Abstract:
Understanding the dynamic shifts in rainfall and temperature holds immense significance
for countries like Bangladesh grappling with critical climate conditions. Natural disasters
such as floods, droughts, and cyclones are intricately tied to variations in these climatic
factors. Therefore, the ability to predict future patterns of rainfall and temperature is crucial
for effective disaster management. While numerous methods exist for forecasting these
conditions, the exploration of predicting future changes in seasonal rainfall and
temperature using diverse machine learning models and assessing their accuracy remains
unexplored. This thesis aims to fill this gap by systematically comparing six machine
learning models such as Linear Regression, Polynomial Regression, Decision Tree,
Random Forest, Adaboost and Gradient Boosting Machines to forecast the seasonal
variations in rainfall and temperature for Bangladesh. The emphasis is on identifying the
most accurate model through rigorous comparisons. The selected optimal machine learning
model is subsequently employed to project seasonal rainfall and temperature conditions for
the year 2025 to 2074. The analysis extends further to examine the percentage change in
seasonal rainfall and change in average temperature between the years 2024 and 2050. By
adopting this approach, the research contributes to advancing the understanding of climate
prediction methodologies, particularly in the context of a country facing acute climate
challenges like Bangladesh.
