Application of artificial neural networks and random forest algorithms for forecasting marketing security among agarwood growers in Assam

Benefits earned from a single agar (Aquilaria malaccensis) plant make it unique and popular among growers in Assam. The present study was conducted in 2024 to assess the marketing security of agar growers using different classifiers viz., artificial neural network (ANN) and random forest. A total of 420 agar growers were selected following the snowball sampling technique. The findings indicated that 45.95 % of the agar growers had a medium level of marketing security. Results of multiple regression analysis showed that 91.55 % variation in the marketing security of agar growers was explained by seven independent variables, namely age, family occupation, annual income, experience in agar cultivation, indebtedness, economic motivation and access to extension services. As per stepwise regression analysis, eight variables viz., age, family occupation, annual income, experience in agar cultivation, indebtedness, economic motivation, access to extension services and risk orientation had a high influence on marketing security, with an R2 value of 96.36 %. The multilayer perceptron neural network (MPN) model indicated that five independent variables viz., age, annual income, economic motivation, access to extension services and risk orientation, generated the greatest number of activated neurons influencing the marketing security of agar growers through the hidden layer H (1:1) and H (2:3). The normalized graph depicted that annual income had the maximum influence on the marketing security of agar growers. The random forest classifier established that age was the most influential variable.