An Enhanced Design and Development of Weather Prediction Model for Transport System

The research aims to predict the weather condition in Andhra Pradesh evaluated using DNN-LSTM performs compared to its predecessor in predicting flow time series. Additionally, both temporal and geographical characteristics may be present in multiple sets of traffic detectors. This system could eventually be used to learn how weather patterns change depending on the amount of precipitation. An unsupervised learning stack of DNNs is created to rapidly sift through data in search of relevant characteristics. The next DNN layer that may be seen is the concealed layer. Each DNN use its own independent instance of the CD learning algorithm for reconstructing the input. Using an up-down technique using labelled data, a fully connected output layer is built on top of the DNN to fine-tune the overall architecture. Instead of the usual pattern identification or classification tasks, this research uses traffic flow data from a single arterial collected using four sequence detectors with spatial-time correlations. Therefore, it is possible to integrate the four prediction tasks such that they can share the training features and weights. The proposed DNN-LSTM models incorporate both an unattended and monitored stack of DNN-LSTM, and their performance is measured using the most common metrics, such as sensitivity, specificity, prediction accuracy, geometry mean, and classification error. By storing long and short time characteristics in memory blocks and memory cells, the LSTM model improves time series prediction.