EcoMind AI

Inspiration

We wanted to see the factors that influenced global warming and rising temperatures. Thus, we made the EcoMind AI website so that users can easily model global warming in their own climates.

What it does

The user inputs values for 32 different weather data factors. Some extremely important factors are outlined in red, while more optional factors that have less impact on the model are outlined in gray. Users then hit a "Submit" button, and an explainable Machine Learning model will predict temperature. We also save user data, so that users can send it back to us to further train our model.

How we built it

For the website, we used a combination of Flask, Bootstrap, HTML, CSS, and Python to build our interface and backend. To develop the model, we used a Random Forest model, which is extremely explainable - it's similar to a Decision Tree, where you are given an input, and based on a threshold, you either go left or right. The Random Forest model uses several Decision Trees to come up with its prediction. We used Pandas to format our data, and LIME and SHAP to create graphics to further to explain the impact of our data columns on our model.

We also Figma to design the UI, and Trello to organize and assign tasks.

Challenges we ran into

There was a ton of missing data! Some row values were missing, and some lines of data had many NaN or empty values. Processing the data to get a proper dataset proved to be very difficult. After this section, we didn't run into too many other issues.

Accomplishments that we're proud of

We're proud of how the look of the website turned out! It was a tiny bit tricky to translate from the Figma diagram to the real website, but once the final frontend was completed, we were happy with our work.

What we learned

We learned how to develop and use explainable models, as well as process data. We also learned how to save and integrate a trained machine learning model into a website.

What's next for EcoMind AI

Improve the website, implement better data processing and treatment methods, and spread awareness of the impact of climate change by reaching out to the public with our product. Furthermore, we can collect data globally, not just from the United States.

Built With

• anaconda
• bootstrap
• explainable-models
• flask
• lime
• machine-learning
• pandas
• python
• scikit-learn
• shap