Skip to main content

Table 1 Summary of datasets and best predictive model

From: Hybrid deep learning approach to improve classification of low-volume high-dimensional data

Dataset

#Samples

#Classes

#Features

Best predictive model

Best accuracy

Class frequency

WISDM

27,452

6

320

Hybrid (1-block, layer: 5)

0.793

Walk 4588

      

Upstairs 4416

      

Downstairs 4438

      

Sit 4678

      

Stand 4698

      

Jog 4634

HAR

10,299

6

561

Hybrid (5-blocks, layer: 10–12)

0.925

Walk 1722

      

Upstairs 1544

      

Downstairs 1406

      

Sit 1777

      

Stand 1906

      

Lie 1944

Amazon

1000

2

100

Hybrid (3-blocks, layer: 5)

0.797

Review > 2 500

      

Review ≤ 2 500

Yelp

1000

2

100

Hybrid (3-blocks, layer: 2)

0.776

Review > 2 500

      

Review ≤ 2 500

IMDb

748

2

100

Hybrid (3-blocks, layer: 5)

0.797

Review > 2 374

      

Review ≤ 2 374

Gene Promoter

106

2

57 × 4

Hybrid (3-blocks, layer: 1)

0.952

Promoter 53

      

Non-promoter 53

  1. For each dataset, the number of available samples, number of classes, and number of features are shown. The Gene Promoter features use a one-hot encoding of 57 base-pairs, each one of four values [a, c, g, t]. The Best Predictive Model is shown along with the number of blocks and the feature extraction layer. The Best Accuracy for this model is also shown. The class frequencies are given in the last column
Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us