I'm Only Unhappy when it Rains: Forecasting Mobile QoS with Weather Conditions

Improving Signal-Strength Aggregation for Mobile Crowdsourcing Scenarios

“…As this paper is the first attempt to challenge existing assumptions about signal-strength aggregation, we performed the following simplifications to the problem of estimating the mathematical expectation of signal strength in an area: We considered areas with signal strength coming from only one base transceiver station (BTS). Even when there may be a time variability of signal strength in the area [ 10 , 28 ], we considered that the mathematical expectation is estimated for a static power configuration of the BTS. …”

“…We simulated the real signal strength on a fine-grained grid G over A with 5 m spacing, considering long-term attenuation due to path loss equation and medium-term variation due to shadowing modeled by a full covariance matrix [28,[54][55][56][57]. Indeed, the real signal strength in G is given by 1P − 10α log 10…”

“…In this way, some works aggregated several signal-strength samples inside the same area into a unique representative signal-strength value to predict user availability [ 9 ], measure the effect of weather conditions in the received signal strength [ 10 , 11 ], analyze network performance [ 12 ], measure the impact on signal strength of indoor-outdoor context [ 13 ] and find correlations between signal strength and other QoS indicators for mobile networks as network congestion [ 15 ], throughput [ 8 ] and TCP goodput and latency inside the same geographic area [ 1 ]. In this paper, we propose an aggregation method that better estimates the expected value of signal strength than the methods used in the papers mentioned above, especially when using measurements taken in Mobile Crowdsourcing contexts.…”

“…Another method to summarize several signal-strength samples is by choosing the median value that separates the higher half from the lower half of all the measurements [ 10 , 47 , 48 , 49 ]. The idea behind using median value to aggregate signal-strength samples is that it is not skewed so much by a small proportion of extremely large or small values, which is a common situation in this case study, because, as shown in Section 3 , signal amplitude could vary very widely among measurements.…”

“…A common methodology to summarize the received signal strength inside a specific area is to aggregate all the individual measurements into one representative value that characterizes the signal strength inside the location area [ 1 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 ].…”