Deep neural net based approach for air pressure prediction

Accurate and real-time air pressure prediction is an important factor in the weather forecast, especially in forecasting various natural disasters like-heavy rainfall, cyclone or storm and depression etc. It requires lots of attention to analyzing the given conditions, low pressure causes heavy rainfall and thus accurate pressure prediction helps to protect life and property and thus it also helps the government to take the necessary steps required. Artificial intelligence approaches like different Machine Learning algorithms have helped in the prediction of pressure. In this paper, a Bidirectional Long Short-Term Memory (BiLSTM) model has been proposed and the result obtained is compared with aLong Short-Term Memory (LSTM) model. Several parameters like-Pressure (millibars), Temperature (Celsius), Temperature (Kelvin), Relative humidity, Saturation vapor pressure, Specific humidity, Water vapor concentration, Airtight, Wind speed, wind direction, etc. are used. The RMSE value is 0.8364 for the LSTM model and 0.41 for the BiLSTM model and both the results are compared. Here in this paper, a more accurate result or prediction using the higher efficient model is calculated.