Research Article
Machine Learning and Deep Learning for Predicting Indoor and Outdoor IoT Temperature Monitoring Systems
@INPROCEEDINGS{10.1007/978-3-030-95987-6_13, author={Nur Indah Lestari and Mahmoud Bekhit and Mohamed Ali Mohamed and Ahmed Fathalla and Ahmad Salah}, title={Machine Learning and Deep Learning for Predicting Indoor and Outdoor IoT Temperature Monitoring Systems}, proceedings={IoT as a Service. 7th EAI International Conference, IoTaaS 2021, Sydney, Australia, December 13--14, 2021, Proceedings}, proceedings_a={IOTAAS}, year={2022}, month={7}, keywords={Deep learning Indoor IoT Machine learning temperature prediction Outdoor}, doi={10.1007/978-3-030-95987-6_13} }- Nur Indah Lestari
Mahmoud Bekhit
Mohamed Ali Mohamed
Ahmed Fathalla
Ahmad Salah
Year: 2022
Machine Learning and Deep Learning for Predicting Indoor and Outdoor IoT Temperature Monitoring Systems
IOTAAS
Springer
DOI: 10.1007/978-3-030-95987-6_13
Abstract
Nowadays, IoT monitoring systems are ubiquitous. These systems utilized sensors to measure the temperature indoors or outdoor. These sensors can be temporarily unavailable for several reasons, such as power outages. Thus, the server that collects the temperatures should find an alternative for predicting the temperature during the downtime of temperature sensors. In this context, there are several machine learning models for predicting temperature. This work is motivated to study the performance gap of predicting outdoor and indoor temperatures. In the proposed study, we utilized a deep learning recurrent neural network called Gated Recurrent Units (GRUs) and four machine learning models, namely, random forest (RF), decision trees (DT), support vector machines (SVM), and linear regression (LR) for predicting the temperature during the downtimes of the temperature sensors. Then, we evaluated the proposed models on a realistic dataset. The results show that predicting the indoor temperature is more predictable than the outdoor temperature. Moreover, the results revealed that the SVM model was the most accurate model for this task.
