1. MOTIVATION

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Typically, a ML model is unidirectional

Given

what's the value of      ?  

X
Y
  • Idealized scenario: if you do not know the exact task on beforehand, this paradigm breaks down.
  • E.g.: anomaly detection (e.g. on sensor data), prediction in spreadsheets, ...

Disadvantage

f_{UD}:
X
\rightarrow
Y

Motivation

A typical ML model can solve this

Given

what's the value of      ?  

X
Y
  • Idealized scenario: if you do not know the exact task on beforehand, this paradigm breaks down.
  • E.g.: anomaly detection (e.g. on sensor data), prediction in spreadsheets, ...

Disadvantage

Our solution

f_1:
X
\rightarrow
Y
f_2:
X
\rightarrow
Y
f_3:
X
\rightarrow
Y

MERCS model

Compact representation:

f_1
f_2
f_3:
f_3

2. OBJECTIVES

3. MERCS MODEL

4. PREDICTION PROBLEM

5. ATTRIBUTE IMPORTANCE

6. CHAINING

7. RESULTS

8. CONCLUSIONS