ORAC

点击翻译按钮获取中文摘要

The aim of the ORAC development project is to advance the results drawn from the InnovateUK Proof of Concept research project. That research into the use of machine learning(specifically neural networks) techniques has shown that Spend360 can significantly improvethe accuracy of data classification within a spend analysis system. This enables effective costreduction and supplier management by providing detailed all-round insight into procurementspend by an organisation.Currently these systems are heavily dependent on human interaction and require considerabledomain-specific knowledge therefore accurate automation is essential. The current state-of-theart for automated data classification in such systems is only 50%-70% accurate and sorequires considerable human intervention to complete the task (this also assumes the autoclassification is capable of addressing a wide range of market sectors). Research by Spend360has shown that when these new machine learning approaches are combined with othertechniques, an accuracy of >95% could be achieved. The innovation will be based upon avariably-sized, multilayered neural network approach: specifically using a general purposeLong Short Term Memory (LSTM) network to map invoice descriptions (sequences of words)to UNSPSC codes.The primary benefit that will accrue from this new innovative approach is that it willsignificantly reduce the human involvement in the classification process and associated dataquality assurance. This will significantly reduce the time to complete the classification andimprove the accuracy. Together, this will result in cost reduction of the solution therebyenabling adoption by a range of organisations that find the current associated service costs toohigh. Furthermore, the automation will enable an organisation to undertake more detailed andvaried forms of spend analysis so trend analysis can be used to identify ways in which thespend profiles can be further optimised.