Use of Machine Learning to Predict and Analyze the Climatic Trends: A Data-Driven Approach

Abstract

Climate change is an alarming problem in the world, in recent years the extreme heat wave has been quite noticeable. For the last couple of years a temperate country like Bangladesh, is suffering from extreme heat waves. The temperature in many areas of Bangladesh rises to 40 degrees Celsius in the summer season. By investigating the data that have been collected for this study, tried to understand the weather trend and forest area trend of Bangladesh. The data was collected from various sources. NASA's POWER Project provides the temperature data, while the Humanitarian Data Exchange (HDX) website provides the rainfall data, with particular attention to Bangladeshi rainfall data. The forest area data was collected from World Bank Data on Forest Area and FAO Bangladesh Country Paper. 48 years of historical data have been used there. There are four types of data present in this dataset. These are Temperature(C), Date, Rain (millimeters), and Forest Area(percentage). 586 sets of data are available in this dataset. By investigating the dataset by exploratory data analysis it was found that there are positive high correlation between rainfall and temperature and a negative moderate correlation between forest area and year. Four machine-learning models were used there to do a comparative analysis between models. The models are gradient boosting, random forest, ridge regression, and lasso regression with an accuracy of 96.01 95.57, 94.07, and 93.89 respectively. The best-performing model was then integrated into a user-friendly web-based interface, allowing users to input specific years and months to receive temperature predictions. The interface also utilizes historical trends in rainfall and forest area to enhance prediction accuracy. This study fills in the current gaps in predictive modeling and provides everyone with a key resource for enhancing climate resilience strategies and making well-informed decisions.