LSTM. (a) and (b) display boxplots presenting the FAR and F1 score for various features, namely Force, Displacement, and F&D, utilizing the fixed number of hidden layers of 100, learning rate of 0.0006, and minibatch size of 128. (c) and (d) showcase boxplots of FAR and F1 score, presenting variations in the numbers of hidden layers maintaining F&D as the feature set, alongside a fixed learning rate of 0.0006, and minibatch size of 128. Finally, (e) and (f) exhibit boxplots demonstrating FAR and F1 score differences based on different learning rates, while keeping the number of hidden layers constant at 60 and utilizing F&D as the feature set. The boxplots are represented with a box having a horizontal line in the center, with horizontal lines extending from each side (referred to as “whiskers”). The box represents the interquartile range (IQR), encapsulating 50% of data points falling between the first quartile and the third quartile in the dataset. The outcomes corresponding to training and validation are distinguished by black and red colors, respectively.