Weather Forecasting and Analysis Based on Machine Learning

Weather forecasting has become a topic of increasing concern due to unnatural causes such as human activities. Although the weather forecasting methods in the past: traditional synoptic method and numerical prediction method can predict the change of the weather very accurately, but these two methods need to monitor and process a large amount of data, which consumes resources and manpower. When people just want to get the weather of a particular area, it's not worth it. Our study provides a new way to predict weather from small pieces of data. The data includes temperature, rainfall, wind direction, wind speed, humidity, pressure, cloud and sunshine which previous research did not use all of them. Firstly, a small part of data was obtained from reliable databases. Then this paper compared different methods (Drop the incomplete samples, using mean value of the column, K-nearest-neighbors) to select the best method for refining data sets. Then this paper compared different models (Random Forest Regression, Logistic Regression, Support Vector machine) to select the best model suitable for small data for machine learning. Finally, this paper formed tables, bar charts and broken line charts to show the weather prediction and evaluation of each model through data visualization, as well as the prediction of evaporation, rainfall and other future weather data. This study created an interactive web page to showcase the final result and make it easy for people to view at any time.