Once the files are gathered, we can use IBM modeler to set up and save a stream. Again, this book is not intending to instruct on modeler, but a stream is the main document type used by IBM modeler. Your streams can be saved, re-loaded, re-edited, and re-executed. Streams can also be set up to run from a script and take parameters, global values, execute other scripts, and so on.
Our simple stream can:
- Read the individual driver files, rename columns, and interpret specific data values. The same logic can be applied to all of the driver files, or each file can have its own unique logic.
- Merge/append all of the individual files into a single file (remember, if you are not on the professional version, you may run into row/record limitations).
- Write out a single, consistently named file (ready for uploading to IBM Watson).
The following screenshot shows our simple SPSS modeler stream:
Once our process has created our combined file, we can upload it to Watson. The steps are as follows:
Once our file is available to Watson, we can start with our analysis. We can start with a Predict. Let's start with a single Target of Trip Issues:
Watson provides our Top Predictors for our Target (Trip Issues), which is shown in the next screenshot. Notice the Collect icon on the upper-left corner:
Clicking on the Collect icon adds to our collection. This is not really what you would expect though; our Bus Runs has been added to the collection (see the following screenshot), but it's not very helpful for later use because the name is so vague:
From an enterprise naming perspective, you should strive to be very specific with your predict name. Once it's created, it is somewhat difficult to rename. For example, going back to the main Welcome page and clicking on the Predict panel, you can click Rename:
Then, you can give it a more descriptive name, like Bus Runs – Trip Issues.
But, be aware that if you open that prediction and click on the Collect icon, Watson still saves the Predict under the original name:
Our collected artifacts can be seen on the screen:
The only way to properly name our artifact is to create a new Prediction (with the same file and target) and name it what we want:
Then, add this prediction to our collection (using the Collect icon):
To demonstrate how Watson uses versioning, let's edit our prediction, changing our Target from Trip Issues to Tip Grade (as shown in the following screenshot):
Next, using the Versions feature of Watson, click on the Properties icon (shown in the following screenshot) in the upper-left corner:
Then, select VERSIONS:
The Versions feature of IBM Watson does supply some help by adding a Change Description:
This new version still Collects using the same predict name:
You can see that, without taking care to use proper naming conventions, versions and collections can quickly become `more of a burden than a useful resource.