Classification of information models in BIM using artificial intelligence algorithms

Abstract

Introduction. The regulatory framework of building information modelling is in the process of proactive development. The development of a construction information classifier is an important step towards effective transition to digital construction. The classifier can serve as the basis for a large number of scenarios, starting from the simplest model navigation and ending with various practically valuable results obtained in the form of project budgets, statements of work amounts, and materials. In practice, classification takes a long time and requires new approaches to process automation. An innovative solution to this problem is artificial intelligence algorithms, which are a forecasting tool employing an automatic method used to enter code into an information model using processed source data and pre-trained AI models. Materials and methods. The material to be studied is the data prepared for a training set based on digital information mo­dels of civil and industrial facilities. Results. Russian and foreign classifiers of construction information were studied; machine learning models were consi­dered; a training set was made and processed using digital information models of civil and industrial facilities, and classification models were evaluated using the processed data. The highest quality classification model was selected using the criteria of preprocessing velocity, training/retraining time and the F1 score. Conclusions. A random forest machine learning model can be used as the main artificial intelligence algorithm to classify construction information. This solution will accelerate the classification process due to the automatic code entry into the model and increase the efficiency of work processes.