Application Research Of Machine Learning Algorithm In Identification Of Subway Prohibited Items

With the improvement of hardware and software technology,computers have be-come more and more powerful,and they have gradually become competent for some tasks that could only be done by humans.Using machines to provide convenience for human work and life has become a new development trend in modern society.Accor-ding to statistics released by authoritative institutions in China,the scale of China's A-rtificial Intelligence industry has been growing in the past few years,indicating that China has great potential in this industry,and AI is widely used in many fields,espec-ially in enterprise technology integration and solutions,key technologies R&D and application platforms,smart robots,new media and digital content,smart medical car-e,smart commodities and retail where AI are the most widely used.Metro security check is a simple but tedious task.In order to ensure absolute saf-ety and prevent any dangerous items from being carried by passengers on the subway,the security inspector not only needs to pay attention to the monitor at all times,but a-lso needs to have certain experience and quick eyesight,especially During peak hours,staff members are often tired because of long-term concentration and repetitive work.Therefore,this paper considers applying Machine Learning and Image Identification technology to provide convenience for subway security inspection.The main work of the paper is as follows:1.Collect sample data,divide the sample data into training set and test set,then based on Statistical Learning method,using the Logistic model,XGBoost model,and Random Forest model to model the sample data.Before the modeling,the grid search method is used to tune the model hyperparameters to find the best parameter combin-ation.The model is used to learn the training set samples,then Forecast test samples.The prediction accuracy rates of the three models are 53.0%,80.7%,and 75.0%.Fin-ally,the learning performance of the three models is comprehensively compared.The results show that the XGBoost model is the best among the three models.2.Use the Convolutional Neural Network built by Python to model the real data,After the hyperparameters are determined by the grid search method,the model is us-ed to learn the training sample set,and the trained model is used to predict the test sa-mple.The prediction accuracy is 86.2%.Calculate The value of an evaluation indicat-or and make an evaluation.3.Compare the scores of various indicators of XGBoost and the Convolutional Neural Network model.The comparison results show that Regardless of Precision,Recall,Macro-F1 or Micro-F1,the Convolutional Neural Network score is higher tha-n the XGBoost model,indicating that the Convolutional Neural Network is a more su-itable model,and should be given priority in practical applications.