ARUBA

该数据集来源于真实公寓和住宅中的实际居民日常生活活动记录，数据包括温度和二进制传感器数据，以及动作和门传感器数据。值得注意的是，该数据集存在不平衡性，某些活动的数据表示不如其他活动充分，并且包含大量未标记的事件，这些事件被归类为“其他”。规模上，该数据集涵盖了数月标记的活动记录。其任务是对智能家居环境中的活动进行识别。

This dataset is derived from real-world daily activity records of residents living in actual apartments and homes. The collected data includes temperature readings, binary sensor data, motion sensor data, and door sensor data. Notably, this dataset exhibits class imbalance, with some activities being underrepresented relative to others, and it contains a large number of unlabeled events categorized as "other". In terms of scale, the dataset spans several months of labeled activity records. The core task associated with this dataset is activity recognition in smart home environments.

搜集汇总

背景与挑战

背景概述

ARUBA数据集是智能家居活动识别研究的一部分，与Milan数据集一起由CASAS提供。数据集经过预处理，支持多种神经网络模型训练，其中FCN + Embedding模型在多种窗口大小下表现出最优性能。