Rainfall Prediction Portal
An entire machine learning based web portal to help common people and researchers test and predict and mail rainfall prediction in a city through atmospheric data.
Created on 30th January 2023
•
Rainfall Prediction Portal
An entire machine learning based web portal to help common people and researchers test and predict and mail rainfall prediction in a city through atmospheric data.
The problem Rainfall Prediction Portal solves
The project helps in predicting the amount of rainfall in Mumbai on a given day of a month and atmospheric conditions using a machine learning model trained on a dataset quite accurately using XGBoost machine learning algorithm which considers the sine wave nature of weather. The project also helps visualize the amount of rainfall against various factors such as humidity and temperature.
The interface and project are backed by a Flask server built entirely by us and when deployed on cloud, it can serve people as well as researchers through a beautifully built web portal which relies on HTML and CSS.
It also stores the result of input and predicted output at an instance and can send the results to the mail inbox of the person utilising the web service.
Ultimately, it provides a complex ML based rain prediction model that can be easily customised through few tweaks to code for researchers as well as common people to use on an elegantly made dynamic web portal.
Moreover the whole training data extracted from real world weather observations and patterns is automatically plotted for research purposes through the web portal itself.
Challenges we ran into
Finding an appropriate dataset, plotting graphs, taking into account non-linear nature of weather patterns.
Making Flask server work with the machine learning model and training dataset was also a problem.
Sending dynamically customised webpage emails through the flask server to the user's email was also a tough task.
Putting together all front-end and back-end elements together for a research as well as common usage based web portal was something to tackle and it took a long time.
Had to migrate the whole prediction function from Random Forest Regressor of Scikit-learn to XGBoost Regressor for better accuracy and results.
Decentralizing the functionality into tailored modules of HTML and Python was one of the biggest challenges we faced as the project has to cater to the needs of researchers who would want to tweak it to their needs too.
Finally, displaying the server generated data plots inside webpages was the last piece of puzzle which also took a lot of research to accomplish.
Implementing blockchain was also really difficult as a new domain to explore but we tried our best.
Tracks Applied (3)
Replit
Replit
Solana
Solana