Weather Predction Due to its numerous applications in industries including agriculture, utilities, and daily life, weather forecasting has been a significant factor. In the past ten years, the world has faced real-time difficulties with weather forecasting. Because of the constantly shifting weather, the prediction is getting more difficult. The goal of weather forecasting is to foresee future changes in the atmosphere. Understanding the numerous contributing elements that lead to weather changes is essential for effective weather analysis. The process of recording meteorological variables, such as wind direction, wind speed, humidity, rainfall, temperature, etc., is known as weather forecasting. Since machine learning techniques are more robust to perturbations, in this project we applied Neural Network with DNN regressor models and LSTM to predict the weather such as temperature, humidity etc. and compare both approaches and analyzed it. We used two different datasets for the same. Coming to result that we got from each approach was quite amazing. In the Neural Network with DNN regressor approach, we got mean absolute error about mean absolute 1.49 mm and median absolute error 0.94 Celsius and explained variant 0.90 when performing rainfall and temperature prediction respectively whereas in the deep learning approach, the mean absolute error was 0.002268 degree Celsius, when performing temperature, wind speed and pressure prediction respectively. We could clearly see the difference between the outcomes.
On the integration of meteorological and physical principles, weather forecasting is built. Weather forecasting allows for the prediction of surface changes caused by atmospheric conditions (snow and ice cover, storm tides, floods, etc.). The basis for scientific weather forecasting is a combination of computer-controlled mathematical models and empirical and statistical techniques, such as measurements of temperature, humidity, air pressure, wind speed and direction, and precipitation.
For future improvements, following step we thought to took-
Replacing model with a latest/different model
Using other robust datasets
Predicting result on more attributes
Training model on higher-end GPU
Weather application
Also, while performing weather forecasting, there was a lot of complexities involved. There are a lot of variables/attributes to consider for forecasting weather and if all or most of them are used, then we need a lot of computation power to get weather information. And, Real time weather forecasting is very difficult to forecast correctly.
In this project, machine learning and deep learning are used to predict the weather forecasting considered Date, Minimum Temperature, Humidity, and Wind Direction as predictors for rainfall, and they have adopted Supplied test set as a test option. The Correlation coefficient of all base classifiers is greater than 0.8. In this study we considered only seven predictors for rainfall prediction, if we use some more climate factors such as atmospheric pressure, sea surface temperature, so we may obtain more accurate predictions. Also, if Ensemble methods have been applied the results may be improved. Weather forecasts are increasingly accurate and useful, and their benefits extend widely across the economy. While much has been accomplished in improving weather forecasts, there remains much room for improvement.