Skip to main content

Table 4 Accuracy (%) of each classifier using the linguistic features in age-controlled experiments

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

  

Age between 60 and 75

 

Age between 76 and 90

 
  

( p = 15, c = 25)

 

( p = 13, c = 26)

 

ID

Features

SVM

Bayes

CTree

NN

p value

SVM

Bayes

CTree

NN

p value

1.1

Question ratio

70.0

67.5

70.0

60.0

0.011

66.7

35.0

38.5

43.6

0.037

1.2

Filler ratio

52.5

65.0

57.5

47.5

0.058

48.7

38.5

46.2

53.9

0.631

1.3

Incomplete sentence ratio

60.0

57.5

52.5

60.0

0.206

61.6

66.7

59.0

64.1

0.037

2.1

Verb freq.

37.5

62.5

47.5

50.0

0.114

69.2

61.5

53.9

59.0

0.016

2.2

Noun freq.

67.5

65.0

60.0

60.0

0.027

59.0

71.8

53.9

59.0

0.006

2.3

Pronoun freq.

72.5

72.5

65.0

47.5

0.004

69.2

66.7

56.4

59.0

0.016

2.4

Adverb freq.

52.5

62.5

52.5

52.5

0.114

59.0

41.0

38.5

43.6

0.262

2.5

Adjective freq.

57.5

47.5

45.0

42.5

0.343

51.3

43.6

51.3

51.3

0.873

2.6

Particle freq.

52.5

60.0

52.5

55.0

0.206

48.7

61.5

77.0

66.7

0.001

2.7

Conjunction freq.

62.5

62.5

65.0

52.5

0.058

53.9

56.4

56.4

46.2

0.423

2.8

Pronoun-to-noun ratio

70.0

70.0

67.5

62.5

0.011

71.8

66.7

61.5

46.2

0.006

3

Unintelligible word ratio

60.0

62.5

55.0

55.0

0.114

82.1

84.6

84.6

82.1

<0.001

4.1

Standardized word entropy

55.0

65.0

45.0

57.5

0.058

74.4

64.1

59.0

56.4

0.002

4.2

Suffix ratio

45.0

52.5

55.0

60.0

0.206

69.2

74.4

64.1

53.9

0.002

4.3

Number ratio

70.0

67.5

70.0

65.0

0.011

48.7

61.5

51.3

48.7

0.150

4.4

Brunet’s index

62.5

57.5

67.5

65.0

0.027

48.7

59.0

61.5

56.4

0.150

4.5

Honore’s statistic

40.0

45.0

65.0

47.5

0.058

43.6

61.5

61.5

71.8

0.006

4.6

Type-token ratio

47.5

45.0

42.5

45.0

1.000

79.5

79.5

64.1

76.9

<0.001

  1. 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.