Data Mining based Geospatial Clustering for Suitable Recommendation system

Precision Agriculture is equipped with the latest generation GPS technology in relation to crop treatment based on its site-specific sensors. It can be termed as a data-driven method to agriculture that is intensely associated with a series of data mining problems. The geographic position of the data records must be clearly considered when establishing a recommended system and evaluating its predictive performance. A crop recommendation model has been developed for selected crops from the sample data collected from six Blocks (Taluks) of Thrissur district, Kerala State. The challenge was to extract information from these raw data. This work emphasizes the exploration of the agriculture data and generating a rank based recommendation system to identify the most suitable crops for a particular location. DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering method with the ball-tree algorithm is used to mine a large amount of crop, soil, and geospatial data. The result obtained from this work is a compressed dataset which is an exact representation of two lakhs of individual farmer's soil test reports. The compressed data set is used to create a rank generator to specify the most suitable five crops for a particular location and also provide the ranking for important crops in that region. Also, validated this recommended suitable crops with currently used crops and observed that 77.18% of farmers are using the crops that are suitable for their soil but the rest of them are not focused on their soil nutrient status and suitable crops.