Principles of advanced data analytics
IBM® Cognos Analytics is a business intelligence tool for managing and analyzing data. It includes self-service features for users to prepare, explore, and share data. Cognos Analytics includes predictive, descriptive, and exploratory techniques, also known as numeric intelligence. Cognos Analytics uses many statistical tests to analyze your data.
It is important to understand the definitions of these tests as they apply to Cognos Analytics.
Numeric algorithms are used as part of the workflow to provide features to the user that communicate information about the numeric properties and relationships in their data.
Unlike traditional statistical software, where the target audience is an experienced data analyst, the algorithms of Cognos Analytics are aimed at users who are familiar with, but not an expert in data analysis. This means that when Cognos Analytics considers tradeoffs, usefulness is chosen over complexity.
Business data is lot more complicated than text-book examples that are used in statistical courses or in web searches and examples. Cognos Analytics uses algorithms that are robust and that are able to cope with a range of varieties of unusual data. Cognos Analytics does this because although the algorithms that are more brittle are able to get slightly better results than robust algorithms, they require you to make sure that they are applicable and build correct data transformations for the results to be meaningful. A minor drop in accuracy is worth the safety that is provided by an algorithm, which does not give wrong results when the data is not as it is supposed to be.
Almost all algorithms require decisions to be made about them; levels of confidence, which combinations of fields to explore, data transformations. The details of these decisions can be found in the descriptions.
Cognos Analytics chooses appropriate values automatically, by examining properties of the data. As a user you might not discover all the decisions that are made.
In Cognos Analytics, the numeric algorithms and procedures are designed to produce trustworthy results automatically. To get the best possible prediction, classification or analysis, a professional statistician analyzes the data by using IBM SPSS Statistics or IBM SPSS Modeler. The goal of Cognos Analytics is to provide qualitative insights that help you to understand your data and its relationships, and to do so automatically for a wide variety of types of data. Cognos Analytics aims at providing results similar to a professional statistician without getting in the way of the business user.