Crop production is a very complicated attribute that depends on a number of variables, including genotype, environment, and their interactions. The functional relationship between yield and these interaction elements must be fundamentally understood for accurate yield prediction to be possible, and both large datasets and potent algorithms are needed to show this relationship. In order to accurately estimate crops, we applied cutting-edge machine learning techniques. Our ability to gather data for huge agricultural areas spread throughout far-flung regions of the globe is made possible by the Internet of Things’ architectural principles. So that we may make forecasts about crops, our machine learning system can use this data. Nitrogen, phosphorus, potassium, temperature, humidity, and rainfall all have a role in determining the best crop to grow. Recommendations are made in light of these factors. There are 2200 total occurrences in the data collection, and there are 8 associated attributes for each one. For each of the 8 possible characteristic combinations, around twenty-two distinct plant species are available as potential options. The most effective model can be produced by utilizing the supervised learning approach and few of the machine learning methods available in WEKA. As prospective possibilities for the methods of machine learning that would be used in the classification process, the decision table classifier and the multilayer perceptron rules-based classifier JRip were selected. The system’s design took into account both the growing Internet of Things and the essential measurements required for farming. The average weighted value of the Receiving Operator Attributes has been found to be 1, the performance measured by the classifiers that were chosen has a value of 98.2272%, and the maximum time needed to construct the model is 8.03 s. Employing machine learning in agriculture aims to increase the production and nutritional value of the plants generated by this sector.