A Layered Features Analysis in Smart Farm Environments

Calculating or predicting the harvest yield of crops in agriculture has been one of its most important techniques. Smart farms, with the power of IoT, provide invaluable information such as real-time environment data such as temperature, humidity, etc. However, there are situations where not all data can be sent over the internet due to private reasons. In this paper, we propose a scalable data analysis framework: the edges preprocess and analyze the private data and send the results to server, and the server gathers and accumulates the results to estimate and predict the total harvest yield. Based on the results obtained from a real tomato farm, error rate is comparable to the one performed at the server only, but the number of features is reduced significantly.