1 Introduction
2 State of the art
3 Methodology for predicting MTF durations
3.1 Data comprehension
3.2 Data preparation
3.3 Modelling
3.4 Evaluation and interpretation
4 Experimental evaluation
4.1 Model functionality and practical applicability
4.1.1 Model behavior with small amounts of data
Para-meter | Configurations | Description |
---|---|---|
n | 2, …, 10 | Number of observations |
α | 0.01 | Very low probability of error |
0.05 | Standard probability of error | |
0.10 | High probability of error | |
0.20 | Very high probability of error | |
σ | 2 | Small scattering |
4 | Medium scattering | |
8 | Large scattering |
N | Significance level α | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
0.01 | 0.05 | 0.10 | 0.20 | |||||||||
j | k | αk–αj | j | k | αk–αj | j | k | αk–αj | j | k | αk–αj | |
2 | 1 | 2 | 0.5000 | 1 | 2 | 0.5000 | 1 | 2 | 0.5000 | 1 | 2 | 0.5000 |
3 | 1 | 3 | 0.7500 | 1 | 3 | 0.7500 | 1 | 3 | 0.7500 | 1 | 3 | 0.7500 |
4 | 1 | 4 | 0.8750 | 1 | 4 | 0.8750 | 1 | 4 | 0.8750 | 1 | 4 | 0.8750 |
5 | 1 | 5 | 0.9375 | 1 | 5 | 0.9375 | 1 | 5 | 0.9375 | 1 | 5 | 0.9375 |
6 | 1 | 6 | 0.9688 | 1 | 6 | 0.9688 | 1 | 6 | 0.9688 | 1 | 6 | 0.9688 |
7 | 1 | 7 | 0.9844 | 1 | 7 | 0.9844 | 1 | 7 | 0.9844 | 2 | 6 | 0.8750 |
8 | 1 | 8 | 0.9922 | 1 | 8 | 0.9922 | 2 | 7 | 0.9297 | 2 | 7 | 0.9297 |
9 | 1 | 9 | 0.9961 | 2 | 8 | 0.9609 | 2 | 8 | 0.9609 | 3 | 7 | 0.8203 |
10 | 1 | 10 | 0.9980 | 2 | 9 | 0.9785 | 2 | 9 | 0.9785 | 3 | 8 | 0.8906 |
\(n_{\hbox{min} } = 8 + l - 1\) | \(n_{\hbox{min} } = 6 + l - 1\) | \(n_{\hbox{min} } = 5 + l - 1\) | \(n_{\hbox{min} } = 4 + l - 1\) |
4.1.2 Interaction effort with different data quality
Parameter/distribution | Configurations | Description |
---|---|---|
Normal (rounded up, with seed) | σ = 2 | Small scattering |
σ = 4 | Medium scattering | |
σ = 8 | Large scattering | |
Log-normal | μ = 0 | Non-normal distribution |
σ = 0.5 | ||
Weibull | Scale parameter = 128 | Investigation without failure class |
Form parameter = 2 | ||
α | 0.05 | Standard probability of error |
n | 100 | Number of observations |
Data sets | 2500 | Number of examinations per distribution |