I'm working my way through trying to make sense of this data, it seems like data munging is going to be an even higher % of the work in the competition then usual.
\n\nI put together a viz to help give an intuition on what's going on. It's still a work in progress, but thought I'd share:https://cole-maclean.github.io/bosch-production-line-performance/
\n\nFeatures are colored on defect rate and sized by total parts that filter through that feature.
\n\nSeems like many features don't have an defective parts that run through them. Seems like an easy place to start in reducing the size of the dataset, but I wonder if there are some creative techniques in utilizing features that only have data for one of the 2 classes. Think I'll go do some research on that.
\n\nUpdate: Updated to show all the unique paths parts can take and the count of parts that take that exact path. Numbers indicate the order in which each station is hit. Should help to visualize the different product types flowing through the factory, and the order stations are hit. The resolution of the timestamps are not very good, so even tho the ordering might change, in reality the two stations were likely hit with the exact same timestamp - so as far as we know at the same time.
", "messages": [ { "id": "133723", "postDate": "09/03/2016 14:33:11", "content": "I'm working my way through trying to make sense of this data, it seems like data munging is going to be an even higher % of the work in the competition then usual.
\n\nI put together a viz to help give an intuition on what's going on. It's still a work in progress, but thought I'd share:https://cole-maclean.github.io/bosch-production-line-performance/
\n\nFeatures are colored on defect rate and sized by total parts that filter through that feature.
\n\nSeems like many features don't have an defective parts that run through them. Seems like an easy place to start in reducing the size of the dataset, but I wonder if there are some creative techniques in utilizing features that only have data for one of the 2 classes. Think I'll go do some research on that.
\n\nUpdate: Updated to show all the unique paths parts can take and the count of parts that take that exact path. Numbers indicate the order in which each station is hit. Should help to visualize the different product types flowing through the factory, and the order stations are hit. The resolution of the timestamps are not very good, so even tho the ordering might change, in reality the two stations were likely hit with the exact same timestamp - so as far as we know at the same time.
", "rawMarkdown": "I'm working my way through trying to make sense of this data, it seems like data munging is going to be an even higher % of the work in the competition then usual.\r\n\r\nI put together a viz to help give an intuition on what's going on. It's still a work in progress, but thought I'd share:https://cole-maclean.github.io/bosch-production-line-performance/\r\n\r\nFeatures are colored on defect rate and sized by total parts that filter through that feature.\r\n\r\nSeems like many features don't have an defective parts that run through them. Seems like an easy place to start in reducing the size of the dataset, but I wonder if there are some creative techniques in utilizing features that only have data for one of the 2 classes. Think I'll go do some research on that.\r\n\r\nUpdate: Updated to show all the unique paths parts can take and the count of parts that take that exact path. Numbers indicate the order in which each station is hit. Should help to visualize the different product types flowing through the factory, and the order stations are hit. The resolution of the timestamps are not very good, so even tho the ordering might change, in reality the two stations were likely hit with the exact same timestamp - so as far as we know at the same time.", "votes": null }, { "id": "133771", "postDate": "09/03/2016 18:00:32", "content": "So awesome!
", "rawMarkdown": "So awesome!", "votes": null }, { "id": "133878", "postDate": "09/04/2016 08:57:42", "content": "Great job! What software did you use to generate this?
", "rawMarkdown": "Great job! What software did you use to generate this?", "votes": null }, { "id": "133890", "postDate": "09/04/2016 10:20:16", "content": "Hey, I did the datamunging with python. The visualization is made using d3
", "rawMarkdown": "Hey, I did the datamunging with python. The visualization is made using [d3][1]\r\n\r\n\r\n [1]: https://d3js.org/ \"d3\"", "votes": null }, { "id": "134028", "postDate": "09/04/2016 18:53:12", "content": "Thanks. That's great
", "rawMarkdown": "Thanks. That's great", "votes": null }, { "id": "329756", "postDate": "05/17/2018 05:58:45", "content": "Hey this is great,could you upload some data on cellular manufacturing or parts movement on shop floor,and it would be ideal data set for classification or cell formation through commonality between parts w.r.t design,mfg,process,material attributes.Thanks.
", "rawMarkdown": "Hey this is great,could you upload some data on cellular manufacturing or parts movement on shop floor,and it would be ideal data set for classification or cell formation through commonality between parts w.r.t design,mfg,process,material attributes.Thanks.", "votes": null }, { "id": "329757", "postDate": "05/17/2018 05:59:21", "content": "Hey this is great,could you upload some data on cellular manufacturing or parts movement on shop floor,and it would be ideal data set for classification or cell formation through commonality between parts w.r.t design,mfg,process,material attributes.Thanks.
", "rawMarkdown": "Hey this is great,could you upload some data on cellular manufacturing or parts movement on shop floor,and it would be ideal data set for classification or cell formation through commonality between parts w.r.t design,mfg,process,material attributes.Thanks.", "votes": null } ], "comments": [ { "id": 133771, "author_name": "jpmiller", "author_url": "", "post_date": "09/03/2016 18:00:32", "content": "So awesome!
", "votes": null, "replies": [] }, { "id": 133878, "author_name": "malten", "author_url": "", "post_date": "09/04/2016 08:57:42", "content": "Great job! What software did you use to generate this?
", "votes": null, "replies": [] }, { "id": 133890, "author_name": "colemaclean", "author_url": "", "post_date": "09/04/2016 10:20:16", "content": "Hey, I did the datamunging with python. The visualization is made using d3
", "votes": null, "replies": [ { "id": 329757, "author_name": "bdukare", "author_url": "", "post_date": "05/17/2018 05:59:21", "content": "Hey this is great,could you upload some data on cellular manufacturing or parts movement on shop floor,and it would be ideal data set for classification or cell formation through commonality between parts w.r.t design,mfg,process,material attributes.Thanks.
", "votes": null, "replies": [] } ] }, { "id": 134028, "author_name": "domcastro", "author_url": "", "post_date": "09/04/2016 18:53:12", "content": "Thanks. That's great
", "votes": null, "replies": [ { "id": 329756, "author_name": "bdukare", "author_url": "", "post_date": "05/17/2018 05:58:45", "content": "Hey this is great,could you upload some data on cellular manufacturing or parts movement on shop floor,and it would be ideal data set for classification or cell formation through commonality between parts w.r.t design,mfg,process,material attributes.Thanks.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "133723": "I'm working my way through trying to make sense of this data, it seems like data munging is going to be an even higher % of the work in the competition then usual.\r\n\r\nI put together a viz to help give an intuition on what's going on. It's still a work in progress, but thought I'd share:https://cole-maclean.github.io/bosch-production-line-performance/\r\n\r\nFeatures are colored on defect rate and sized by total parts that filter through that feature.\r\n\r\nSeems like many features don't have an defective parts that run through them. Seems like an easy place to start in reducing the size of the dataset, but I wonder if there are some creative techniques in utilizing features that only have data for one of the 2 classes. Think I'll go do some research on that.\r\n\r\nUpdate: Updated to show all the unique paths parts can take and the count of parts that take that exact path. Numbers indicate the order in which each station is hit. Should help to visualize the different product types flowing through the factory, and the order stations are hit. The resolution of the timestamps are not very good, so even tho the ordering might change, in reality the two stations were likely hit with the exact same timestamp - so as far as we know at the same time.", "133771": "So awesome!", "133878": "Great job! What software did you use to generate this?", "133890": "Hey, I did the datamunging with python. The visualization is made using [d3][1]\r\n\r\n\r\n [1]: https://d3js.org/ \"d3\"", "134028": "Thanks. That's great", "329756": "Hey this is great,could you upload some data on cellular manufacturing or parts movement on shop floor,and it would be ideal data set for classification or cell formation through commonality between parts w.r.t design,mfg,process,material attributes.Thanks.", "329757": "Hey this is great,could you upload some data on cellular manufacturing or parts movement on shop floor,and it would be ideal data set for classification or cell formation through commonality between parts w.r.t design,mfg,process,material attributes.Thanks." }, "source": "meta" }