Weather is a continuous, data-intensive, multidimensional, dynamic and chaotic process and these properties make weather prediction a big challenge. Building a hybrid approach, that has the capability to deal with linear and nonlinear relationships which is best suitable in most applications, is the objective here.
Weather forecasting has become an important field of research in the last few decades. Establishing a linear relationship between the input weather data and the corresponding target data is always preferred. But with the discovery of nonlinearity in the nature of weather data, the focus has shifted towards the nonlinear prediction of the weather data.
So, it has two parts as mentioned below:
- Prediction of Rain i.e. whether it will rain or not.
- Prediction of different weather factors which will help us to predict rainfall. To achieve this second goal, a Hybrid Time series prediction model is built, which will try to learn the relationship between the weather factor and time.
The data for mean annual rainfall over Bhubaneshwar region of Odisha, India has been used for the study. It consists of the data values of the factors such as date, temperature, dew point, humidity, wind speed, wind gust, direction, viscosity, pressure, precipitation, precipitation rate, total precipitation, condition, fog, rain, snow, hail, thunder and tornado. The data is recorded at a time interval of 30 minutes and consists of one year data, starting from January 2017 to January 2018.
Environment : Python3
- Pandas
- Numpy
- Matplotlib
- Seaborn
- Statsmodels
- Sklearn
- Tensorflow
- Keras