Neural Network-Based Approach for Predicting Trust Values Based on Non-uniform Input in Mobile Applications

Abstract

Recently, there has been much research focus on trust and reputation modelling as one of the key strategies for the formation of successful business intelligence strategies, particularly for service in mobile applications. One of the key trust modelling activities is trust prediction. During this process, the accuracy and reliability of the predicted trust values play an important role in the making of informed business decisions. Key factors to be considered at this stage are the variability and the high levels of distortion in the input series that have to be captured when predicting the trust values at a point in time in the future. In this paper, we propose a Multi-layer Feed Forward Artificial Neural Network to predict the future trust values of entities (services, agents, products etc.) for a future point in time based on data series input. We use four different non-uniform’ data input series and measure the accuracy of the predicted values under different experimental scenarios for benchmarking and comparison with existing approaches. Results indicate that the model is reliable in predicting trust values even in scenarios where there are only limited data available on training the neural network and a high level of distortion is present in the input series.