively faster in delay prediction and high in performance. However, tuning hyper-parameters also plays a vital role in enhancing accuracy. A Coati optimization algorithm and the hybrid network model are introduced for optimal hyperparameter tuning. Finally, the performance analysis is determined by evaluating standard metrics like accuracy, RMSE, and MAE. This method also produces a better accuracy of 96.66% compared to existing models that utilize feature representation using buckets.