Lucrative Startups Screening for Seed Accelerators: A Data-Driven Selection Criteria Pipeline

As startup accelerators become more prevalent, a greater number of startups seek such programs to accelerate their growth. Thus, the staggering number of applicants in each round of intake has made the process of selecting the most lucrative startups costly and overwhelming. This thesis proposes a framework that infers features with the highest prediction power in lucrative startup selection. First, the study extracted 35 criteria with respect to early-stage start-ups. Then, leveraging graph analytics techniques, it extracted another 12 features with respect to the interaction of startups as members of a cohort. I extensively studied the effectiveness of the proposed pipeline and criteria on 35,647 companies founded between 20122015 as well as 763 startups admitted to accelerators in the same period. Results show that the proposed pipeline can predict the success of the final admitted startups with a high performance in terms of AUC (88%) and F1-score (79%).