Configuration of Efficient Returning Farmers Data Set for Algorithms Validation based on ANN and Random Forest

Since 2010, as the number of urban residents returning to farming and returning to rural areas has increased, various policies and service models such as education have been supported. However, as the number of failures and dissatisfaction cases for returning to farming and returning home increases, it is urgent to prepare a support service model. After all, in addition to farming technology, it is necessary to collect and prepare a lot of information, such as selecting competitive crops, needing to check how to secure housing/farmland, and recognizing legal process information such as registration of farmers/businesses. Therefore, in this paper, algorithm verification was performed to improve the importance of key variable items in order to efficiently compose the returnee data set. To this end, the algorithm was verified to be able to implement a service model that can recommend regions, items, and information with high reliability to returning farmers by applying artificial neural networks and random forest techniques. The artificial neural network and random forest technique were applied as methods for deriving effective variables and validating algorithms to secure the reliability of the returnee data set, which is the goal of the study. For this purpose, algorithm verification was performed using Ridge regression and Lasso regression among artificial neural network techniques. And, algorithm verification was performed using IncMSE and IncPurity methods among random forest techniques. In addition, negative binomial distribution regression was additionally applied to increase the reliability of the verification. Algorithm verification results for deriving effective variables and measuring importance of the returnee data set, a total of five variables that obtained relatively high scores from all methods were derived: the number of direct sales stores, the land area, the number of libraries, the number of hospitals/clinics, and the number of specialized retail businesses.