In this letter, a novel deep learning-based IoT device transmission interval management scheme for enhanced scalability that reduces the redundancy data measurement in LoRa networks is proposed. For this purpose, a Local LSTM prediction model of each cluster is proposed in which the devices are clustered based on the features of the extracted data using an autoencoder. By adjusting the device transmission interval based on the prediction results, the amount of redundantly collected traffic in the LoRa environment is reduced. The proposed scheme is validated using a simulation-based experiment with the Intel lab IoT dataset. Here, we consider the physical characteristics of LoRa and the data pattern of Intel lab data. As a result, the scalability of the proposed scheme can be improved by 31% on average with a 0.3 MAPE prediction error threshold compared to the base model to which the proposed scheme is not applied.

中文翻译：

---

一种基于深度学习的新型物联网设备传输间隔管理方案，用于增强 LoRa 网络的可扩展性


在这封信中，提出了一种新颖的基于深度学习的物联网设备传输间隔管理方案，以增强可扩展性，减少 LoRa 网络中的冗余数据测量。为此，提出了每个集群的本地 LSTM 预测模型，其中使用自动编码器根据提取数据的特征对设备进行集群。通过根据预测结果调整设备传输间隔，减少LoRa环境中冗余收集的流量。使用英特尔实验室物联网数据集进行基于模拟的实验来验证所提出的方案。在这里，我们考虑 LoRa 的物理特性和英特尔实验室数据的数据模式。结果，与未应用该方案的基础模型相比，该方案的可扩展性在 0.3 MAPE 预测误差阈值下平均提高了 31%。