#1 Class on 14th Aug 2017
Course outline
1. Introduction to Big Data Potential, Pitfalls & Strategy Thinking
2. Supervised and Unsupervised Learning
3. Clustering and Classification, why this is important?
4. Game - Can your job be replaced by a machine?
Note-taking
Definition of "strategy"
- where to compete
- how to compete
- Michael Porter's 5 forces vs. 6 forces (+ compliment)
* Compliment: e.g., air ticket + hotel room or ground transportation
Option1: good strategy + bad execution
Option2: bad strategy + good execution
Option2 is the way preferable over option1. That is, the execution is very very important in real business. Even if you set the bad strategy and do good execution, you can figure out quickly when business goes wrong and then do something to solve the problems. Before executing your strategy, we cannot make sure what will happen. To see whether our strategy will work for us or not, we MUST do execution fast and change our strategy if needed (resilience).
There is no sustainable strategy but transitional strategy. So, how fast you can execute your strategy is the most important thing.
Machine Learning: the ability to learn without being explicitly programmed.
Big Data: 3V (Volume - IOT, Velocity, Variety)
- Big Data doesn't give us 100% correct answer.
- Big Data just shows us "correlation", not "causality (why)."
- Big Data doesn't have the concept of "outliers." (Unlike statistics - R-sqr, etc, there is no thumb of rule for cut-off indicating whether the result is good enough or not.)
- Big data Analytics: predicting the future helping companies for their future successes, not solving the problems.
Supervised Learning (e.g., lecture) vs. Unsupervised Learning (e.g., jungle)
- Labels (S.L) vs. Groups (U.L)
- Unsupervised learning is hard to measure.
Classification vs. Clustering
- classification: supervised learning problem (label)
- clustering: unsupervised learning problem (group)
- The clustering is the kind of grouping but you may not know what each group is exactly.
- For example, imagine the picture of a bird on one's hand with green background
- classification: bird, hand, green background
- clustering: 3 groups by shape, object, etc. (The computers don't know what they are.)
Human insight vs. Decision-making using data analytics
- when and where human insight is more helpful for business and vice versa
Course outline
1. Introduction to Big Data Potential, Pitfalls & Strategy Thinking
2. Supervised and Unsupervised Learning
3. Clustering and Classification, why this is important?
4. Game - Can your job be replaced by a machine?
Note-taking
Definition of "strategy"
- where to compete
- how to compete
- Michael Porter's 5 forces vs. 6 forces (+ compliment)
* Compliment: e.g., air ticket + hotel room or ground transportation
Option1: good strategy + bad execution
Option2: bad strategy + good execution
Option2 is the way preferable over option1. That is, the execution is very very important in real business. Even if you set the bad strategy and do good execution, you can figure out quickly when business goes wrong and then do something to solve the problems. Before executing your strategy, we cannot make sure what will happen. To see whether our strategy will work for us or not, we MUST do execution fast and change our strategy if needed (resilience).
There is no sustainable strategy but transitional strategy. So, how fast you can execute your strategy is the most important thing.
Machine Learning: the ability to learn without being explicitly programmed.
Big Data: 3V (Volume - IOT, Velocity, Variety)
- Big Data doesn't give us 100% correct answer.
- Big Data just shows us "correlation", not "causality (why)."
- Big Data doesn't have the concept of "outliers." (Unlike statistics - R-sqr, etc, there is no thumb of rule for cut-off indicating whether the result is good enough or not.)
Operational Analytics vs. Big data Analytics
- Operational Analytics: finding the exact solution to the current problem (real-time, 5 whys)- Big data Analytics: predicting the future helping companies for their future successes, not solving the problems.
Supervised Learning (e.g., lecture) vs. Unsupervised Learning (e.g., jungle)
- Labels (S.L) vs. Groups (U.L)
- Unsupervised learning is hard to measure.
Classification vs. Clustering
- classification: supervised learning problem (label)
- clustering: unsupervised learning problem (group)
- The clustering is the kind of grouping but you may not know what each group is exactly.
- For example, imagine the picture of a bird on one's hand with green background
- classification: bird, hand, green background
- clustering: 3 groups by shape, object, etc. (The computers don't know what they are.)
Human insight vs. Decision-making using data analytics
- when and where human insight is more helpful for business and vice versa
