Recently, people are paying increasing attention to the air quality issues. In order to explore the effects of buildings on the change from the outdoor PM_2.5 to indoor ones, an effective model-fitting approach based on known knowledge is proposed to analyze the collected data from signals and systems perspective. With our presented approach, some interesting physical laws are discovered. It is found that the building produces a filtering effect on the outdoor PM_2.5. The statistical distribution of indoor PM_2.5 per hour follows Gaussian distribution in most cases and the indoor PM_2.5 has a positive correlation with the outdoor PM_2.5. Moreover, the indoor PM₂.g consists of two parts: one is from the outdoor PM_2.5 penetrating into the building, and the other is from the indoor background PM_2.5. It is also shown that the B-J model is the best mode in characterizing the memory effects of the building both for long-time scale and short-time scale analysis. For the long-time scale memory effect, the memory of current indoor PM₂.5 to the historical indoor PM₂.g is about 2 hours and that to the outdoor PM_2.5 is about 7 hours, while for the short-time scale memory effect, the memory of current indoor PM_2.5 to the historical indoor PM_2.5 is about 2 hours and that to the outdoor PM₂.5 is about 5-8 hours. This paper also indicates that the model-fitting approach from signals and systems perspective provides a very effective way to analyse PM_2.5 data.