Prediction in a Dynamic System using Nodal Patterns

Nowadays, communication networks have become an important part of our daily lives. The communication between any two points happens with the transfer of data between the nodes. But due to the non-functionality, non-availability, and long passway of nodes, the transmission of data over the networks is consuming an excessive amount of energy. It would be better if the behavior of nodes is predicted in the passway before the transmission of data. Our idea is to predict the patterns between two nodes and find out the possibilities of knowing the behavior of nodes based on their delay characteristics. Since communication between networks happens in milliseconds, it is too fast to discuss this idea in a highly dynamic system like this. So, we opted for a similar type of lower dynamic system which can reflect the characteristics of the telecommunications system. In this study, we have employed the Blekinge traffic data provided by NetPort Science park and predicted the pattern between two consecutive stations. In the initial step, we had investigated the dataset and found missing points in it. To fill the gaps, present in the data, we had determined different methods for reconstructing the data. Among them, the spline method in the interpolation technique is providing the best and stable results compared to others. Later, we had modeled a Long-Short-Term Memory (LSTM) neural network by dividing the 80% of the dataset for the training and 20% of the dataset for the testing. The results show that there is an important need for pre-processing the data with an appropriate reconstructing method before modeling any system. The pre-processing has a great effect on the accuracy of the model. The obtained results are indicating that the proposed model's training time is short and suitable for modeling the high dynamic systems