Time-Aware Service Ranking Prediction in the Internet of Things Environment

. 2017 Apr 27;17(5):974. doi: 10.3390/s17050974

Algorithm 1 Algorithm for time series forecasting

Input: QoS data D
Output: Predicted QoS values

{q_{t}^{i j} | t = n, n + 1, . . ., n + m}

1:
Analyze the QoS data D. Services pairs are selected out based on the pairwise comparison model, the comparison is obtained for each pair ${(s_{i}, s_{j}) | q_{t}^{i j} = \frac{1}{2} + \frac{w_{i} - w_{j}}{2 (w_{i} + w_{j})}, t = 1, 2, . . ., n}$
2:
for Each service pairs do
3:
if the p-value of LB-test $p v_{l b} < α$ then
4:
This series has serial dependency
5:
if the p-value of the ADF test $p v_{a d f} < α$ then
6:
${q_{t}^{i j}} \leftarrow diff ({q_{t}^{i j}})$
7:
end if
8:
Identify the models for QoS series
9:
Estimate the parameters of the identified models
10:
Check the significance of all candidate models, remove the non-significance models from the candidate models
11:
Select the best model from significance models as the fitted model
12:
Obtain the predicted QoS comparison values ${q_{t}^{i j} | t = n, n + 1, . . ., n + m}$ by time series forecasting
13:
else
14:
Forecast the future QoS comparison values by (8)
15:
end if
16:
end for