A Hidden Markov Model-based Approach for Weather Prediction in Bangladesh

In the context of Bangladesh, a country prone to diverse and often unpredictable weather patterns, reliable weather forecasts are critical for making informed decisions and mitigating the impact of extreme weather events. In this research, we present an innovative approach to predict the average temperature and rainfall of Bangladesh using Hidden Markov Models (HMM). Our HMM-based approach leverages the past weather data as the observation in order to predict future weather patterns. For this study, we used the Bangladesh Weather dataset from Kaggle which contains monthly average temperature and rainfall data from 1901 to 2015. Experimental results show that the proposed HMM model was able to successfully capture the trends of the weather pattern with an MAE of 0.74 and 77.01 for the temperature and rainfall prediction respectively.