With this project, we are replying to the DLR challenge to deliver a tool for users to explore estimated property values drawing from different sources, including geospatial data.
In the interactive Jupyter Notebook, the user can check the values for neighbourhoods delivered by a machine learning algorithm for major German cities. This feature makes use of different large data sets on population, public infrastructure, building substance, and transport.
First, we had to ensure that data sets from different sources were interoperable by transforming and matching location to coded information in the datasets. Second, we analysed public infrastructure from maps (e.g. kindergartens, industrial neighbourhoods) and fed this into the data frame. Third, a machine learning algorithm was trained to infer the effect of various features of neighbourhoods on land and property prices.
Handling geospatial data as non-experts is difficult!
We are proud that we started from scratch and no prior knowledge of the data types and environments and, after three days of learning, can present the results of our hard work.
Coming from different academic backgrounds that do not regularly interact with geospatial data, we learned a lot about the handling, application, and analysis of these data. Despite the technical hurdles, we also practised development workflows and coding in teams.
Hopefully win the ifohack 2023 :) We have different ideas for improving the user experience, including a tie in with device GPS data and map exploration. If combined with augmented reality tools, the user can explore the data based on their live location (e.g. pointing their phone at a house or street and receiving information from the database). Further fine-tuning of the data sets and algorithms can help to improve the accuracy of the computed values.
Look at our userstories.txt file. There are awesome ideas who could use our application for which purpose.
Ludwig Meyer, Thomas Meyer, Tobias Weber, Florian Korn