Modeling for Tabular Competitions
Until 2017, there was no need to distinguish too much between competition types and, since the vast majority of competitions were based on tabular data, you could not even find mention of “tabular competitions” on Kaggle forums. Suddenly, something changed. After a relative shortage of competitions (see https://www.kaggle.com/general/49904), deep learning competitions took the upper hand and tabular competitions became rarer, disappointing many. They became so rare that Kaggle recently had to launch a series of tabular competitions based on synthetic data. What happened?
By 2017-2018, data science had grown to full maturity and many companies had initiated their data journeys. Data science was still a hot topic, but no longer such an uncommon one. Solutions to problems similar to those that had populated Kaggle for years at the time had become standard practice in many companies. Under these circumstances, sponsors were less motivated...
