Skip to main content

Table 4 Fraction of predicted stages for cells that contain previously unseen functions. CORAL accurately categorizes common data analysis functions as frequently belonging to their expected stage

From: CORAL: COde RepresentAtion learning with weakly-supervised transformers for analyzing data analysis

Function

Expectation

IMPORT

WRANGLE

EXPLORE

MODEL

EVALUATE

pandas.DataFrame.dropna

Wrangle

0

0.93

0.07

0

0

pandas.DataFrame.groupby

Wrangle

0

0.52

0.12

0.02

0.34

seaborn.jointplot

Explore

0

0.00

0.98

0.00

0.02

seaborn.countplot

Explore

0

0.01

0.98

0.00

0.01

sklearn.linear_model.SGDClassifier

Model

0

0

0

0.67

0.32

sklearn.linear_model.PassiveAggressiveClassifier

Model

0

0.06

0

0.61

0.39

sklearn.metrics.f1_score

Evaluate

0

0

0.01

0.05

0.94

sklearn.metrics.log_loss

Evaluate

0.02

0.01

0.02

0.26

0.70