Meta-Feature Based Data Mining Service Selection and Recommendation Using Machine Learning Models

Quality of Service (QoS) based web service selection has been studied in the service computing community for some time. However, characteristics of the input dataset are not usually considered in the selection process, even though they might have an impact on the QoS values of the service. To address this issue, we propose a QoS-based service selection process that considers the impact of dataset features and we focus on data mining services because their QoS values could be highly dependent on dataset features. We have used a meta-learning algorithm to incorporate dataset features in the selection process and studied the use of different machine learning algorithms (both classification models and regression models) as meta-learners in recommending data mining services for the given dataset. We have also investigated the impact of the number of dataset features on the performance of the meta-learners. Out of the five classification models examined here, Support Vector Machine (SVM) showed the best results in predicting the recommended service for the given dataset with an accuracy of 78%. When it comes to regression models, Multilayer Perceptron (MLP) was the best regressor. We recommend considering only simple meta-features that can be collected for most datasets, as those proved to be sufficient to achieve good prediction accuracy.