Validation is really important for problems like this. I baffled by lots of people validating on last month without any gap. Replicating test set split was essential for the validation scheme so I used 8 folds of 2 months gap and 3 months validation with expanding window over the entire dataset. I also added 3 additional folds with their training period ending on December for simulating new year transitions. I had 11 folds but I didn't necessarily use all of them at the same time.
\nI tried some models but none of them improved my validation score so I had to use last value baseline. These are the models and techniques I tried:
\nAs everyone knows, this was a quite strong baseline. I wasn't able to beat this with other models so I decided to improve it by making data-driven decisions.
\nFirst thing I did was adjusting last values to new year which is mentioned here by @cdeotte. I got similar validation score boost when I tested this adjustment on my last three folds that were simulating new year transitions.
\nThe other thing I did was checking last 5 target values of each county:
\nLots of time passed since submissions are closed so I don't remember my exact validation scores but I remember that I got a significant boost with this approach. I didn't use Kaggle notebooks on this competition either. I was generating submissions on local and uploading them. This repository is the only thing I can share…
\nhttps://github.com/gunesevitan/godaddy-microbusiness-density-forecasting/
and this is the script that generates 67th place predictions.
\nhttps://github.com/gunesevitan/godaddy-microbusiness-density-forecasting/blob/main/src/county_models/last_value.py
Last value was really strong! I used constants for March, April, May for all counties but using a rule-based approach would have been better. I was afraid of overfitting, but looks like it worked out of sample!
", "votes": 1, "replies": [ { "id": 2307396, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2023-06-18T06:42:42.933000", "content": "I selected constants that were minimizing my validation score so I didn't think I was overfitting at any time.
", "votes": 1, "replies": [] } ] }, { "id": 2306376, "author_name": "Machine Lying", "author_url": "", "post_date": "2023-06-17T08:40:01.813000", "content": "Congratulations, thank you for your proposal, this competition really showed me that sometimes simple models can bring good results, the main thing is the insight of the data😃
", "votes": 1, "replies": [] }, { "id": 2318453, "author_name": "RogerOcean", "author_url": "", "post_date": "2023-06-26T11:25:53.410000", "content": "thanks for sharing, I forgot to use rule-based model😂 even if I wrote a notebook about it Godaddy: Select cfips with strong linear trend. But in the final analysis, I worried about this method may cause overfitting😭
", "votes": 0, "replies": [] } ], "comment_count": 4, "messages": [ { "id": 2306265, "postDate": "2023-06-17T07:34:58.033Z", "content": "Validation is really important for problems like this. I baffled by lots of people validating on last month without any gap. Replicating test set split was essential for the validation scheme so I used 8 folds of 2 months gap and 3 months validation with expanding window over the entire dataset. I also added 3 additional folds with their training period ending on December for simulating new year transitions. I had 11 folds but I didn't necessarily use all of them at the same time.
\nI tried some models but none of them improved my validation score so I had to use last value baseline. These are the models and techniques I tried:
\nAs everyone knows, this was a quite strong baseline. I wasn't able to beat this with other models so I decided to improve it by making data-driven decisions.
\nFirst thing I did was adjusting last values to new year which is mentioned here by @cdeotte. I got similar validation score boost when I tested this adjustment on my last three folds that were simulating new year transitions.
\nThe other thing I did was checking last 5 target values of each county:
\nLots of time passed since submissions are closed so I don't remember my exact validation scores but I remember that I got a significant boost with this approach. I didn't use Kaggle notebooks on this competition either. I was generating submissions on local and uploading them. This repository is the only thing I can share…
\nhttps://github.com/gunesevitan/godaddy-microbusiness-density-forecasting/
and this is the script that generates 67th place predictions.
\nhttps://github.com/gunesevitan/godaddy-microbusiness-density-forecasting/blob/main/src/county_models/last_value.py
Last value was really strong! I used constants for March, April, May for all counties but using a rule-based approach would have been better. I was afraid of overfitting, but looks like it worked out of sample!
", "rawMarkdown": "Last value was really strong! I used constants for March, April, May for all counties but using a rule-based approach would have been better. I was afraid of overfitting, but looks like it worked out of sample!", "votes": 1, "replies": [ { "id": 2307396, "postDate": "2023-06-18T06:42:42.933Z", "content": "I selected constants that were minimizing my validation score so I didn't think I was overfitting at any time.
", "rawMarkdown": "I selected constants that were minimizing my validation score so I didn't think I was overfitting at any time. ", "votes": 1 } ] }, { "id": 2306376, "postDate": "2023-06-17T08:40:01.813Z", "content": "Congratulations, thank you for your proposal, this competition really showed me that sometimes simple models can bring good results, the main thing is the insight of the data😃
", "rawMarkdown": "Congratulations, thank you for your proposal, this competition really showed me that sometimes simple models can bring good results, the main thing is the insight of the data😃", "votes": 1 }, { "id": 2318453, "postDate": "2023-06-26T11:25:53.410Z", "content": "thanks for sharing, I forgot to use rule-based model😂 even if I wrote a notebook about it Godaddy: Select cfips with strong linear trend. But in the final analysis, I worried about this method may cause overfitting😭
", "rawMarkdown": "thanks for sharing, I forgot to use rule-based model😂 even if I wrote a notebook about it [Godaddy: Select cfips with strong linear trend](https://www.kaggle.com/code/roger92/godaddy-select-cfips-with-strong-linear-trend). But in the final analysis, I worried about this method may cause overfitting😭" } ], "complete": true, "source": "live", "fetched_at": "2026-10-19T19:28:19Z", "competition_id": 41881, "competition_slug": "godaddy-microbusiness-density-forecasting", "team_id": 9628036, "solution_url": "/competitions/godaddy-microbusiness-density-forecasting/writeups/daddy-62th-place-solution-last-value-and-rule-base" }