Once the data has been loaded, we can get to know it a bit. One way of doing this is to create a prediction on our file, but starting with a new analysis, using the following steps:

1. Click on Predict and select our file. On the Create a new analysis page, you can click on the link that says Edit this workbook's field properties, as shown in the following screenshot:

2. Once you click on the link, you will see the Field Properties page, as shown in the following screenshot:

3. On this page, every column found in the file is listed down the left-hand side of the page. Clicking on the column selects it and allows you to see the name of the column, the label being used for the column, its role, and its measurement level.

Although we won't actually change anything at this time, scrolling through the data columns in this way is an easy way to get to know our data file. Scrolling through this list, we can see some interesting data columns, including those associated with patient history, listed as follows:

• Number of illnesses
• Number of hospital stays
• Number of surgeries

I would think that it would be of value to determine whether there is anything within the dialog data that may drive up the number of illnesses, hospital stays, and/or surgeries for patients. In other words, can we use Watson Analytics to find a pattern within our data that can be used to predict an outcome?

Going further, we can identify some prediction targets, such as the number of illness, (hospital) stays, and surgeries. However, at the moment, we don't know how to identify one or more predictive drivers within our dialog data. One option could be the patients' responses to questions such as Do you smoke? or Do you use recreational drugs?, or perhaps more emphasis should be on dialog statistics, such as the (total) length of time that the physician speaks during the dialog (or the length of time that the patient speaks).