Smart Agricultural Practices using Machine Learning techniques For Rainfall Prediction: A case Study of Valkenburg station, Netherlands.

Abstract

Forecasting weather in general and rainwater or downpour in particular is significant for cropping-plan decision-making, water resource management which helps to determine impending irrigation potentials and other important aspects that help the farmer to have better and quality yield. Meteorological conditions and the precise prediction of weather patterns are important for agriculture and allied sectors. Predicting weather conditions like rainfall which is a type of weather pattern that is influenced by various weather parameters like wind direction, wind speed, humidity, geographical location, temperature etc. and have inherent connection with the agricultural activities.  Adoption of  disruptive technologies like artificial intelligence & machine learning for rainfall predictive analytics have achieved better results with appreciable performance and accuracy in predicting rainfall as compared to the  traditional statistical methods. In this work, we have implemented Naive Bayes technique for the rainfall prediction. The historical weather data is collected from Royal Netherlands Meteorological Institute (KNMI) which is available on http://www.sciamachy-validation.org/climatology/daily_data/selection.cgi.Out of 39 available weather attributes, 5 most relevant attributes are selected for rainfall prediction using genetic algorithm and are more relevant for better rainfall prediction. The weather data set of 7670 daily weather instances of Valkenburg station from 1990 to 2010 is used to build the model using Naïve Bayes approach and its accuracy is tested on a test data set having 1826 daily weather samples from 2011 to 2105. The experimental results using Naïve Bayes algorithm show 71.2% accuracy rate for rainfall prediction which is appreciable and significant for predictive analytics.