Classification and detection of inconsistent data from a crash report in field of a GPON broadband access network using Artificial Intelligent Algorithms

In the dynamic landscape of GPON broadband access networks, the need for robust and efficient mechanisms to handle inconsistent data from crash reports is paramount. The first scenario is to create a similarity group (cluster) on a Kohonen map on a real basis of a GPON massive failure analysis report, find the discrepant data, investigate possible analysis errors by maintenance specialists. The second objective is to automate the prediction of data traffic behavior on real broadband network access interfaces using the LSTM recurrent neural network. The average error found was 24.6%, MSE 31.41%, relative RMSE was 0.49%, RMSE 56.05%, $\mathrm{R}^{2}$ was 96.17% and R was 98.07%. The average error, MSE, RMSE, $\mathrm{R}^{2}$ and R obtained were 24.6%, 31.41%, 0.49%, 56.05%, 96.17% and 98.07% respectively.