Weather Forecasting Using Machine Learning Approach

Abstract

Weather forecasting has been playing an important factor due to it’s applications in the various sectors such as in agriculture, utility company and in day to day life, thus it has been evolved as an important field of research for the few years. Since developing a linear interrelation between weather data and target output is a tedious job, nonlinear prediction of weather data has gained focus. The per minute weather data collected from NIT Rourkela was pre processed, missing values were cleaned, other parameters were normalized or scaled between 0 to 1 using min-max normalization. The per minute rainfall is combined and the per hour rainfall is calculated. The rainfall is then labeled into five classes depending on the quantity of rainfall in an hour. Various classifiers were employed and accuracy was calculated for rainfall. The classification was made by taking the parameters such as timestamps, ambient temperature, wind speed, wind direction, solar radiation, relative humidity, atmospheric pressure, net solar radiation as input parameters and based on these the target output rainfall is predicted.