SPSS Modeler: advanced data preparation (0A058G)

• Teams with streams so slow nobody runs them twice.
• Anyone with dates in five different formats in the same field.
• Analysts with a badly skewed target — 2% fraud — and models that learn nothing.

• Functions to cleanse and enrich data: date, conversion, string, statistical and missing value functions
• Further field transformations: the Filler node, binning continuous fields, and changing distributions with the Transform node
• Working with sequence data: sequence functions, counting an event across records, the Restructure node and Space-Time-Boxes
• Sampling, partitioning and balancing: the Sample node, the Partition node for training and testing, and the Balance node
• Improving efficiency: SQL pushback, outliers and missing values with the Data Audit node, global values, parameters, looping and conditional execution

• Labs on IBM SPSS Modeler with real data: streams built, run and fixed during the course.
• Taught directly by IBM Champion instructors from the SIXE team.
• Credly badges available: credly.com/organizations/sixe/badges
• Flexible format — online, classroom or onsite — in closed groups, priced per attendee, in the language of your choice (ES / EN / FR).

What is SQL pushback and why does it matter?
Modeler translating part of the stream into SQL and running it in the database instead of pulling the data over. On large tables it can change the run time completely, and the course shows how to check which nodes take advantage of it.

Does it address imbalanced classes?
Yes, with the Balance node: when to reduce the majority class, when to boost the minority one, and what that does to the probabilities the model returns.

Group size and language?
Closed groups of four to ten people, in Spanish, English or French. We run it with fewer than four as well, though it usually pays to wait and fill the group.

• 0A069G — Modeler foundations
• 0A079G — machine learning models
• 0A038G — advanced predictive modeling