Application of machine learning and IoT to enable child safety at home environment

Safety of children is of utmost importance in any home environment. IoT when combined with machine learning is found to ofer tremendous benefts in creating smart and safe homes to the society. The aim of this research is to, apply machine learning models, in order to detect the anomaly on the dataset gathered from three IoT devices. The vironmental parameters for which the anomaly is detected are smoke emission, light illumination, LPG gas emission, CO emission, motion detec- tion, humidity changes and temperature-level changes. The research makes use of three machine learning models namely K-Means clustering, Isolation Forest and
Inter-Quartile Range to detect anomalies. In addition to that, it also uses Facebook Prophet Model to predict the daily trends in the data predicted by the three models. The evaluation of performance shows that the ccuracyofpredicting anomaly is greater for the Inter-quartile range model when compared with that of the remain- ing two machine learning models. The accuracy obtained by the IQR model is 99% whereas the models K-means and Isolation Forest render an accuracy of 94% each. The study also provides a scheme of a hardware as a part of the future work that could be implemented in order to implement child safety in a better way in the near future.