This presentation focuses on Feature Engineering and the Heuristics which can be extracted pre modelling whihc can be used post modelling for change detectors, explain ability etc.
2. Data Science Life Cycle
1- Scale Wide Data
Explicit Feature
Selection
-Dimensionality
Reduction
Methods
2- Modelled Data
Implicit Feature
Selection
- Modelling
/Segmentation
Algorithms
3 – Un Checked Bounds
Unbounded at
Leaf Nodes
Un Checked
attributes -
greedy attribute
split
4- Un Processed Dimensions
Features Left in
Steps 1 ~3
5 - Unprocessed
Samples
Left by
Modelling
Algorithms to
Reduce Variance
3. Data Science Life Cycle
1- Scale Wide Data
Explicit Feature
Selection
-Dimensionality
Reduction
Methods
2- Modelled Data
Implicit Feature
Selection
- Modelling
/Segmentation
Algorithms
3 – Un Checked Bounds
Unbounded at
Leaf Nodes
Un Checked
attributes -
greedy attribute
split
4- Un Processed Dimensions
Features Left in
Steps 1 ~3
5 - Unprocessed
Samples
Left by
Modelling
Algorithms to
Reduce Variance
Extract Per Class Feature Heuristics
Automate Bound Checks
Automate - Confidence Score Generation
, Bound Check Violations
Configurable Bound Checks
Index
4. References – PHD Thesis / Book
• Feature Selection Problem • Feature Selection for High
Dimensional Data