Agriculture is believed to be as the backbone of Indian economic system. For the past few decades, agriculture field has seen lots of technological changes to improve better productivity. Day by day the population is increasing leading to increasing demand for resources but the amount of resources required has been reducing and falling down. Therefore, there has been extensive endeavors to create imaginative and technological advances methodologies for manageable harvest generation. Using prediction methods, farmers can enhance the productivity of crops. These strategies are utilized to find the required number of crops, seeds, moistness, water level and other supplements. Since prediction refers to a statement about an uncertain event, hence modeling the prediction would a good solution to adopt. Predictive modeling uses statistics to predict outcomes. Quantifying the yield is essential to optimize policies to ensure food security. This paper aims at providing a new method to predict the crop yield of food crops grown above the ground level based on big-data analysis technology, which differs with traditional methods in the structure of handling data and in the means of modeling. Firstly, the method can make full use of the existing massive agriculture relevant datasets and can be still utilized with the volume of data growing rapidly, due to big-data friendly processing structure. Secondly, the "nearest neighbors"modeling, which employs results gained from the former data processing structure.