Event-triggered reinforcement learning-based internet data bandwidth allocation technique as a metric for balanced QoS and QoE

<span lang="EN-US">This research work studies the performance and management of the internet services of institutions of higher learning in Nigeria. Data were collated from a federal, state, and private university designated as FEDERAL1, STATE1, and PRIVATE1, respectively, in this research study. The reinforcement learning-based internet data bandwidth allocation model was developed for bandwidth allocation and prediction to enhance balanced quality of service (QoS) and quality of experience (QoE) of the users. The linear Lagrange’s method of interpolation, the LILARINT model, was developed and implemented to predict and allocate effective internet data bandwidth for the significantly increasing number of internet users in each of the institutions. The problem of inability to predict and allocate acceptable internet data bandwidth with the corresponding number of internet users was solved by the LILARINT model. The Allen’s PRESS regression, <em>R</em><sup>2</sup> of the LILARINT models, was very close to unity, which is an indication that the models developed stood at the very best fit. In this research work, it is clear that PRESS regressions, <em>R</em><sup>2</sup> for the selected institutions, were better than the regression, <em>R</em><sup>2</sup> obtained from Nielsen’s institution. Using the measured and simulated results, we found out that PRESS regression, <em>R</em><sup>2</sup> has significantly performed best in the LILARINT model developed. In the overall comparative analysis, the FEDERAL1 LILARINT model emerged as the most reliable model developed and implemented. The model has a regression, <em>R</em><sup>2</sup> of 0.9999, mean squared error (MSE) of 1.455, mean absolute deviation (MAD) of 122.6920, Standard Deviation (σ) of 8.2975, and mean absolute percentage error (MAPE) of 0.6274%.</span>