Design And Implementation Of Decision Support System For Iron And Steel Enterprises Based On Machine Learning

Since its birth,artificial intelligence has changed our way of life,and it is also accelerating into the stage of changing production methods.Machine learning is the key way to make computers intelligent,and the primary driving force of its development is how to acquire knowledge quickly and accurately.At present,the decision support system for iron and steel enterprises lacks timeliness and effectiveness when analyzing massive data,and lacks a knowledge base that automatically absorbs the wisdom of managers and can be easily interacted.Therefore,this paper applies machine learning technology in it,focusing on the research on the application of intelligent prediction in the field of product quality.For the knowledge generated by intelligent prediction,speech recognition and knowledge graph are used to realize intelligent interaction.This paper has completed the functional design and implementation of the decision support system for iron and steel enterprises based on machine learning,which adopt a three-tier architecture of data processing,data analysis,and data interaction.In the data analysis tier,firstly establish product quality(composition,performance,etc.)predictions and other intelligent prediction models,use massive historical data containing all features as the training sets to test the prediction model,calculate its prediction accuracy,and then choose the optimal model and algorithm.In the data interaction tier,the bidirectional long short-term memory network speech recognition system is used to identify the user's analysis needs description.Based on the knowledge inferred from the intelligent prediction model,build a simple knowledge graph,and display the intelligent prediction results through intelligent interaction.This paper conducts experiments on intelligent prediction and intelligent interaction models based on different data sets.Intelligent prediction uses the alloy performance prediction model as a pilot to compare the accuracy of generalized linear models,random forests,XGBoost,etc.on actual data sets.Intelligent prediction selects the XGBoost model with significant advantages to optimize its learning ability.We build a speech recognition model based on bidirectional long short-term memory network,and optimize it for complex application scenarios.Under the experimental conditions of this article,the accuracy of prediction models and speech recognition reached about 85% and 90% respectively.