Mining of soil data for predicting the paddy productivity by machine learning techniques

Crop yield prediction is a challenging task towards precision agriculture. In particular, paddy is one of the world’s significant cereal crops and thus crucial for crop management and decision making. Despite the number of crop yield prediction models, better performance in paddy yield prediction is still desirable. Keeping this in mind, the present study aimed to determine the most influencing features that impact paddy production. We employed a machine learning algorithm alongside the best data sources for paddy yield prediction in this study. A total of 5 regression machine learning algorithms were developed using the 16 input variables obtained from the soil health card. Note that we have carried out multiple approaches to improving the model performances. The model results were also validated using Monte Carlo methods. The result from our analysis depicts that XG boost ensembled random forest has demonstrated the highest prediction accuracy of 86% of the other models investigated in our study. It is worth mentioning that this is the first study on paddy crop yield prediction from the features of a soil health card. Indeed, farmers and agronomists could use this model to plan their paddy cultivation and procure the maximum yield.