Home Page of RB-Pred

Rice blast (Pyricularia grisea) continues to be the most destructive disease of rice despite decades of research towards its control. Weather has a very important role to play in the appearance, multiplication and spread of the blast fungus. Weather-based forecasting systems reduce the cost of production by optimizing the timing and frequency of application of control measures and ensure operator, consumer and environmental safety by reducing chemical usage. Thus, if a sound forewarning system is developed, the explosive nature of the disease could be prevented by timely application of the control measures.

‘RB-Pred’ web-based server, a first of its kind worldwide, is an attempt for forecasting leaf blast severity based on the weather variables which may help the farmers and plant pathologists in timely prediction of rice blast in their areas and ultimately, in their decision making process.

RB-Pred predicts rice leaf blast severity (%) based on the weather parameters input by the user. It uses the regression module called support vector regression (SVR) of a powerful machine learning technique called support vector machine (SVM). The SVM learns how to classify from a training set of feature vectors, whose expected outputs are already known. The training enables a binary classifying SVM to define a plane in the feature space, which optimally separates the training vectors of two classes. When a new feature vector is inputted, its class is predicted on the basis of which side of the plane it maps.

INTRODUCTION