An improved naive bayes model for air temperature prediction

In this paper, a Temperature Probability Prediction with Multi Factor (TPPMF) based on Naive Bayesian algorithm (NB) is designed for weather data including multiple types and forms. The defect of continuous variable analysis is solved through the variable - probability transformation. A temperature increment algorithm is proposed to improve the accuracy of NB. And a compressed encoding method for feature vector of incompatible incidents achieve the dimensionality reduction of data expression, which reduced the training time for about 20%. The experiment results show that the root-mean-square of the mean-absolute-error minimum reached 1.698°C, in the 24-hour air temperature prediction of 12 typical cities.