arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2403.15461eess.SPphysics.optics
收藏

主成分分析与人工神经网络在南非自由空间光链路服务质量预测中的应用

Application of Principal Component Analysis and Artificial Neural Networks for the Prediction of QoS in FSO Links over South Africa

S. O Adebusola, P. A Owolawi, J. S Ojo, P. S Maswikaneng, A. O Ayo

AI总结:

本文利用主成分分析与人工神经网络,基于南非五地十年气象数据构建FSO链路QoS预测模型,并在乔治取得最佳RMSE表现,为南非电信业务优化提供实用工具。

摘要:

自由空间光通信(FSO)相比射频(RF)能够提供更多的无线带宽,在免费许可下运行,并且启动成本更低。然而,其对大气气象条件变化的脆弱性是一个值得关注的问题。本研究的最终目的是利用主成分分析(PCA)与人工神经网络(ANN)为地面FSO通信连接设计一个服务质量(QoS)预测模型。为实现既定目标,从南非气象局(SAWS)档案中收集了十年期间五个不同地点——乔治、约翰内斯堡、金伯利、布隆方丹和波罗克瓦尼——的能见度、风速和海拔等气象数据。在布隆方丹、约翰内斯堡、金伯利、乔治和波罗克瓦尼各站点,PCA中第一主成分(PC1)和第二主成分(PC2)的特征值分别为7.624和1.020、7.234和0.984、6.204和1.723、7.354和0.876、7.104和0.865,表明它们被保留为QoS变量以训练人工神经网络(ANN)模型,因为它们提供了对原始变量数据最有说服力的解释。所有研究地点每个所提模型的RMSE值分别为0.1437、0.2131、0.2329、0.1101和0.1977。基于RMSE,所提模型在乔治表现更好。为每个研究地点开发了一个现实且准确的预测模型。因此,所开发的模型将成为在南非维持FSO网络服务良好QoS和改善电信业务的重要工具。

英文摘要:

Optical Communication in Free Space (FSO) bids more radio bandwidth, operates under a gratis license, and has a lower startup cost as compared to Radio Frequency (RF). Nonetheless, its vulnerability to variations in atmospheric meteorological circumstances is a concern. Ultimately, the purpose of this study is to use Principal Component Analysis (PCA) with Artificial Neural Networks (ANN) to design a QoS prediction model for a terrestrial FSO communication connection. To accomplish the specified goal, meteorological data such as visibility, wind speed, and altitude were collected from the Weather Services in South Africa (SAWS) archive during a ten-year duration at five different locations: George, Johannesburg, Kimberly, Bloemfontein, and Polokwane. The eigenvalues of the first Principal Component (PC1) and the second Principal Component (PC2) in the PCA across the stations Bloemfontein, Johannesburg, Kimberly, George, and Polokwane are 7.624 and 1.020, 7.234, and 0.984, 6.204 and 1.723, 7.354 and 0.876, and 7.104 and 0.865, respectively, demonstrating that, they are kept as QoS variables to train the Artificial Neural Network (ANN) model as they provide the most compelling interpretation of the original variable data. The RMSE values of every proposed model across all the study locations are 0.1437, 0.2131, 0.2329, 0.1101, and 0.1977, respectively. Based on the RMSE, the proposed performed better over George. A realistic and accurate predictive model is developed for each of the study locations. Thus, the developed model will serve as a valuable tool for maintaining good QoS in FSO network services and improving telecom businesses in South Africa.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑