Skip to main content

Table 6 Accuracy (%) of each classifier using the linguistic features in education-controlled experiments

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

  

High school and below

 

College and above

 
  

( p = 18, c = 24)

 

( p = 10, c = 27)

 

ID

Features

SVM

Bayes

CTree

NN

p value

SVM

Bayes

CTree

NN

p value

1.1

Question ratio

54.8

38.1

59.5

45.2

0.217

73.0

75.7

70.3

62.2

0.002

1.2

Filler ratio

40.5

35.7

35.7

38.1

1.000

62.2

59.5

70.3

62.2

0.014

1.3

Incomplete sentence ratio

52.4

54.8

57.1

59.5

0.217

56.8

59.5

64.9

67.6

0.033

2.1

Verb freq.

61.9

61.9

61.9

47.6

0.123

64.9

54.1

64.9

73.0

0.005

2.2

Noun freq.

57.1

40.5

59.5

50.0

0.217

78.4

78.4

86.5

73.0

<0.001

2.3

Pronoun freq.

76.2

71.4

61.9

54.8

0.001

75.7

73.0

64.9

51.4

0.002

2.4

Adverb freq.

38.1

35.7

50.0

40.5

1.000

70.3

81.1

59.5

62.2

<0.001

2.5

Adjective freq.

52.4

45.2

54.8

57.1

0.355

54.1

51.4

56.8

56.8

0.411

2.6

Particle freq.

54.8

47.6

61.9

50.0

0.123

43.3

62.2

54.1

62.2

0.139

2.7

Conjunction freq.

42.9

35.7

35.7

42.9

1.000

73.0

73.0

75.7

64.9

0.002

2.8

Pronoun-to-noun ratio

73.8

71.4

57.1

59.5

0.002

78.4

73.0

51.4

62.2

0.001

3

Unintelligible word ratio

64.3

57.1

57.1

57.1

0.064

78.4

78.4

62.2

67.6

0.001

4.1

Standardized word entropy

64.3

69.1

57.1

57.1

0.014

70.3

56.8

59.5

56.8

0.014

4.2

Suffix ratio

57.1

57.1

59.5

57.1

0.217

62.2

37.8

54.1

62.2

0.139

4.3

Number ratio

57.1

50.0

50.0

35.7

0.355

59.5

56.8

56.8

43.3

0.250

4.4

Brunet’s index

40.5

61.9

54.8

52.4

0.123

67.6

54.1

48.7

62.2

0.033

4.5

Honore’s statistic

40.5

52.4

45.2

54.8

0.537

56.7

56.8

54.1

48.7

0.411

4.6

Type-token ratio

59.5

64.3

45.2

50.0

0.064

67.6

37.8

59.5

56.8

0.033

  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.