QMADM-S: data-driven algorithms for enhancing cloud service decision-making through imputation

With the rapid advancement of cloud computing, users encounter difficulties in selecting the best cloud service providers (CSPs). These providers offer numerous services, evaluated based on key quality of service (QoS) attributes such as availability, latency, reliability, response time, and throughput. Multi-attribute decision-making (MADM) techniques are widely used to assess CSP performance. However, in real-world applications, unavailable QoS values often result in incomplete decision matrices, potentially affecting the accuracy of CSP rankings. This study evaluates the impact of three imputation techniques, min, max, and mean, on handling unavailable performance data across different levels of unavailability in seven QoS MADM with Spearman's rank correlation coefficient (QMADM-S) algorithms to assess ranking consistency. The analysis uses the QoS for web services (QWS) dataset and sensitivity analysis to identify the most effective imputation technique. Simulation results indicate that the mean imputation technique maintains ranking stability better than the other imputation techniques.