Detecting Anomalies in Environmental Data With Jetson Nano

This paper evaluates the effectiveness of various machine learning and statistical analysis methods used to detect anomalies in environmental data. The key focus is on the adaptability, computational complexity, and performance of each of the methods under consideration, namely: isolation forests, support vector machines (SVMs), autoencoders, LSTM neural networks, and statistical anomaly detection methods. This analysis is based on the study of the possibilities of applying the considered methods on the Jetson Nano embedded computing platform. The paper emphasizes the challenges associated with optimizing computing resources and adaptability of algorithms in this environment. It identifies prospects for further research and provides methodological recommendations for improving the results. This research aims to identify the advantages of a hybrid approach in applying machine learning methods together with statistical analysis to effectively detect anomalies in environmental data. The hybrid approach compensates for the weaknesses of individual methods and emphasizes the use of their strengths to achieve superior anomaly detection accuracy. The work also proved that machine learning models can be effectively used to track potential environmental pollution risks and adapt to different conditions and scenarios. The next step of this research was to establish the need to optimize and adapt the algorithms for embedded computing and application on the Jetson Nano platform, as this will facilitate the development of robust and flexible anomaly detection systems. The result is that the study of current anomaly detection methods is an important area in the field of environmental data monitoring and efficient use of resources of modern energy systems.