The Spotfire recommendations engine

The recommendations engine gives you instant insight into your data and some suggestions of which visualizations to use. Spotfire's analyst clients have an advanced feature called AI-Powered Suggestions. This is Spotfire's new way of helping make sense of any type of data, regardless of its size or shape. The basic premise is that you should select a "target" column in the data panel and Spotfire will do the rest. Spotfire runs a specialized algorithm over all the columns in the data and selects those that most strongly drive or influence the target. Those columns are called "predictors." It then produces suggested visualizations for the target and selected predictors.

The recommender is available in the web clients too, but (at the time of writing) the web clients do not have the AI element, where predictor columns are automatically selected. I hope that the feature will be made available at some point. In the meantime, if you're using the web clients, you can follow a slightly different path to create visualizations. I'll point out how to do that along the way.

Let's get started! In the case of the Titanic data, the most obvious target column is survived. In other words, we'd like to know which columns best predict, influence, or explain whether passengers survived the Titanic disaster or not:

1. In the data panel, select the survived column. In analyst clients, Spotfire will produce something that looks like this:

Interesting! Immediately, we can see that the strongest predictors of survival are pclass and sex. The very first visualization is always just the row count of each of the values in the target column, so the second visualization is the one that begins to explain the target.

1. To add the visualization to your analysis, just click on the visualization that shows the relationship with pclass and sex. Your Spotfire session should now look something like the following screenshot:

1. You can produce the same effect in Spotfire web clients by selecting the survived column, the pclass column, and the sex column. Hold down the Ctrl key while clicking to select multiple columns:

1. I think it's also interesting to explore the male/female ratio on board the Titanic, so we need to add a bar chart visualization that shows just survived and sex. The AI recommender will choose these columns—just scroll down the panel a bit to find the visualization that shows this relationship, or manually select the columns in the web clients, then click the visualization to add it to your analysis.
   Now, let's pause and look at what we've created. In a few clicks, we have loaded some data and created two visualizations that really explains a lot of findings (insights) all in one go! Here is the analysis without the data panel (collapse it by clicking the double arrow toward the top-right corner of the panel):

Here are my notes on interpreting these visualizations. Of course, what I say is only a matter of opinion, so feel free to draw your own conclusions:

1. Notice how many more men there were on board than women?
2. Look at how many more men died than survived! The survival rate of women was much higher—this would be borne by the "women and children first" policy of filling the lifeboats.

1. Look at the left-hand visualization. It is trellised by survived—this means that Spotfire has split it into panels, one for each data value.
2. First-class female passengers had the greatest chance of survival—suggesting that class and socioeconomic factors played a role in the survival rates.
3. More third-class males survived than first- or second-class ones. Might they have had stronger fighting instincts or been pushier? Were they more willing to cram into overcrowded lifeboats? All are potential explanations.