I found this competition a couple days ago and have been learning from other participant's notebooks and discussion. Before diving into the competition and pouring in hours of work, I want to make sure that web-scrapping is not playing a role at this competition.
\nHere is the question:
\nWill the use of information from H&M's website give significant advantage?
I took a sample of articles.csv and found that most of the articles still have a product page on H&M's website. The product page contains two sections that can be informative: ![\"product_webpage\"](\"https://i.imgur.com/d7dE2fC.jpg\")
\n(1) Others also bought: it shows 12 items, just like the competition's prediction.
\n(2) Style with: also recommends 12 items.
@cdeotte's notebook shows that recommending items frequent purchased together is effective, where the items (that purchased together) are computed from the competition dataset. What if we use the items from the others also bought session from the product webpage instead? Those could contain information about what users bought AFTER the training period.
Please share your thought. Thanks!
", "messages": [ { "id": "1763677", "postDate": "04/21/2022 17:46:02", "content": "I found this competition a couple days ago and have been learning from other participant's notebooks and discussion. Before diving into the competition and pouring in hours of work, I want to make sure that web-scrapping is not playing a role at this competition.
\nHere is the question:
\nWill the use of information from H&M's website give significant advantage?
I took a sample of articles.csv and found that most of the articles still have a product page on H&M's website. The product page contains two sections that can be informative:
\n(1) Others also bought: it shows 12 items, just like the competition's prediction.
\n(2) Style with: also recommends 12 items.
@cdeotte's notebook shows that recommending items frequent purchased together is effective, where the items (that purchased together) are computed from the competition dataset. What if we use the items from the others also bought session from the product webpage instead? Those could contain information about what users bought AFTER the training period.
Please share your thought. Thanks!
", "rawMarkdown": "I found this competition a couple days ago and have been learning from other participant's notebooks and discussion. Before diving into the competition and pouring in hours of work, I want to make sure that web-scrapping is not playing a role at this competition.\n\nHere is the question: \nWill the use of information from H&M's website give significant advantage?\n\nI took a sample of `articles.csv` and found that most of the articles still have a product page on H&M's website. The [product page](https://www2.hm.com/en_us/productpage.0762558201.html) contains two sections that can be informative: \n(1) `Others also bought`: it shows 12 items, just like the competition's prediction. \n(2) `Style with`: also recommends 12 items.\n\n\n\n@cdeotte's [notebook] (https://www.kaggle.com/code/cdeotte/recommend-items-purchased-together-0-021) shows that recommending items frequent purchased together is effective, where the items (that purchased together) are computed from the competition dataset. What if we use the items from the `others also bought` session from the product webpage instead? Those could contain information about what users bought **AFTER** the training period.\n\nPlease share your thought. Thanks!", "votes": null }, { "id": "1763935", "postDate": "04/22/2022 01:40:20", "content": "Thank you for pointing it out.
\nI agree that web-scraping should not play any role at this competition.
\nI hope the host clearly prohibit this approach.
Thanks for sharing.
\nI think scrapping from the website is an interesting approach in competition, just like the easter eggs in the game.
Personally I don't want to participate in a competition where web-scrapping gives advantages, but it's interesting to think about how one can utilize such information.
\nAnother potential usage of H&M's websites is to identify the sell country of a product (and hence the customer who purchase the product): One can use article_id to find its product webpage. If the webpage is only available at hm.com.cn and not at other countries' websites, we know the product is only sold in China.
", "rawMarkdown": "Personally I don't want to participate in a competition where web-scrapping gives advantages, but it's interesting to think about how one can utilize such information.\n\nAnother potential usage of H&M's websites is to identify the sell country of a product (and hence the customer who purchase the product): One can use article_id to find its product webpage. If the webpage is only available at hm.com.cn and not at other countries' websites, we know the product is only sold in China.", "votes": null }, { "id": "1770738", "postDate": "04/28/2022 14:41:32", "content": "I want to know that the data if someone scrapped from H&M website should make it public according the competition rule?
", "rawMarkdown": "I want to know that the data if someone scrapped from H&M website should make it public according the competition rule?", "votes": null } ], "comments": [ { "id": 1763935, "author_name": "tomooinubushi", "author_url": "", "post_date": "04/22/2022 01:40:20", "content": "Thank you for pointing it out.
\nI agree that web-scraping should not play any role at this competition.
\nI hope the host clearly prohibit this approach.
Thanks for sharing.
\nI think scrapping from the website is an interesting approach in competition, just like the easter eggs in the game.
Personally I don't want to participate in a competition where web-scrapping gives advantages, but it's interesting to think about how one can utilize such information.
\nAnother potential usage of H&M's websites is to identify the sell country of a product (and hence the customer who purchase the product): One can use article_id to find its product webpage. If the webpage is only available at hm.com.cn and not at other countries' websites, we know the product is only sold in China.
", "votes": null, "replies": [] } ] }, { "id": 1770738, "author_name": "juzqyxs", "author_url": "", "post_date": "04/28/2022 14:41:32", "content": "I want to know that the data if someone scrapped from H&M website should make it public according the competition rule?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1763677": "I found this competition a couple days ago and have been learning from other participant's notebooks and discussion. Before diving into the competition and pouring in hours of work, I want to make sure that web-scrapping is not playing a role at this competition.\n\nHere is the question: \nWill the use of information from H&M's website give significant advantage?\n\nI took a sample of `articles.csv` and found that most of the articles still have a product page on H&M's website. The [product page](https://www2.hm.com/en_us/productpage.0762558201.html) contains two sections that can be informative: \n(1) `Others also bought`: it shows 12 items, just like the competition's prediction. \n(2) `Style with`: also recommends 12 items.\n\n\n\n@cdeotte's [notebook] (https://www.kaggle.com/code/cdeotte/recommend-items-purchased-together-0-021) shows that recommending items frequent purchased together is effective, where the items (that purchased together) are computed from the competition dataset. What if we use the items from the `others also bought` session from the product webpage instead? Those could contain information about what users bought **AFTER** the training period.\n\nPlease share your thought. Thanks!", "1763935": "Thank you for pointing it out.\nI agree that web-scraping should not play any role at this competition. \nI hope the host clearly prohibit this approach.", "1763943": "Thanks for sharing.\nI think scrapping from the website is an interesting approach in competition, just like the easter eggs in the game.", "1764056": "Personally I don't want to participate in a competition where web-scrapping gives advantages, but it's interesting to think about how one can utilize such information.\n\nAnother potential usage of H&M's websites is to identify the sell country of a product (and hence the customer who purchase the product): One can use article_id to find its product webpage. If the webpage is only available at hm.com.cn and not at other countries' websites, we know the product is only sold in China.", "1770738": "I want to know that the data if someone scrapped from H&M website should make it public according the competition rule?" }, "source": "meta" }