A robust regression model based on optimal feature set for simple decision making in indoor farms

Abstract

This paper proposed a robust regression model for simple decision making in smart indoor farms. In our proposal, there are several steps to ensure the time-series data set which collected from sensor nodes in smart indoor farms are expanded to its features into new data set. The step tries to maximize features, then high corelated features with outcome in new data set will be filtered with strong threshold value. Moreover, we use statistical tests to remove the features in original regression model for finding out the final model. The approach not only interprets curve fitting but also produces small features for equation in the final equation. Simulation results shown that R-square value of the final model is close to R-squared value of original model while outcome in the final equation just depends on small features. The results shown that our proposal can make optimized decisions making in practical applications of agricultural systems.