Representation And The Price Prediction Of Elevator Cost Data With Complex Features From Multiple Sources

Elevator manufacturing enterprises are currently facing two situations,firstly,the diversification of elevator demand,the complicated calculation of elevator cost under the customization and personalized demand,and secondly,the low utilization value of multisource data of elevator business,which is difficult to transform.Enterprises need to achieve fast cost calculation and elevator quotation to win customers in response to different customers' personalized demands.This thesis takes elevator service order history data sets as the research object,proposes the data representation method of hierarchical clustering based on the knowledge system of knowledge graph,realizes the transformation of raw business data to the data set used for machine learning,completes data improvement to enhance data model performance based on data equalization and threshold-based feature selection method,and completes fast model prediction of raw input based on dynamic rule-based mapping method.Finally,based on the combination of knowledge graph technology architecture and machine learning framework platform,the elevator cost prediction system is designed and implemented in development environments such as Python and Pycharm,and the core processes of this system are tested using test datasets of elevator costs.The specific work and contributions of this thesis are as follows.(1)The knowledge system based on the knowledge graph completes the extraction of entity attributes related to elevator cost and complete the construction of the initial original data set of elevator cost;based on the knowledge technology system of knowledge graph,a data representation method based on hierarchical clustering is proposed to represent the original data into data that can be recognized and learned by machine learning algorithms.(2)Based on the Tensorflow machine learning framework,a data model of elevator cost data is constructed using DNN neural network algorithm,and the data is optimized by combining data equalization and threshold feature selection method based on feature contribution degree,thus completing the performance improvement of the data model;in the prediction stage,a dynamic rule-based data mapping method is proposed to realize the transformation of the original data to the accurate data input to the model.(3)The architecture technology system based on knowledge graph,combined with integrated environment of Python and Pycharm,and Django-Web framework combined with the above technical process,designed the architecture of elevator cost prediction system,and realiszed the elevator cost estimation system.Data decision making based on enterprise production and service-related data is currently a hot topic of research in the Internet and industry,etc.In this thesis,data attributes are constructed through data in service departments in elevator enterprises,and data knowledge representation as well as data modeling are studied and applied to an intelligent service data prediction and decision making system.Through Python data processing library and Tensorflow machine learning platform,data representation and data modeling of the original elevator dataset are carried out,and the elevator cost system construction is completed in combination with Django-Web framework,which can realize the management,pre-processing,training,viewing and other visualization operations of data,models and predictions.The system's service and interaction methods and prediction accuracy can greatly meet the needs of current users.