Journal: Heliyon
Article Title: A new method applied for explaining the landing patterns: Interpretability analysis of machine learning
doi: 10.1016/j.heliyon.2024.e26052
Figure Lengend Snippet: Overview of the proposed workflow for data collection, classification, and explanation in automated landing pattern recognition. A The single-leg landing movements of the subjects before and after the fatigue intervention were collected, and the three-dimensional kinematics and kinetics data of the landing leg during the landing phase were used as the input data of the model. B The three-dimensional kinematics and kinetics data as input signals to explore the recognizability of the two class landing patterns by three classical classification and recognition algorithm models and ZeroR classifier. C The ANN with the best performance in classification and recognition accuracy between classes was used as the forward propagation classifier to compute the input signals, and the output signals of ANN were used as the input of LRP to calculate the RS that can explain the difference of landing patterns through backward propagation. D The application of 1-SPM to evaluate the LRP results from a statistical perspective. E The results of these two aspects were evaluated and discussed together.
Article Snippet: For the implementation of SPM, the open-source MATLAB script (paired-samples T-test) of One-Dimensional SPM (SPM 1D) was employed to test the statistical differences, and the significance threshold was set as 0.05 [ , ].
Techniques: