Table 2 Overview of components used in the classification pipeline. (a) classification algorithms, (b) computed features, and (c) preprocessing steps.
Classifier name | Mathematical background | # of snippets |
|---|---|---|
(a) Classification algorithm | ||
CatBoost | Gradient boosting on decision trees24 | 8416 |
H2O GBM | Gradient boosting machine25 | 8416 |
Decision Tree | Decision tree classifier | 8416 |
Extra trees | Extremely randomized tree classifier | 8416 |
KNN | k-nearest neighbors | 8416 |
SVM | Support vector machines (supervised learning) | 8416 |
Random Forest | Random forest meta estimator | 8416 |
Naive Bayes | Gaussian Naive Bayes | 8416 |
Neural networks (MLP) | Log-loss function LBFGS or stochastic gradient descent26 | 8416 |
Light GBM | Gradient Boosting Decision Tree27 | 8416 |
XGBoost | Distributed gradient boosting28 | 8416 |
Feature type | Description | Â |
|---|---|---|
(b) Summary of extracted features | ||
FFT features | Frequency components of sensor data | Â |
RMS features | Measure of signal energy | Â |
Wavelet features | Time-frequency decomposition | Â |
Jerk RMS | Measure of motion smoothness | Â |
Signal magnitude area (SMA) | Statistical measure of the magnitude | Â |
Correlation (XY, XZ, YZ) | Measure of relationship between acceleration axes | Â |
Preprocessing step | Description | Â |
|---|---|---|
(c) Summary of preprocessing methods | ||
Rolling mean smoothing | Moving average within a window29 | Â |
Rectification | Absolute values of signal30 | Â |
Savitzky–Golay filter | Polynomial smoothing31 |  |
Butterworth low-pass filter | Lowpass filter 10 Hz31 | Â |
Z-score normalization | Mean of 0 and standard deviation of 131 | Â |
