Research And Application Of Key Algorithms For Construction Cost Big Data

The arrival of the era of big data and the development of machine learning have brought new opportunities for Chinese construction cost field.Through the combination of construction cost field and big data field,we can make full use of the massive data wealth accumulated in the construction cost field for a long time,dig out the development patterns and knowledge of the construction cost industry,and provide guiding opinions for future project construction and supervision.At present,Chinese construction cost big data field is still in the initial stage of platform construction,data collection and standardization,and there are a lot of issues.Among these issues,the solution to the issue of the standardized classification of lists is of great significance to the analysis of the rationality of the cost structure of projects from multiple levels based on uniform standards,and the solution to the problem of abnormal data detection for comprehensive unit price of lists is of great significance for correct calculation of construction cost and research of construction cost trend.This thesis selects checklist normalization classification and list unit price abnormality data detection as the starting point.The specific work are as follows:(1)In the field of construction cost in China,due to historical reasons,the classification criteria of construction cost list in the original data is confusing.The traditional rule-matching-based lists normalization classification method in the field has problems such as inefficient manual summarization of rules and poor general use of rules.This thesis found the features of list data such as large quantity of proper nouns,scattered distribution of features.Based on these features,this thesis implements and compares the effects of several lists classification methods,analyzes the causes of the results and refers to the classification effect of traditional methods.After these works,this thesis proposes a multinomial Bayes-based lists standardized classification method.(2)There are many abnormal data in the original construction cost list data.For the abnormal data detection of list unit price,the detection dimension of the traditional method is relatively single,and only abnormal data with a large difference between the overall unit price and the historical distribution of historical data can be detected.However,there are still a large number of integrated unit prices in the actual situation that are in line with the overall distribution,but far higher than the price that the work in the list description should have.This kind of abnormal data is difficult to detect for traditional methods.This paper proposes a method for detecting abnormal data by using the integrated unit price as a classification label and studying its association with the list description according to the list classification method.The effectiveness of this method was verified by experiments.Based on these two methods,this thesis designs a system architecture with list normalization classification and list integrated unit price anomaly data detection capabilities.Applying this system architecture to the data standardization of the construction cost big data platform can reduce labor costs and brings convenience to data analysis.