SDS 437: Data Science at a World-Leading Hedge Fund — with Claudia Perlich

Winning Strategies

Claudia Perlich shares her insights on winning data science competitions, emphasizing the importance of understanding data quality and preprocessing. She recounts her experience with early Kaggle competitions, where identifying data leakage was crucial for success. Claudia notes that winning requires significant time investment, and she has since transitioned to organizing competitions herself 1.

When you win this thing three times in a row, there's really nowhere to go. So I retired and had instead then started to run data competitions myself.

--- Claudia Perlich

Her journey began with the KDD conference, where she first encountered data science competitions, highlighting the evolution of these events over time 2.

   

Data Quality

Data quality issues are a common challenge in data science competitions, and Claudia Perlich emphasizes the importance of skepticism in data analysis. She shares experiences where exploiting data flaws led to competition wins, but stresses that such practices are not viable in real-world applications 3.

The performance is too good in some ways, or there are certain ways that algorithms behave, like the relationship between the performance of the logistic regression and the tree just doesn't make sense.

--- Claudia Perlich

Claudia advocates for setting expectations before analysis to identify discrepancies and learn from them, which can lead to significant insights or reveal flaws in the data or analysis 4.