Optimizing Crop Recommendation System Using Threshold and Decision-Level Fusion

Nowadays crop recommendation systems face many obstacles in recommending suitable crops. A main challenge that has to be encounter is changes in environmental conditions and handling diverse dataset like soil and climate oriented dataset. Existing algorithm's like decision tree, random forest have limitations in adapting to local variability. To overcome this and to enhance the agricultural activities, this paper presents an innovative crop recommendation system based on threshold (CRST) values, through which crops are chosen by considering properties of soil, climatic parameters, and season factors such as Kharif, Rabi, and Zaid. Critical limit threshold values are set for most important features like fertility of soil, nutrition of the soil, pH, temperature, humidity, and rainfall to ensure optimal growth conditions. In order to recommend the suitable crop, a rule-based decision-fusion method is tied with confidence level analysis. To evaluate the effectiveness of the threshold method, ordinary least squares (OLS) regression is used. CRST model achieves an accuracy of 90% compared to existing algorithms. Integration of seasonal adaptation and threshold determination greatly enhances crop management decision-making and provides a strong answer to agricultural problems.