WEATHER PREDICTION USING HIDDEN MARKOV MODEL

Abstract

Weather forecasting is avital role in meteorology and has been one of the most particularly problems all over the world. Data mining techniques are useful for objective recognition, forecasting, and various kinds of pattern extraction. Forecasting is important for many people. This system forecasts weather. Therefore, patterns on updating weather conditions are needed to observe. The aim of the proposed system is to cluster weather data and predict the forthcoming weather conditions based on daily and hourly. The proposed system uses K-means clustering algorithm and Hidden Markov Model (HMM) to predict the weather condition. K-means clustering is used for clustering weather data and generating the observations from the input datasets. The result clustering groups of K-means become the observation matrix of Hidden Markov Model. HMM is used for prediction of upcoming weather conditions. The implementation of the proposed system is developed using C# programming language. Three different datasets, weatherHistory, New_York_Hourly_climate, and Bostonweather_clean, are used as case study. The accuracy of prediction result is used as the performance measure. The accuracy of three data sets, WeatherHistory, Bostonweather_clean and New_York_Hourly_climate, are 85%, 78% and 75% respectively.