E model performance when combining ML-SA1 site 3D-ACC with the ECG signal.It
E model functionality when combining 3D-ACC with the ECG signal.It can be essential to mention that for ten out of 14 subjects we observe “Stairs-Walking” improvement after adding the ECG signal to 3D-ACC, even so, in three out of 14 circumstances adding the ECG signal does not increase the “Stairs-Walking” GLPG-3221 supplier classification. Moreover, in 1 case, the model completely distinguishes between “Stairs-Walking” by just using the 3D-ACC, leaving no space for improvement for the 3D-ACC and ECG fusion model. 6.2. Cross-Subject Cross-subject models give a extra insightful analysis, since these models missclassify activities much more generally, in comparison with subject-specific models. As depicted in Figure 7, utilizing only the 3D-ACC signal, we obtained an F1-score of 83.16 which is somewhat decrease than the model overall performance within the subject-specific setup. Following a detailed investigation in confusion matrices of your 3D-ACC educated model, we after once again recognize that the activities “stairs” and “walking” are miss-labeled. Furthermore for the described pair of activities, an additional pair is miss classified in cross-subject models, namely, “sitting” and “playing table soccer”. We when once again examine the confusion matrices connected models educated with 3D-ACC (Situation 1) signal versus the model trained with each 3D-ACC and ECG signals (Scenario 4). We observe that the ECG signal drastically aids the model recognize “Stairs-Sensors 2021, 21,17 ofWalking”, however, it will not add any worth when it comes to distinguishing the “SittingTable-Soccer” pair. Figure 10 depicts each confusion matrices connected to topic quantity 7 in the cross-subject model. The left side of Figure 10 is associated to the model overall performance when considering only 3D-ACC; note the enormous portion of “Walking” situations which are miss-classified as “Stairs”. Having said that, around the suitable side of Figure 10, it is apparent that after adding the ECG signal, the “Stairs-Walking” detection enhances noticeably.Figure 10. Comparison among confusion matrices in cross-subject models. On the left: the model efficiency when thinking of only 3D-ACC. Around the proper: the model overall performance when combining 3D-ACC with all the ECG signal.It is worth noting that for 9 out of 14 subjects, we observe “Stairs-Walking” improvement immediately after adding the ECG signal to a pure 3D-ACC model. In 3 out of 14 circumstances, adding the ECG signal yielded no significant impact; and, in two out of 14 situations, the ECG signal addition resulted inside a decline in the “Stairs-Walking” classification. 6.3. Feature Importance We’ve got shown that fusing 3D-ACC and ECG signals yielded the ideal functionality in classifying human activities in our study. Even so, which features from each signals were the most relevant to our model In this section, we present the feature significance ranking of the model that combines 3D-ACC and ECG (Situation 4) utilizing the cross-subject model, as we want to investigate the most effective attributes across multiple subjects. We calculate the feature significance utilizing the Mean Lower in Impurity (MDI) of our random forest model [59]. To aggregate the value score for each model evaluated on a single topic, we calculate the typical score for each function over each of the subjects and rank their value score. As Table 5 shows, out of leading 20 features, 16 features are associated towards the 3D-ACC signal and four of them to the ECG signal. Naturally, as 3D-ACC supplies the ideal signal with the person signal models (situation 1), we expect to find out a dominance of 3DACC capabilities inside the top-20 ranking.