Research Article
Weighted Classification of Machine Learning to Recognize Human Activities
Table 1
The different characteristics of ML and deep learning as regards the HAR system.
| Process | ML | Deep learning |
| Preprocessing | Preprocessing is required and necessary for the performance | Preprocessing is not required; also normalization method |
| Features collection | (i) Depends on manual extraction | (i) Learns to extract the features directly | (ii) Avoids complicated activities | (ii) Temporal and spatial dependence with scaling features automatically | (iii) Considers feature selection with reduction technique | | (iv) Fails to handle variability and similarity class | |
| Learning process | (i) Considers small training data | (i) Works with large data | (ii) Less memory and computational time are needed | (ii) Works well with high complexity and computation | | (iii) Needs special hardware to perform training |
|
|