🌍 Must-Read: LB score no better than random...

INEGI UN-GGIM Human Settlement Detection Challenge by ITU

Helping Mexico

1 000 CHF

Completed (over 1 year ago)

Skills you will learn

Classification

208 joined

84 active

Info Data Chat Leaderboard

Start

Sep 02, 24

Sep 28, 24

Reveal

Oct 10, 24

nymfree

LB score no better than random guesses ~0.5

Data · 7 Sep 2024, 11:55 · 6

My LB scores seem to be no better than random guesses for NN or GBT models, despite local AUC CV being 0.81 for NN and 0.76 for GBT. Has anyone else experienced this?

Discussion 6 answers

not_today

Facing same problem. Anyone here who can explain this

10 Sep 2024, 19:30

Upvotes 0

nymfree

Figured it out. you have to use the provided id_map file

10 Sep 2024, 19:50

Upvotes 0

not_today

I don't understand what you said. Can you explain

replied to nymfree10 Sep 2024, 22:22

Upvotes 0

not_today

Can you share any sample notebook?

replied to nymfree12 Sep 2024, 10:58

Upvotes 0

nymfree

Basically after making predictions, assuming all your predictions are in a list, do:

df = pd.read_csv('id_map.csv')
my_dict = {}
for i,row in df.iterrows():
    my_dict[row['id']] = row['ID']
    
sub = pd.read_csv('SampleSubmission.csv')
for i in range(len(sub)):
    sub.at[i,'class'] = test_predictions[my_dict[sub.at[i,'id']]]

replied to not_today12 Sep 2024, 11:19

Upvotes 3

not_today

thanks

replied to nymfree12 Sep 2024, 11:29

Upvotes 0

Join the largest network for
data scientists and AI builders

About FAQs

Status