In India, we see the same crop being grown year after year on the same soil with no mixing or variations, for example rice in northern regions of Haryana and Punjab. This is known as monocropping, and it is shown to be leading to many soil problems like nutrient depletion in the soil as the soil does not get enough time to recover and it becomes very deficient in key nutrients due to the long-term stress of growing the same crops. This study addresses this problem with a machine-learning based recommendation system which can suggest key crops to grow in rotation with existing crops. We use an extensive dataset which has many parameters related to soil content and the weather to train models with techniques like KNN and Random Forest which result in a very high accuracy. We have used official government dataset of soil nutrients for various states in India and tested them in our predictive models. The resultant recommended crop is compared with the predominant crop being grown in that particular state and if this crop is different from the existing one, this is proposed as an alternative crop that can be used in crop rotation to help the soil recover and increase the quality and yield of the soil, leading to ecological balance. These findings can be used as the base of future integrations with IoT enabled smart devices which can help farmers analyse their soil and crop conditions in real time.