Rice yield prediction using adaptive Neuro-fuzzy inference system (ANFIS)

Abstract

In agriculture yield prediction is toughest task around the globe. The agriculture yield depends on various factors such as water, weather, soil characteristics, crop rotation, pest, disease etc., This paper presents a model designed using Adaptive Neuro-Fuzzy Inference System (ANFIS) to predict the yield of rice. ANFIS helps to determine the incompleteness in decision making made by human expert using the learning mechanism. Fuzzy inference system and neural network are combined in ANFIS, the input parameters are passed through input layer and output could be viewed through output layers. Training is involved with iterative adjustment of parameters of the ANFIS using hybrid learning process to diagnosis the yield of rice. ANFIS uses five layers, each layer has its own nodes. Layer1 has the input variables with membership function. T-norm operator that perform the AND operator can be used in layer2. The sum of all rules firing strengths are assigned in layer3. The nodes in layer4 are adaptive and perform the consequent of the rules. Single node that computes the overall output in layer5. With the input parameters Leaf Folder pest incidence(LFI), Sheath Blight disease (SB), Number of Tillers Hill(NH), No. of grains per panicle(GP) and 1000 grain weight(GW) the algorithm is developed to predict the yield of rice. The proposed Fuzzy Prediction Model is effectively “hand crafted” to achieve the desired performance and also used for rice yield prediction.