Timeframe and Context

The project was carried out from October 2022 to March 2023 as part of my bachelor’s thesis in cooperation with OFYZ GmbH / aumentoo GmbH.

The goal was to create a predictive model to forecast the success of startups based on historical company data.

As the sole developer, I bore full responsibility for the project and was responsible for the development, training, and evaluation of the predictive mode

Implementation and Tech Stack

The implementation began with the definition of the target variables and the creation of a project plan. The company data was then collected, understood, and preprocessed before a prediction model was developed and its performance evaluated.

Python was used for modeling and evaluation, and MySQL was used for data management.

Challenges and Results

The data was highly imbalanced (<5% successful startups), which necessitated the use of oversampling. In addition, the subject matter was complex, with many influencing factors, and the available time frame was limited. Since I had little experience in AI and Data Science at the beginning, extensive independent training was necessary.

The model was able to correctly predict the failure of startups with 95% accuracy, while the probability of success was predicted with 35% accuracy.