Recently, air quality issues have attracted much more attention. This paper aims to find an effective way to analyse the buildings' effects on the air quality variation between indoor and outdoor. To do so, we treat the building as a dynamic filter system by regarding the outdoor PM_2.5, the indoor PM_2.5, and the building as the input, the output, and a response system, respectively. To analyze the filtering effect produced by buildings, the statistical distribution of the indoor PM_2.5 per hour is studied, and the interrelationship between the indoor and the outdoor PM_2.5 is explored in time domain. Some interesting physical laws are discovered by using the collected data. First, the indoor PM_2.5 per hour follows Gaussian distribution in most cases. Second, the indoor PM_2.5 has a positive correlation with the corresponding outdoor one. Third, a linear regression model with high accuracy on analyzing the indoor PM_2.5 is presented, which indicates that the indoor PM_2.5 consists of two parts-one comes from the outdoor PM_2.5 penetrating into the building and the other comes from the indoor environment. Fourth, by applying different system identification methods, it is found that the B-J model is the best one in characterizing the memory effects of the building for both long time and short time scales. Particularly, for the long time memory effect, the indoor PM_2.5 average memory duration (AMD) is about 2 h, and the indoor PM_2.5 AMD to the outdoor PM_2.5 is about 7 h, while for the short time memory effect, the indoor PM_2.5 AMD is also about 2 h but that to the outdoor PM_2.5 is about 5 h. Additionally, the continuance of outdoor PM_2.5 has much greater effect on the indoor one than its concentration.