Improve the efficiency of network traffic in IoT using dimensionality reduction

IoT has changed our lives through the increased convenience of automating mundane tasks, enhancing home security systems, wearable devices to improve health and wellness, and improved connectivity. Vast volumes of data, such as images, are generated and then transmitted to the cloud server for processing or storage. This leads to an increase in latency and network traffic. To address these challenges, the concept of edge computing has been introduced as a computing paradigm that moves processing closer to the point of data generation. This paper includes the merging of cloud and edge computing in the realm of IoT. It explores techniques that utilize machine learning, employing the autoencoder-based deep learning method and principal component analysis, to perform image dimensionality reduction at the edge. Subsequently, the encoded data is transmitted to the cloud server, where it can be directly utilized for various machine learning tasks without significantly impacting the accuracy of the processed data in the cloud. The suggested approach has been assessed on an image classification task using a set of 20,580 images from the ILSVRC dataset. The results indicate that a 61% reduction in data did not significantly affect the accuracy of the image classification task.