This site uses different types of cookies, including analytics and functional cookies (its own and from other sites). To change your cookie settings or find out more, click here. If you continue browsing our website, you accept these cookies.
Hi, I'm starting with Alteryx, I have an interesting dataset to try to correlate certain behavior with time variables, such as "month of the year", "day of the month", "day of the week" and "hour of the day". I have used a Neural Network block, and it seems it works, I'm attaching a "Browse" block to the "R" connector (Report) and I see a beautiful report showing "effect plots" for each variable and other interesting data (I say "beautiful" because of the "so relevant" information it gives about the problem :)
My question is this: I don't know what to do with the "O" connector of the Neural Network block, I've read the Help Page on this subject, https://help.alteryx.com/9.5/Neural_Network.htm and it says the Output is "a table of the serialized model with its model name", I don't understand this, I'm seeing an Object with 500.000 or so bytes long. ¿What is that? ¿How can I "browse" it? The Browse block only says "View Data Table in Results window". I looked for examples but found none.
The O output of the Neural network outputs the actual model that was created by the neural network tool, so that you can use that model against other datasets. You can use the score tool to score/predict additional data. This is one nice way to look at how much overfitting exists with your model. Just plug the O output from the neural net into one of the inputs of the score tool, and the additional data into the other input of the score tool.
As far as to whether the serialized format of the model is usable outside of Alteryx, that I do not know. Hope this helps!
Thanks for your response. It works perfectly, the model works ok! :) Now I have another question: It is possible to "export inside Alteryx" the model somehow? I ask because I have a very large "first" dataset (and a little "second", dataset (to be predicted)), so it takes a very long time to run the workflow and build the model, Can I "save" the generated model somehow and use it directly as a block against my second dataset?
If you add an Output Data tool to the O output of the Neural Network tool and save to an Alteryx Database (yxdb) format that will preserve the trained model for future use.
You can then create a new workflow, drop in an Input Data tool and point to this saved file (the model from the earlier step) and send this directly into one of the two inputs for a Score tool. The other input will be the data you wish to test in your second dataset.
Hope this helps? It's a common workflow pattern - use one workflow to generate the predictive model, then use the model in other workflows (where the scoring step is typically much faster).