Complex Event Analysis - Report 2020/02/29

No momentum supporting factor found

Challenge supporting factors

(leakage, mse)

Work-in-progress supporting factors

(leakage, model)

(model, performance)

(leakage, performance, population)

(leakage, population)

(performance, population)

(better-than-random, leakage)

(better-than-random, leakage, performance)

(better-than-random, performance)

Complex Event Time Series Summary - REPORT

Time Period	Challenge	Momentum	WIP
Report 2020/02/29	4.21	0.00	95.79

High Level Abstraction (HLA) combined

High Level Abstraction (HLA)	Report 2020/02/29
(1) (leakage,model)	100.00
(2) (model,performance)	55.86
(3) (leakage,performance,population)	40.69
(4) (leakage,population)	35.86
(5) (performance,population)	35.17
(6) (better-than-random,leakage)	33.79
(7) (better-than-random,leakage,performance)	28.28
(8) (better-than-random,performance)	21.38
(9) (leakage,mse)	18.62

Complex Event Analysis - REPORT 2020/02/29

Back to top of page

Supporting narratives:

challenge (Read more)

If no such explanation exists, be suspicious.
Examine several performance measures: in too many projects we
see a naive definition of performance measures, such as merely
accuracy (or minimum MSE). Naive performance measures may
hide local data leakages. For example, let us assume that, in a
classification task, we have a brutal data leakage which affects only
10% of the population
High Level Abstractions:

(leakage,mse)
Inferred entity relationships (3)
(leakage,performance,population) [inferred]
(leakage,population) [inferred]
(leakage,model) [inferred]

WIP (Read more)

Here are a few practices that will help you to be aware of
and hopefully even avoid (or at least minimize) the impact of data
leakage.
Decide on a benchmark model first and assess its performance: A
benchmark model is a simple yet better-than-random prediction
model
High Level Abstractions:

(leakage,model)
(model,performance)
Inferred entity relationships (30)
(model,trade) [inferred]
(leakage,mse) [inferred]
(model,nlp,pretrained) [inferred]
(model,revenue) [inferred]
(model,ott,pretrained) [inferred]
(performance,population) [inferred]
(model,models) [inferred]
(leakage,model) [inferred]
(model,ott) [inferred]
(model,pretrained,syntax) [inferred]
(leakage,performance,population) [inferred]
(model,ton) [inferred]
(model,pretrained,trade) [inferred]
(model,reading_test) [inferred]
(performance,pretraining) [inferred]
(model,pretrained,ton) [inferred]
(model,models,pretrained) [inferred]
(model,pretrained,pretraining) [inferred]
(model,pretrained,quanta_magazine) [inferred]
(model,pretrained,train) [inferred]
(model,quanta_magazine) [inferred]
(model,nlp) [inferred]
(model,performance) [inferred]
(model,pretrained,rst) [inferred]
(model,pretraining) [inferred]
(model,rst) [inferred]
(leakage,population) [inferred]
(model,product,revenue) [inferred]
(model,train) [inferred]
(model,quanta_magazine,reading_test) [inferred]

WIP (Read more)

For example, let us assume that, in a
classification task, we have a brutal data leakage which affects only
10% of the population.
High Level Abstractions:

(leakage,performance,population)
(leakage,population)
(performance,population)
Inferred entity relationships (6)
(leakage,mse) [inferred]
(leakage,performance,population) [inferred]
(performance,pretraining) [inferred]
(leakage,population) [inferred]
(performance,population) [inferred]
(leakage,model) [inferred]

WIP (Read more)

Here are a few practices that will help you to be aware of
and hopefully even avoid (or at least minimize) the impact of data
leakage.
Decide on a benchmark model first and assess its performance: A
benchmark model is a simple yet better-than-random prediction
model.
High Level Abstractions:

(better-than-random,performance)
(better-than-random,leakage)
(better-than-random,leakage,performance)
Inferred entity relationships (9)
(leakage,performance,population) [inferred]
(performance,pretraining) [inferred]
(better-than-random,leakage) [inferred]
(better-than-random,leakage,performance) [inferred]
(leakage,mse) [inferred]
(better-than-random,performance) [inferred]
(leakage,population) [inferred]
(performance,population) [inferred]
(leakage,model) [inferred]

Target rule match count: 33.0 Challenge: 0.02 Momentum: 0.00 WIP: 0.48