QuadReal

Inspiration

Environmental Problems have increasingly getting people's attention. QuadReal is a project that with more than 1GB data provided and environment-related work, our group is really interested into this.

What it does

We actually built a neural network model and use that to predict future environmental details with column given.

How we built it

actually we tried a lot of network models, by splitting up the 1GB css training file into several small training files by building_id, using those models to fit the data is quick and reliable. After the trails, we found out that the MLPRegressor model fits the best.

Challenges we ran into

We really have problem in computer power. Training AI is a consuming work, we knew we might not doing well in the final correctiveness, but our researching progress makes us pride.

Accomplishments that we're proud of

As mentioned above, we're really proud of doing a project as a team and moving problem together, we came up with many different ideas and having much conflicts during the model selection and finally we found a mathematical way to solve all that confilcts.

What we learned

Thanks to this competition, we learned a lot about the AI and group work, making great power for four people to focus on one thing is an amazing time for us.