Inspiration My colleagues and I were curious about the relationship between socioeconomic data and land prices in different cities in Germany, so we decided to examine it with real data.
What it does We came up with an interactive map, where you can deploy predicted prices for land in 5 German cities, based on population distribution, ethnic background and country of origin. It is a very user-friendly interface and all you have to do is open it on your computer, select your city and click on a neighborhood to see the different prices and segregation levels.
How we built it First, we gathered census data on German population and aggregated it on the neighborhood level for 5 different German cities using R. While using the census data, we also created a segregation feature improve the performance of our model. Then, we tested three different ml algorithms, including linear regression, random forest regression and svr to predict land price, all using Python. The most accurate model was a random forest regression. On parallel, we created a stream-lit app with a very user-friendly interface to deploy the algorithm and visualize the predictive prices to every neighborhood.
Challenges we ran into We ran into a bunch of problems starting with the installation of the conda environment, then bringing all the members of the group to work together on GitHub Repo and joining geospatial data together, learning about ml algorithms, but we managed to solve that. When building the app, among the challenges we faced were displaying open-street-map, but main thing the app is working!
Accomplishments we are proud of We are proud of joining the geospatial dat, building models and creating the stream lit app from scratch.
What we learned We learned a lot about geospatial data, machine learning models and building an app with stream-lit.
Whats next for Brazy We don’t know that yet, it's only the beginning!
Built With
- mlflow
- python
- r
- scikit-learn
- stata
- streamlit
Log in or sign up for Devpost to join the conversation.