Soil Fertility Detection and Crop Prediction using IoT and Machine Learning

Abstract

India has huge agriculture heritage. This is the major source of livelihood for most Indian families. Farmers are seen to use fertilizers in inappropriate proportion to enhance crop yield which results in infertile land. To overcome this issue, to check the fertility level of the soil, environment conditions and predicting suitable crop and fertilizers required is need of hour. Soil fertility depends on nutrients like Nitrogen (N), Phosphorous (P), and Potassium (K). It is also affected by environmental factors such as temperature, moisture, humidity, etc. The proposed system provides a cost-effective solution using IoT and Machine Learning based approach to check the NPK concentration present in the soil. Based on which, user can predict the soil suitable crop. The technique used comprises an integrated light transmission and detection system which consists of three LEDs with different wavelengths. Photodiode (LDR sensor module) is used for light detection purposes. The output obtained from the photodiode is handled using a Arduino UNO microcontroller. Based on the inputs received from LDR module, NPK concentration can be evaluated. The model is trained with the Crop Prediction dataset to predict the crop using LightGBM algorithm. The proportion of NPK nutrients and the predicted crop is sent to the user as a text message through the GSM module and ThingSpeak cloud platform.