Build classifier to detect is there contact data in advertisement or not.
Dataset train.csv
could be downloaded using script ./data/get_train_data.sh
or by clicking on
link
As a model metric averaged ROC-AUC
on each advertisement category is used.
Source code to train model is located in ./lib/train_model.py
.
Baseline model score is 0.91 and must be overcome by a trained model.
For training python==3.8.6 is used.