The emerging of new models in machine learning enhances the performance of algorithms proposed to address several challenging tasks such as object recognition, classification and identification purpose. Nowadays. the deep learning algorithms are playing a massive role in accurately addressing complex problems due to their capability of learning various complex features from data. One limitation of deep learning is a lack of sufficient data for training. In this study, we proposed variants of Long Short Term Memory (LSTM) model and ensemble learning methods such as XGBoost, AdaBoost, Bagging, Stacking and Random forest. The experimentation is carried out on Time series data generated from the Internet of Things (IoT) devices. To validate the proposed method, we have used a freely available dataset on the web namely Smart-Fall datasets. To measure the performance of the proposed method, we have used standard performance measures namely, accuracy, precision, recall, f-score, specificity, geometric mean and confusion matrix. A set of experimental details are carried out on the SmartFall dataset and the experimental results exhibit that the Random forest algorithm performs better when compared with a single deep LSTM model and different ensemble techniques.

机器学习中新模型的出现增强了为解决一些挑战性任务（例如对象识别，分类和识别目的）而提出的算法的性能。如今。深度学习算法具有从数据中学习各种复杂特征的能力，因此在准确解决复杂问题方面发挥着巨大作用。深度学习的局限性之一是缺乏足够的训练数据。在这项研究中，我们提出了长期短期记忆（LSTM）模型的变体和整体学习方法，例如XGBoost，AdaBoost，Bagging，Stacking和Random forest。实验是从物联网（IoT）设备生成的时间序列数据上进行的。为了验证所提出的方法，我们在网络上使用了免费的数据集，即Smart-Fall数据集。为了衡量所提出方法的性能，我们使用了标准性能指标，即准确性，精度，召回率，f评分，特异性，几何均值和混淆矩阵。在SmartFall数据集上进行了一组实验细节，实验结果表明，与单个深度LSTM模型和不同的集成技术相比，随机森林算法的性能更好。