machine learning for farming decisions

The robust and efficient Machine learning model for smart farming decisions and allied intelligent agriculture decisions

Shraban Kumar Apat, Jyotirmaya Mishra, K. Srujan Raju, Neelamadhab Padhy


Crop Yield Prediction is essential in today's rapidly changing agricultural market (CYP). Accurate prediction relies on machine learning algorithms and selected features. Any machine learning algorithm's performance might well be enhanced by introducing a diverse set of features into the same training dataset. Crop yield prediction includes parameters such as temperature, humidity, pH, rainfall, as well as crop name in forecasting the yield of the crop based on historical data. It offers us an indication of the best crop to expect in terms of weather conditions in the field. Crop prediction is a difficult task in the agricultural realm. The primary purpose of this research is to offer a novel machine learning approach for a heterogeneous data environment containing IoT-sensed data about the environment, agricultural conditions, plants' features, demands, etc. We have used data set of the following five crops (rice, ragi, gram, potato & onion) collected from Andhra Pradesh, Kaggle repository. In this work, we utilized diverse machine learning as well as deep learning algorithms such as Regression methods, Decision tree, Naive Bayes, SVM, K-Means, Expectation-Maximization (EM), and AI techniques (LSTM, RNN). It seems that among machine learning techniques, the Random Forest algorithm outperforms with 99.27% training accuracy for crop yield prediction. However, among sigmoid, ReLu, and tanh activation, sigmoid achieves 99.71 percent accuracy with four hidden layers for predicting the crop yield prediction.


Agriculture decisions; smart farming decisions; IoT; Machine learning; Deep Learning

Full Text: