## DEADLINE IS EXTENDED TO ALLOW LATE SUBMISSIONS FOR YOUR OWN INTEREST ## NO PRIZE SHOULD BE EXPECTED ## The challenge Your challenge is to accurately forecast the demand of 300 retail products across 20 retail stores. Accurately forecasting demand is crucial for retailers because it allows them to make informed decisions regarding inventory management, pricing strategies, and sales projections, all of which can have a significant impact on a retailer's bottom line. In short, demand forecasting is an essential tool for retailers looking to optimize their operations and maximize profits so they can stay competitive in a crowded retail landscape. ## Join our Data Science Hackathon and beat our model with Cyclic Boosting Are you ready to take on this challenge and prove your data science skills? Join our hackathon and use the newly open sourced Cyclic Boosting machine learning library to beat our model with better accuracy. Cyclic Boosting is a cutting-edge machine learning library that provides high-performance algorithms for classification and regression tasks. It is based on a novel cyclic boosting algorithm that achieves state-of-the-art accuracy on many benchmark datasets. To participate in the hackathon, you just need to download the Cyclic Boosting library (instructions can be found on [cyclicboosting.org](https://cyclicboosting.org/)) and train your own model. Use Cyclic Boosting or try any other model to beat our baseline. Everybody, who beats the benchmark model will win a free beverage at our booth in the exhibition hall at the PyCon DE & PyData Berlin conference. Come to the booth and let's discuss your solution. In addition, two winners will be selected: the highest score with the Cyclic Boosting library and with your any other library. Don't miss this opportunity to showcase your data science skills and learn about the latest advancements in machine learning. Join our hackathon and beat our baseline! ### How to participate 1. Register for the hackathon as a team and add your ticket id to the team name if you want to get prizes (Team tab) 2. Check the example notebook with CB on kaggle (Code tab) 3. Train your own models using the provided datasets locally or on Kaggle 4. Submit your results before the end of the conference. **Please provide your ticket ID (Booking reference, can be found in your confirmation email) for the PyCon DE & PyData Berlin 2023 on your submission or just prove it is your submission at the booth ** Also, please add the model you used in the submission description. You can submit either through csv file through "Submit Predictions" button above or like int the template notebook on Kaggle (Code tab) 5. Get the free drink (beer or soft drink) after the first submission at the booth. Also, beat the benchmark model (lower MAE than for the CB_simple_submission 5.5) on public leader-board and get additional prizes at the BY booth if you participate in the conference on-site 6. Private leaderboard results will be revealed on Tuesday after 4pm and two winners will be selected: top score with CB and with any other model from the conference on-site participants Winners will get practical BY thermal bottles to celebrate their achievements P.S. Also, try to beat CB_master_submission with MAE of 4.87 to amaze our Corporate Vice President (Data Science) and creator of Cyclic Boosting Felix Wick ### Resources - [Cyclic Boosting documentation](https://cyclic-boosting.readthedocs.io/en/latest/index.html) - [submitting predictions](https://www.kaggle.com/docs/competitions#making-a-submission) Good luck!