Challenges

1. ifo Business Survey prediction (ifo Institute)
2. Natural Language Processing (KfW)
3. Spatial data and prediction (DLR)
4. ifoGPT (ifo Institute)

Get started

1. In the ifo Business Survey every month approximately 9,000 German firms are surveyed on their current situation (status quo), recent developments (ex-post), and their plans and expectations for the near future (ex-ante). Build a model which predicts the categorical answers of firms on various questions in the next 12 months. This model can be based on past answers and/or firm-specific information in the micro data of the survey (e.g., firm-size, region, sector).
2. You are provided with meta information on a set of entity articles published after 01.03.2022. Create visualizations and/or tables which highlight news events containing coverage peaks for certain companies/special topics or other anomalies/patterns/cluster in the news article data. Build a model in order to predict the likelihood of an article being read.
3. This spatial ifoHack challenge aims to investigate the relationship between land prices and spatial factors that characterize living environments, in large and diverse cities in Germany, like Berlin, Bremen, Dresden, Frankfurt am Main and Köln. By exploring information extracted from satellite images and other relevant spatial datasets, we aim to guide participants into exploring non-conventional data sources, to answer relevant economic questions related to housing markets and spatial inequalities. Are land prices positively associated with a greater availability of spatial determinants of well-being, such as green and open spaces, or walkable and bicycle paths? Is proximity to amenities, educational, leisure or work opportunities a factor that drives higher land prices? Is there a spatial segregation effect in terms of land prices and specific population subgroups?
4. Use a text corpus of past ifo press releases (use the one provided by us or collect your own) to develop a large language (e.g. transformer) model such as GPT or BERT and provide an intuitive ML-frontend to automatically generate a press release based on keywords such as "positive", "manufacturing sector" and "macroeconomy".

• Idea and use case (20%)
  Problem identification, solution approach, prediction accuracy
• Implementation (50%)
  Tech stack (languages and frameworks) (20%) Clean Code / Reproducibility (10%) Progress towards production (prototype) (20%)
• User experience (20%)
  Web/mobile app, user interaction, deployment
• Live Demo Pitch (10%)
  Prototype >> fancy slides, team engagement, entertainment