Skip to main content

Table 8 Accuracy (%) of each classifier using the linguistic features in gender-controlled experiments

From: Evaluation of linguistic and prosodic features for detection of Alzheimer’s disease in Turkish conversational speech

  

Male

 

Female

 
  

( p = 18, c = 31)

 

( p = 10, c = 20)

 

ID

Features

SVM

Bayes

CTree

NN

p-value

SVM

Bayes

CTree

NN

p-value

1.1

Question ratio

47.1

50.0

69.0

60.8

0.007

70.7

68.0

58.7

61.0

0.022

1.2

Filler ratio

43.7

40.2

37.1

39.4

1.000

66.7

51.0

63.7

69.0

0.031

1.3

Incomplete sentence ratio

46.1

43.9

54.5

52.9

0.295

53.7

47.7

47.7

52.0

0.379

2.1

Verb freq.

51.4

55.5

45.5

51.0

0.242

46.7

54.0

52.0

48.0

0.335

2.2

Noun freq.

48.6

46.7

55.3

64.7

0.031

80.7

70.7

76.0

69.7

0.001

2.3

Pronoun freq.

59.2

59.7

63.6

52.8

0.016

59.0

71.5

65.5

50.0

0.012

2.4

Adverb freq.

47.1

46.7

50.8

48.0

0.458

57.0

45.0

44.7

41.7

0.183

2.5

Adjective freq.

53.7

52.2

54.7

65.3

0.030

36.0

42.0

44.0

50.7

0.473

2.6

Particle freq.

65.3

66.9

66.7

59.2

0.010

39.0

38.0

41.7

41.0

1.000

2.7

Conjunction freq.

45.3

65.3

50.8

49.2

0.021

53.0

71.0

64.7

69.7

0.010

2.8

Pronoun-to-noun ratio

65.0

61.1

39.4

39.4

0.032

67.0

72.0

65.0

69.5

0.014

3

Unintelligible word ratio

66.7

66.1

58.0

62.2

0.022

59.7

58.0

57.0

72.0

0.009

4.1

Standardized word entropy

57.1

60.8

42.2

46.7

0.092

56.0

53.7

52.7

58.7

0.148

4.2

Suffix ratio

56.9

48.6

46.1

42.4

0.196

69.0

73.0

66.7

58.7

0.008

4.3

Number ratio

65.1

65.3

59.4

55.3

0.031

51.7

58.7

60.0

49.0

0.120

4.4

Brunet’s index

41.7

52.9

51.6

62.4

0.061

48.0

51.0

42.0

41.7

0.459

4.5

Honore’s statistic

47.1

53.3

55.1

52.9

0.249

39.7

40.7

39.0

41.0

1.000

4.6

Type-token ratio

64.1

62.0

59.0

51.6

0.041

51.7

46.7

54.0

51.7

0.308

  1. The number of patients (p) and control subjects (c) in each group is given in parenthesis. Classifiers with significant results are denoted in italics. Features with p values less than 0.05 are considered significant and are shown in bold. p value of the best classifier is reported for each feature.