Linear pooling of machine learning techniques in Multi-Agents Systems (MAS)

No Thumbnail Available
Kumar, Arun
Issue Date
MSc in Data Analytics
Dublin Business School
Items in eSource are protected by copyright. Previously published items are made available in accordance with the copyright policy of the publisher/copyright holder.
A global solution derived from a weighted linear way improves the prediction with desirable results in multi-agent systems (MAS). In this paper, we provide two novel contributions. First, we propose a linear pooling approach to merge data from multiple agents which outperforms to standard single Machine Learning (ML) algorithms. This approach can be applied for common agents’ target variable with different agents’ descriptive variables. Second, we also examine the proposed method versus various ML techniques with different splits and compared them in terms of a standard measure of performance i.e. root mean square error for each agent with global RMSE and noticed robust results. The proposed approach is implemented into a real-world dataset (superstore) for two agents and compared to the standard ML techniques. The results show that our approach significantly outperforms the available ML methods.