Skip to main content
Journal of Big Data

Table 6 Experimental design scenarios

From: Text based personality prediction from multiple social media data sources using pre-trained language model and model averaging

Scenario

System baseline

Batch size

Learning Rate

1

BERT

16

1.00E−05

2

16

3.00E−05

3

16

1.00E−05

4

16

3.00E−05

5

32

1.00E−05

6

32

3.00E−05

7

32

1.00E−05

8

32

3.00E−05

9

Roberta

16

1.00E−05

10

16

3.00E−05

11

16

1.00E−05

12

16

3.00E−05

13

32

1.00E−05

14

32

3.00E−05

15

32

1.00E−05

16

32

3.00E−05

17

XL Net

16

1.00E−05

18

16

3.00E−05

19

16

1.00E−05

20

16

3.00E−05

21

32

1.00E−05

22

32

3.00E−05

23

32

1.00E−05

24

32

3.00E−05

25

BERT + NLP Statistical Features

16

1.00E−05

26

16

3.00E−05

27

16

1.00E−05

28

16

3.00E−05

29

32

1.00E−05

30

32

3.00E−05

31

32

1.00E−05

32

32

3.00E−05

33

Roberta + NLP Statistical Features

16

1.00E−05

34

16

3.00E−05

35

16

1.00E−05

36

16

3.00E−05

37

32

1.00E−05

38

32

3.00E−05

39

32

1.00E−05

40

32

3.00E−05

41

XLNet + NLP Statistical Features

16

1.00E−05

42

16

3.00E−05

43

16

1.00E−05

44

16

3.00E−05

45

32

1.00E−05

46

32

3.00E−05

47

32

1.00E−05

48

32

3.00E−05

49

Proposed Method (Model averaging (BERT + ROBERTA + XLNet)) + NLP Statistical Features

16

1.00E−05

50

16

3.00E−05

51

16

1.00E−05

52

16

3.00E−05

53

32

1.00E−05

54

32

3.00E−05

55

32

1.00E−05

56

32

3.00E−05

Back to article page