Community Tip - You can subscribe to a forum, label or individual post and receive email notifications when someone posts a new topic or reply. Learn more! X
Hello,
is it possible to test a time series model by passing the input data row values with a file (test values/file with test data)?
Is my understanding correctly, that the upload input file button is only used to provide the possibility to pass an file by url in the input data row values?
Solved! Go to Solution.
Hi @skef
Based on your previous post you are using ThingPredictor for scoring.
The aim of ThingPredictor is mostly for real time scoring, therefore limited amount of data at one time.
If you have got a new csv file with new rows to score against a previously created model, you should use the Analytics Server scoring feature as opposed to ThingPredictor (they all use the same engine behind the scene)
For this:
In ThingWorx Analytics Builder select Models > select your model > View > Predictive scoring > New > Upload New data, load your csv file with data to be scored. Crete a filter to segregate those new data (otherwise the whole dataset including the training data will be scored).
Once the scoring job is complete you will have a link to download the result.
Hope this helps
Christophe
Hi @skef
Can you please confirm whether you are looking to build a fresh model by time series data? In that case you need to have this time series data in .CSV and Json format to upload on Analytics builder and create model.
Regards,
Mohit
No, I want to test a model based on time series with data from a csv file. Currently, I have to enter the data for every single row manually. I want to put this data in a file and make twx analytics to read the values in the file and test the model based on this data.
Hi @skef
Based on your previous post you are using ThingPredictor for scoring.
The aim of ThingPredictor is mostly for real time scoring, therefore limited amount of data at one time.
If you have got a new csv file with new rows to score against a previously created model, you should use the Analytics Server scoring feature as opposed to ThingPredictor (they all use the same engine behind the scene)
For this:
In ThingWorx Analytics Builder select Models > select your model > View > Predictive scoring > New > Upload New data, load your csv file with data to be scored. Crete a filter to segregate those new data (otherwise the whole dataset including the training data will be scored).
Once the scoring job is complete you will have a link to download the result.
Hope this helps
Christophe
Regarding this Questions, I am wondering what is the "Upload models with file inputs" as explained in http://support.ptc.com/help/thingworx_hc/thingworx_analytics_8/#page/thingworx_analytics_8%2FAnalysisServices_UploadingDifferentModels.html for?
The Analytics Manager-Test seems to provide user with the feature to input by file instead of manually have to key in one by one, but it does not provide an example of the file format and etc which make it impossible to use this feature and it becomes quite redundant?
Hi pcheah
The Upload Input Files menu is for a different purpose than uploading input data.
This is only used if a custom model takes a file as input, it is not to upload all input data from a file.
I give you a theoretical example to make it hopefully clearer.
Let's assume that you have a image recognition model.
When needing to score a new image, you would need to upload the new image file. This is where this Upload Input Files will be used, because the model would take a file as input.
None of the OOTB agents have this feature, so this is provided in case of need from custom agent .