You are browsing as a guest. Sign up (or log in) to start making projects!

Machine learning rain calculator model.

By @zuhair_chattha

  • 3 Devlogs
  • 2 Total hours

Weather Desk is a machine-learning weather prediction project that predicts whether it will rain tomorrow using temperature, humidity, pressure, and wind speed. I started with an open weather dataset from an AI competition and experimented with different classification models, including Logistic Regression, Decision Trees, Random Forest, and Gradient Boosting. After testing different approaches, I selected a stronger model and turned it into an interactive Streamlit web app. Users can enter weather measurements and get a rainfall prediction with its estimated probability. The project is hosted on GitHub and deployed publicly so others can try it.

Source
Ship #1 Pending review

@zuhair_chattha on Machine learning rain calculator model.

I engineered a localized weather computation model inspired by the severe monsoon cycles designed to process atmospheric variables like humidity, wind speed, and barometric pressure. The primary technical hurdle was programmatically handling erratic weather spikes without breaking the calculation engine, which required extensive debugging to ensure extreme data inputs didn’t corrupt the predictive output. I am deeply proud that I built this functional system completely on my own at just 15 years old, proving that youth is no barrier to creating impactful tools for real-world issues. To test the project, reviewers simply need to clone the repository, run the main script, and input custom values for humidity, pressure, and wind speed (or use the provided sample dataset) to see how the algorithm dynamically updates its calculations.

  • 3 devlogs
  • 2h
Try project → See source code →
Open comments for this post

@zuhair_chattha on Machine learning rain calculator model.

1h 9m 28s logged

Weather Desk — Development Log
Stage 2 — Trying Different Models
After getting 64.9% with Logistic Regression, I had my first baseline.
I started trying different models to see if I could improve it. I tested a Decision Tree and Random Forest, and later tried Gradient Boosting as well.
The results were interesting because the models didn’t all perform the same way. Gradient Boosting ended up being one of the strongest versions I had tested so far.
I now have a model that looks promising, but I’m not stopping here. I want to test it properly and see if I can improve the final result.
Still experimenting. 🌦️

0
0
4
Open comments for this post

@zuhair_chattha on Machine learning rain calculator model.

27m 54s logged

Weather Desk — Development LogStage 1 — Getting Started With My First ModelI started this project by trying to understand what the dataset was actually showing instead of jumping straight into a model.The main goal was to predict whether it would rain tomorrow using today’s temperature, humidity, pressure, and wind speed.I left out the id column because I understood that it was just a record number and not an actual weather measurement.Since the answer is either 0 (no rain) or 1 (rain), I learned that this is a binary classification problem.I split the 5,000 training rows into 80% for training and 20% for testing. I did this so I could check how the model performs on data that it didn’t learn from.For my first model, I chose Logistic Regression. I picked it because it was simple, quick to run, and gave me something to use as a baseline before trying more advanced models.The first result was 64.9% accuracy.This wasn’t my final goal. I mainly wanted to have a starting point so I could compare the next models against it and see whether my changes were actually helping.

0
0
3
Open comments for this post

@zuhair_chattha on Machine learning rain calculator model.

48m 56s logged

just started my project Got inspiration through the question (How does really weather forecasting would be working?)And started to build a weather forecasting machine learning model. Started from small. Using some open-source train and test CSV datasets from a competition to train my model.

0
0
1

Delete project?

Are you sure you want to permanently delete this project? This action cannot be undone.

All devlogs, followers, and associated data will be removed.

Followers

Loading…