Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
Do you really want to delete this prezi?
Neither you, nor the coeditors you shared it with will be able to recover it again.
Make your likes visible on Facebook?
Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.
Data Science vs. Analytics: Approaches to Problem Solving
Transcript of Data Science vs. Analytics: Approaches to Problem Solving
My Boss: email@example.com
Cool Person: firstname.lastname@example.org Nick Kolegraff, Lead Data Scientist @ Accenture Puzzles Two Types: pieces exist. we just need to put them together. All pieces might not exist. Solutions are unknown. Problem deconstruction becomes critical. "How" Approach + Abstract Problem = Cool solution. Wrong problem. Solving for the "How" without understanding the "Why" and you get a solved problem for the wrong reasons. Assumptive When variable relationship is better understood. Temperature = Hot | Cold | Warm Wear = Shorts | Coat | Pants Assume nothing-ish and attempt to account for everything, kinda. when variables have a complex relationship. Being = Mother | Girlfriend | Grandmother | Dog
Action = Kiss | Call | Hug | Play Target specifically and adjust accordingly. "Why" Approach + Abstract Problem = Good Solution. Understanding the "Why" then solving the "How" and you get a solved problem for the right reasons. After multiple failed attempts of me trying to explain what I do differently to my boss, Vince, he suggested this. Not Just Database Structured
Signal Data Types
Data Size Millisecond,
Days Speed: BS Statistics, BA Computer Science Culture is Key to solving these problems Hard to structure innovation, better to enable it. Abstract Non-Assumptive generally don't ask why, the pieces are in the box we just need to assemble them. How? We Have to ask 'why' in order to understand the unknowns. WHY: We need both and use both so really,
Data Science + Analytics Product vs Project Why xTB, do we need that?
Could we sample down, then scale up? Asking 'why' naturally forces you to deconstruct a problem to a more pure form. Standards Structure Efficiency and Consistency. Unstructured-ish Process Innovative With puzzles, we don't have to ask 'why'
we just define how we do it. Discriminative Modeling Generative Modeling Creativity, Curiosity, Creation How is difficult because Why wasn't solved. <Many Business Problems Go Here> When a problem makes it here, we call it "industrialized" Industrialize? Psssh. #innovationalize Ugh. It didn't work. If we are industrialized ... how do we innovationalize? Why? vs. Understanding what pieces we need for a possible solution requires deep questions. Make you question that something different might exist. You Decide. Problems
Types of Modeling
Decisions Not just infrastructure needs to support these.
models do too. Trying to solve abstract problems in a 'how' culture is difficult. Cool... Cool... Tying this in to modeling .. But, not only do we have different modeling approaches. Infrastructure is different too. Types: Size Not limited to disk, billions of ids are large too Creates many abstract unsolved problems. (Puzzle) (Abstract) Come talk during office hours right after this. Accenture Data Science. Solve for why and how becomes easy. There is more... Calculus helps us understand our world.
Statistics helps us live in our world. Translation: We have decisions to make. Conscious Instinctive Data Repetition. aka practice. (think athletics) You don't know why, you just know. (think athletics again) Critical for business leaders. (many decisions are still instinct) I don't think that will change anytime soon.
Imagine ... (hold on while I take you down) But how do we get better? ... we have data, models, infrastructure, decisions with specific requirements ... Because we don't really know what we are doing. We have xTB of data, we want to predict the probability of <event> How do we do that? .... You'll focus on scale when you probably don't need it. Still Difficult.