Overview: Data-driven decisions from big data, data science, and predictive analytics has quickly become among the most important areas of growth in organizations. Skills in these areas are specialized and often lacking in traditional education.
Effective predictive modeling does not require a PhD in mathematics, statistics, or hard science fields to do well. Many effective and even famous data miners and predictive modelers have BS or BA degrees in non-technical fields. However, it does require a qualitative understanding of what the predictive modeling process s about, what algorithms do, what their limitations are, how to change their behavior, and what kind of data is necessary for building predictive models.
Even individuals with experience in analytics understand that predictive modeling requires not only an understand of the science, but also decisions throughout a modeling project that are not (indeed cannot be) governed fully by the science; there is "art" and tradeoffs we as analysts make at every stage. These aren’t guesses, but are rather governed by the principles of predictive modeling: sampling, data distributions and their effects on summary statistics and the modeling algorithms, and how to determine if a model is good or not.
Areas Covered in the Session:
CRISP-DM - what are the main steps in the predictive modeling process
Key steps in defining modeling objectives
The most important principles in setting up data for modeling
Brief overview of key modeling algorithms
Matching model accuracy to business objectives
Who Will Benefit:
Big Data Analysts
Functional Analytic Practitioners
Anyone Overwhelmed with Data