**5. Discussion**

Context mining of sedentary behaviour and visualization of individual patterns may promote self-awareness to reduce it. In this regard, technology and hand-held smart devices can play a significant role. Whether a person spends much time in sedentary activities is somewhat dependent on the age group, health status, environmental conditions and life roles [12]. Our research goal is to promote self-awareness to reduce sedentary behaviours. Our approach identifies sedentary behaviour based on daily, weekly and monthly patterns. This information can be used to intervene in sedentary behaviour across the working hours, as well as leisure time. Furthermore, it can be used to predict the subject alarming condition while being sedentary in the future. In order to provide rationales for our study, we provide the comparison of one subjects' two-week comparison of recognized sedentary behaviour. The visualization of recognized contexts is presented in Figure 15.

**Figure 15.** Two-week comparison of the mining contexts.

In Figure 15, the inner circle presents the first week recognized context in percentage, while the outer circle presents the second week contextual information. We can observe quite subtle differences in the recognized sedentary behaviour. For instance, after knowing the sedentary routines, the user took more short breaks during prolonged sitting. In Figure 15, we can observe a 50% increase in short breaks, and sedentary time is reduced from 71% to 66%. The participant also reported that after knowing his/her sedentary behaviour patterns, he/she started making small changes in his/her daily routines. For instance, he/she preferred to use stairs instead of elevators and took short breaks during prolonged sitting.

In Figure 16, we also present the accumulated results of active or still contexts in terms of hours to get abstract information about the sedentary behaviour.

**Figure 16.** Time spent active or still during two weeks.

It is obvious from Figure 16 that the participants are becoming more active after knowing the sedentary patterns. Several issues with the study approach were noted throughout. In particular, our trained classifier module classifies the context "sedentary-context unknown" in certain situations where the user is working on a PC and listening to music in the background or using a PC in public places. However, we consider both situations as a sedentary context. It can be seen in Figure 17 that we presented the 57-min contextual data while participants were working on a PC and listening to music. On the other hand, the subject may not have carried his/her smartphone all the time, which may introduce errors in context recognition.

We also collected the participant feedback to find out more about the UX (i.e., user experience). The participants commented that the information they received about the sedentary behaviour is more helpful than they had expected. They found themselves checking the progress and daily status of sedentary behaviour frequently and adjusting their activity accordingly.

**Figure 17.** Misrecognized context while working on a PC.
