Please enable JavaScript.
Coggle requires JavaScript to display documents.
Chapters 2-7 (Chapter 4 acquire subject matter exercise (without subject…
Chapters 2-7
Chapter 4 acquire subject matter exercise
without subject matter expertise or access to it, you will not be capable of providing complete insights
constitutes deep experience in s specific domain such as accounting, marketing, supply chain, or medicine
need subject matter expertise is setting realistic expectations for model performance
we also expect the subject matter expert to suggest ideas for data collection, that is, to know where relevant data resides, including and external data
Chapter 2 Automating Machine Learning
lifecycle
define project objectives
Acquire exploring data
Model data
interpret & communication
implement, document &maintain
looks linear but is no a linear process
auto ml is everywhere and a lot of companies are adopting ml to better their business
it is not automatic ml
decisions need to be made by the analyst
they need to decide what is worthy to be interpreted
two types of tools and platforms that are the main ones
context-specific tools
these are implemented within another system or for a specific purpose
general platforms
those that are designed for general-purpose machine learning, splits into two types
open-source
tools tend to be developed by and for computer and data scientists and generally require knowledge of programming languages
commercial
provided by a commercial vendor for money
8 criteria for AutoML excellence
Accuracy
most important, without this there is no reason
productivity
Ease of us
should be easy to use
understanding and learning
resource availability
process transparency
Generalize across contexts
recommend actions
Chapter 3 specify business problem
start with a business problem
keep in mind that problem is equal to saying opportunity
perfection is bad if the wrong problem is addressed
make sure you specify the problem carefully
once we ensure that we have the requisite data, the project must then be described in a brief to be shared with stakeholders
Chapter 5 Decide on Unit Analysis
this involves intricate understanding of a problem's real-world context
unit of analysis
what, who, where, and when of our analysis
determining unit of analysis
general sense of unit of analysis as a concept
one way to determine the appropriate unit of analysis is to first to think about what the prediction target is, which will then oftentimes supply an obvious choice
unit analysis is the test subject
sometimes the unit of analysis isn't obvious
Chapter 6 Define Prediction target
critically important skill to develop as a well-founded understanding of prediction targets may enable the inspiration to use ML in new and ingenious ways
prediction target
is the behavior of a "thing" we need to know about the future
why need a target
you find out more why once we learn more about ML
in a simple way
without a target, there is simply no way for humans or machines to learn what associates drive an outcome
two kinds of targets
classification
predicts the category to which a new case belongs
Regression
to predict the target numeric values
Chapter 7 Success, risk, and continuation
identify success criteria
evaluating project objectives, a better understanding of success criteria will develop over time, both within an individual project and across many projects
being involved early increases the success
looking at risk
this is hard to calculate
to get a the risks, you will nee to be creative and often play the devil's advocate
model risks might relate to the model being insufficiently predictive or, alternatively, "simple" mistakes such as target leakage features in the model, which result in models that are too predictive
Decide whether to continue
after weighing the risks against the rewards, seriously evaluate whether to move forward with your project
change the cost and benefit analysis
getting feedback from subject matter experts on that pilot should help important here to evaluate whether additional data could move a project above the required threshold for success
quick pilot projects can also be a wonderful way of garnering interest and buy-in for your project idea, particularly because it immediately communicated the minimum level of success management and the company expert