BlueEdge neural network approach and its application to automated data type classification in mobile edge computing.

Owing to the increasing number of IoT gadgets and the growth of big data, we are now facing massive amounts of diverse data that require proper preprocessing before they can be analyzed. Conventional methods involve sending data directly to the cloud, where it is cleaned and sorted, resulting in a more crowded network, increased latency, and a potential threat to users’ privacy. This paper presents an enhanced version of the BlueEdge framework—a neural network solution designed for the automated classification of data types on edge devices. We achieve this by utilizing a feed-forward neural network and optimized features to identify the presence of 14 distinct data types. Because of this, input data can be preprocessed near its source, and not in the cloud. We utilized a comprehensive dataset comprising 1400 samples, encompassing various data formats from around the world. Compared with rule-based methods, experimental assessment achieves better performance, and results in reduced data transmission (reduced by 62%) and processing latency (78 times faster than cloud-based systems), with resource efficiency comparable to low-end mobile devices. Additionally, our strategy demonstrates strong performance under various data conditions, achieving accuracy levels of over 85% on datasets that may include variations and a noise level as high as 20%. The approach used here is capable of processing data for IoT devices used in education, which can lead to more efficient connections with the cloud and better privacy preservation.