A combination model based on a neural network autoregression and Bayesian network to forecast for avoiding brown plant hopper

In this paper, we propose a new approach that is to combine a model of time series with Bayesian networks to create a new forecasting model to predict the occurrence of harmful pests of rice. Ability of forecast system knows when immigrant brown plant hopper (BPH) peaks to make a calendar of sowing rice to avoid them. This is indeed helpful for experts as well as farmers to sow rice of seeds actively and simultaneously on a large scale for each new rice crop. Using knowledge of experts and processing historical data combined with data at present time are able to create more highly accurate forecasts than just relying on historical data. This model is a decision support system for experts in the Plant protection centre of the South of Vietnam to guide farmers to sow cultivation in a specific area, pilot for 22 Southern provinces.