A Machine Learning Technique for Rice Blast Disease Severity Prediction Using K-Means SMOTE Class Balancing

Introduction

Agriculture has become a vital source of food for the world's ever-increasing population. In India, farmers produce agricultural products and there is increased consumption of agriculture products due to the high population growth. In south India, rice is the main ingredient in all the food items prepared daily. Hence, maximum farmers have adapted cultivation of paddy crop than other field and horticulture crops. The yield of the rice crop is affected by the various diseases caused by different agents. Huge loss in rice cultivation is incurred due to the existence of diseases from continuous variation in environmental conditions and pests. The paddy crop may be infected by either fungal or bacterial diseases. Among various diseases, the rice blast disease creates grain losses nearly up to 70 to 80%, according to the report of Gandhi Agriculture University (GAU). Blast disease severity is an important parameter which measures the level of blast disease affected to paddy plant and it can also be used to decide preventive measures. Initially disease affects very small part of the leaf as severity increases disease may affect to the whole plant leading to the major loss of crop production. Accurate and effective prediction of severity of rice blast disease would help farmers to take preventive actions at the early stage. Over the last few decades traditionally, disease severity is decided by farmers or experts’ visual inspection, which leads to poor disease management in precision agriculture. Constant climate change, along with high urbanization pressure and unsustainable agricultural activities, has put ecosystems at risk in gradual deterioration of rice crop production. To reduce the human interaction in agriculture disease management and ensuring food standards proper monitoring is essential. Decision support systems that use weather-based forecasting models can provide important information about blast disease severity management planning. By regulating the time and frequency of control methods, the cost of production and crop losses can be decreased.

In Davangere region, paddy is one of the most significant food crops. Rice blast disease imposed by the Magnaporthe Grasia pathogen, is among the most serious crop disease, resulting in significant production losses. Blast disease management is complicated since it involves several elements. Rice blast disease and severity is reported to be influenced by three factors known as the disease triangle, each of which represents a host system, a pathogenic, and a favorable climate. Adoption of control methods at right time by the farmers would reduce severity of rice blast disease. Machine learning is a fascinating field which has variety of techniques that could be used to design decision support system for controlling the severity of rice blast disease in paddy field.