Table 3.

Evaluation benchmark, real faults: the Fault Type can be real (R) or artificial (A); Source shows the dataset of origin; SO Post # / Subject shows ID from the dataset of origin; the models are divided into two groups of classification (C) or regression (R) task

Fault	Id	SO Post #	Source	Task	Faults
Type		/Subject
R	D1	31880720	DeepFD	C	Wrong activation function (7)
R	D2	41600519	DeepFD	C	Wrong optimiser | Wrong batch size
					Wrong number of epochs
R	D3	45442843	DeepFD	C	Wrong optimiser | Wrong loss function
					Wrong batch size | Wrong activation function (0,1)
					Wrong number of epochs
R	D4	48385830	DeepFD	C	Wrong activation function (0,1)
					Wrong loss function | Wrong learning rate
R	D5	48594888	DeepFD	C	Wrong number of epochs | Wrong batch size
R	D6	50306988	DeepFD	C	Wrong learning rate | Wrong number of epochs
					Wrong loss function | Wrong activation function (1)
R	D7	51181393	DeepFD	R	Wrong learning rate
R	D8	56380303	DeepFD	C	Wrong optimiser | Wrong learning rate
R	D9	59325381	DeepFD	C	Wrong data preprocessing
					Wrong activation function (5,6) | Wrong batch size
R	D10	024	Defect4ML	R	Wrong optmiser | Wrong number of epochs
					Missing validation set
R	D11	068	Defect4ML	C	Wrong activation function (7)
R	D12	098	Defect4ML	C	Wrong data preprocessing
R	D13	099	Defect4ML	C	Missing layer | Wrong number of neurons (0)
					Wrong activation function (1)
R	D14	48221692	DeepLocalize	R	Wrong activation function (1)
R	D15	50079585	DeepLocalize	C	Wrong number of neurons (13) | Wrong loss function
					Wrong activation function (14)
					Wrong data preprocessing
R	D16	kerasma	DeepLocalize	C	Wrong number of neurons (1)