I am not sure, which math model predicts more than one recommendation!!! How do we generate 3 priority outputs then ?
\n\nShall we run the model 3 times hoping that recommendations will be different each time (though that makes model less stable)?
", "messages": [ { "id": "121256", "postDate": "05/25/2016 07:10:40", "content": "I am not sure, which math model predicts more than one recommendation!!! How do we generate 3 priority outputs then ?
\n\nShall we run the model 3 times hoping that recommendations will be different each time (though that makes model less stable)?
", "rawMarkdown": "I am not sure, which math model predicts more than one recommendation!!! How do we generate 3 priority outputs then ?\r\n\r\nShall we run the model 3 times hoping that recommendations will be different each time (though that makes model less stable)?", "votes": null }, { "id": "121258", "postDate": "05/25/2016 07:28:52", "content": "Most classifiers are also capable of generating a vector with probability estimations (e.g. for scikit-learn you can use the function predict_proba).\nYou can sort them & select the top 3 values. If you use Python, here is a way to do this: http://stackoverflow.com/questions/32461246/how-to-get-top-3-or-top-n-predictions-using-sklearns-sgdclassifier
", "rawMarkdown": "Most classifiers are also capable of generating a vector with probability estimations (e.g. for scikit-learn you can use the function predict_proba).\r\nYou can sort them & select the top 3 values. If you use Python, here is a way to do this: http://stackoverflow.com/questions/32461246/how-to-get-top-3-or-top-n-predictions-using-sklearns-sgdclassifier", "votes": null }, { "id": "121266", "postDate": "05/25/2016 08:17:40", "content": "Thanks Daniel for your help. Do we have something similar in R?
", "rawMarkdown": "Thanks Daniel for your help. Do we have something similar in R?", "votes": null }, { "id": "121317", "postDate": "05/25/2016 15:35:50", "content": "For R, look up the help documentation for the classifier you are using (more specifically, usually its predict method). Most likely, there will be some toggle such as "probability=TRUE" to have the model generate probabilities for you. An example for randomForest's predict function:
\n\n"type\none of response, prob. or votes, indicating the type of output: predicted values, matrix of class probabilities, or matrix of vote counts."
\n\nso we can set it up by doing something like:
\n\npredictions <- predict(rf_model, test_data, type='prob')
", "rawMarkdown": "For R, look up the help documentation for the classifier you are using (more specifically, usually its predict method). Most likely, there will be some toggle such as \"probability=TRUE\" to have the model generate probabilities for you. An example for randomForest's predict function:\r\n\r\n\"type\r\none of response, prob. or votes, indicating the type of output: predicted values, matrix of class probabilities, or matrix of vote counts.\"\r\n\r\nso we can set it up by doing something like:\r\n\r\npredictions <- predict(rf_model, test_data, type='prob')", "votes": null } ], "comments": [ { "id": 121258, "author_name": "danielheres", "author_url": "", "post_date": "05/25/2016 07:28:52", "content": "Most classifiers are also capable of generating a vector with probability estimations (e.g. for scikit-learn you can use the function predict_proba).\nYou can sort them & select the top 3 values. If you use Python, here is a way to do this: http://stackoverflow.com/questions/32461246/how-to-get-top-3-or-top-n-predictions-using-sklearns-sgdclassifier
", "votes": null, "replies": [] }, { "id": 121266, "author_name": "sandeepak", "author_url": "", "post_date": "05/25/2016 08:17:40", "content": "Thanks Daniel for your help. Do we have something similar in R?
", "votes": null, "replies": [] }, { "id": 121317, "author_name": "fireiceearth", "author_url": "", "post_date": "05/25/2016 15:35:50", "content": "For R, look up the help documentation for the classifier you are using (more specifically, usually its predict method). Most likely, there will be some toggle such as "probability=TRUE" to have the model generate probabilities for you. An example for randomForest's predict function:
\n\n"type\none of response, prob. or votes, indicating the type of output: predicted values, matrix of class probabilities, or matrix of vote counts."
\n\nso we can set it up by doing something like:
\n\npredictions <- predict(rf_model, test_data, type='prob')
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "121256": "I am not sure, which math model predicts more than one recommendation!!! How do we generate 3 priority outputs then ?\r\n\r\nShall we run the model 3 times hoping that recommendations will be different each time (though that makes model less stable)?", "121258": "Most classifiers are also capable of generating a vector with probability estimations (e.g. for scikit-learn you can use the function predict_proba).\r\nYou can sort them & select the top 3 values. If you use Python, here is a way to do this: http://stackoverflow.com/questions/32461246/how-to-get-top-3-or-top-n-predictions-using-sklearns-sgdclassifier", "121266": "Thanks Daniel for your help. Do we have something similar in R?", "121317": "For R, look up the help documentation for the classifier you are using (more specifically, usually its predict method). Most likely, there will be some toggle such as \"probability=TRUE\" to have the model generate probabilities for you. An example for randomForest's predict function:\r\n\r\n\"type\r\none of response, prob. or votes, indicating the type of output: predicted values, matrix of class probabilities, or matrix of vote counts.\"\r\n\r\nso we can set it up by doing something like:\r\n\r\npredictions <- predict(rf_model, test_data, type='prob')" }, "source": "meta" }