I imported the data set in the Quick Start - the import step shows all Green. However I am not able to move forward - unable to figure out what is the issue!
@kprasad - To be able to Quick Start the project you would have to select the target. You would then see the start button ready to start AutoML.
If still you dont see it active. Please try and refresh as suggested by @dalilaB above. If this happens often please make sure you use Google Chrome as that is preferable browser.
Feel free to reach out if you are still having problems starting the project. Thanks
hi @kprasad the only thing I will add here is that if you are working on an unsupervised learning project, either clustering or anomaly detection, you do not need to select a target variable. You can click on a button called 'no target' right under the box where you are supposed to enter the target name and move forward from there.
I am also having problems being able to click on the Start. Refreshing does not seem to do anything. Also, can we be recommended a sample data file to practice on?
Actually neither as the picture hopefully shows how I am starting with a simple regression. My Dependent or Response variable here is OrderWeight - and it is a continuous/numeric field. One idea that I am thinking could be the culprit here is the InvoiceDate field. Typically when it is read in via Praxata it sees it as a string field to be cast as Date in one of the Praxata steps. I think that the date was autodetected here shortly after reading in the .csv. Am thinking maybe that the model is waiting for me to absolutely declare it as Date??? but just an idea.
I tried removing the InvoiceDate already hoping that it would Trigger the Start, but it did not.
But the instructions on my screen are different from what you suggest. Below is a picture of my available options for TS forecasting & nowcasting.
If I next click on the Forecasting button then it brings me to:
It does not permit me to use the Truck as my series name.
But if I click on Show Advanced Options it looks pretty different from the provided example:
That specifically was where I got the idea of my date not being formatted correctly.
Am also thinking that perhaps my InvoiceDate is faulty because of the gaps in there. It is not continuous, and it is mainly weekdays.
Just an idea...
If you have weekdays mainly, go to first AI Catalog, upload your dataset to AI Catalog, and then you will see a humberger on your top right, click on it, and then choose prepare dataset for time-series, else:
If the invoiceDate is not continuous, when choosing TS, choose row instead of date base.
Here is an example:
After deciding on the date, go to Automated Time Series, and if you have series add the series, but then
If you get an error, choose row. You can also click on Time Series Data Prep which will take you to AI Catalog where the data can be cleaned.
Wow is getting a little exciting. Thank you. Unfortunately the devil resides is some details. Please see the following screen shot:
This is what the interactive menu showed before I did anything related to the Series ID. Having explored the documentation it looks like maybe I should have used my Truck field for the Series ID, but am not sure. Please advise here.
Some other interesting tidbits + features (pun intended) follow:
I am next going to review what I think is the new dataset within the AI Catalog that is more suitable for TimeSeries analyses.
Again, DataRobot is very good at removing the coding aspect that I am accustomed to over the course of many weeks/months. Miss it a bit though... Imputation can be fun!
So I was able to see the datafile within the AI Catalog. I thought I was being clever in downloading the file from the hamburger, but I think it just downloads the original .csv file again. Below are some screen-shots of the file residing in the AI Catalog:
Again I downloaded it from the hamburger in the upper-right:
I renamed the downloaded dataset to something "Interpolated."
But where should I retrieve this data from?
This is where I am currently at. Am a little embarrassed that I am drawing a blank on what to do next.
Please take an easy look at some of my other questions within this thread for other specific areas such as the need, or not, of defining the Series and where/when should that occur, as well as how to check the quality of my work through the many steps.
If you are just forecasting order weight irrelevant of a truck, than you don't have a series, and you should just ignore the truck series id. So, when filling Prepare dataset for Time-series, just don't fill the series id section.
Here are the three steps:
1. Fill the form for prepare dataset for time series
This is what you will get, as you notice a suffix was added at the end of the original dataset name, and it is still registering.
whenNow that the dataset is registereted, just click on create Project
I completely agree that doing the DRU courses will be very valuable. A problem that I am having is with some of the basic or beginning labs:
This is very early in the instructions.
Even my graph in the upper-right is different from the instructions:
The following snapshot of the Time-Aware modeling sets the Default prediction window to 7 days. I set it to 30 per the instructions and receive the following error:
I still get the same error when setting the window to 7 days instead of 30.
I normally would not care a lot about minor differences especially with different versions, but I worry how difficult it will be to follow the lessons after even having problems with the early ones. Am currently thinking instructor-led courses may be an ideal solution? Please advise on the best strategy, and this is a summary of why I wish to try and bypass some of the DRU offerings. Please let me know if or if not this is the correct forum for discussing such issues.
email@example.com would you like me to delete the previous 2 posts (the one about DRU feedback and the one saying you are trying to delete it)? Not suggesting to do so, but if you'd like me to, I'd be happy to.