Rain Prediction In Australia With Active Learning Algorithm
Zhixuan He
Abstract
In the past, weather forecasting usually uses meteorological satellites for forecasting. Nowadays, with the development of computer technology, machine learning is increasingly used for weather forecasting. This research aims at using an active learning algorithm to make the prediction on the rain in Australia. The data set which includes weather data is from Kaggle. The active learning contains a pool-based sampling method and the query strategy is entropy sampling. The classification model is logistics regression model. The research also provides a comparison of the prediction accuracy between active learning and random sampling. In the experiment, The prediction accuracy of active learning with pool-based sampling algorithm is 82.02%, the prediction accuracy of the random sampling is 82.0%. The experiment presents that these two algorithms have almost the same prediction accuracy.