nPlan's Experimental Research Department

Research at nPlan

Research is in the foundations of nPlan. It is currently done in nERD - nPlan’s Experimental Research Department. We operate as an independent multidisciplinary team, with our long term goals aligned with nPlan’s business objectives. We also publish our research contributing to the broader ML audience.

Mission

nERD’s mission is to advance the state of the art in project forecasting enabling robust and predictable projects. We rapidly transfer innovative technologies into nPlan products ensuring that nPlan products have a future.

Brief

Our research interests cover a wide range of topics in ML inspired by our dataset and business needs. Our dataset is composed of over 400`000 construction project schedules. These are DAGs, where each node represents an activity and edges are constraints between activities. Each node has numerical and textual features, while edges have types, weights and directions. In addition, our graphs have a temporal component representing how each project changed over time.

Since nPlan’s products are based on forecasting on project plans, our research efforts have always been focused on probabilistic modelling and uncertainty estimation. In the past we have also built in-house language models, however, with the rise of large pre-trained models we have shifted our focus on graph neural networks on temporal DAGs. Another area of active research in nERD is studying how humans interact with, trust and take action using forecasts given by AI. This includes explainable forecasting and recommendations.

Publications

On Forecasting Project Activity Durations with Neural NetworksP Zachares, V Hovhannisyan, C Ledezma, J Gante, A MoscaInternational Conference on Engineering Applications of Neural Networks, 103-114
A graph-based approach for unpacking construction sequence analysis to evaluate schedulesY Hong, H Xie, V Hovhannisyan, I BrilakisAdvanced Engineering Informatics 52, 101625
Determining Construction Method Patterns to Automate and Optimise Scheduling–A Graph-based ApproachY Hong, V Hovhannisyan, H Xie, I Brilakis2021 European Conference on Computing in Construction
How to Calibrate your Neural Network Classifier: Getting True Probabilities from a Classification ModelN Culakova, D Murphy, J Gante, C Ledezma, V Hovhannisyan, A MoscaProceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 3499–3500, 2020.