IoT-Based Localization in Urban LoRaWAN Networks Using CNN–LSTM: Supporting Dataset and Model

This dataset and trained model accompany the article “IoT-Based Localization in Urban LoRaWAN Networks Using CNN–LSTM Deep Learning Model” accepted for publication in the Journal of Universal Computer Science (JUCS). The dataset includes two preprocessed RSSI feature sets (processed_dataset_final and processed_dataset_quantile) derived from the LoRaWAN measurements collected in Antwerp, Belgium, covering 72 static gateways and multiple mobile nodes. The provided trained model file (cnn_lstm_v4.keras) corresponds to the final version (CNN–LSTM v4) described in the paper, optimized using QuantileTransformer and RobustScaler preprocessing, Mish activation, batch normalization, and AdamW optimizer. Researchers may use these files to reproduce, validate, or extend the reported results in urban IoT localization tasks.本数据集与配套训练模型，对应已被《通用计算机科学期刊》（Journal of Universal Computer Science，JUCS）接收发表的论文《基于物联网的城市LoRaWAN网络本地化：采用CNN–LSTM深度学习模型》。 该数据集包含两份经预处理的接收信号强度指示（Received Signal Strength Indicator, RSSI）特征集——processed_dataset_final与processed_dataset_quantile，其数据源自比利时安特卫普市采集的LoRaWAN网络测量数据，覆盖72个静态网关与多个移动节点。 本次提供的训练模型文件cnn_lstm_v4.keras，对应论文中所述的最终版本CNN–LSTM v4。该模型采用分位数变换器（QuantileTransformer）与鲁棒标准化器（RobustScaler）进行预处理，使用Mish激活函数、批量归一化（batch normalization）以及AdamW优化器完成训练优化。 研究人员可借助上述文件，复现、验证或拓展该论文中报道的城市物联网本地化任务相关研究成果。