Study On KNN Improved Algorithm And Its Application In Take-out Software

The mobile information age has arrived,and people's lifestyles have changed dramatically.Ordering food from a takeout App on a smartphone is one of them.Nowadays,China's takeout industry has been developing rapidly,and more and more people start and frequently use the takeout App to order food,so the takeout market is expanding rapidly.However,at present,the service of the takeout industry cannot keep up with the volume of the takeout market,and there is a phenomenon of "high-end charging and low-end service".In order to improve users' experience of ordering takeout,this paper studied the function of "expected delivery time" in the takeout App.In order to evaluate the accuracy of expected delivery time,the author assumes construct a model a evaluation model formed by learning the data of previous takeout orders,and the accuracy of the estimated delivery time for new orders can be judged by the established model.Therefore,an improved KNN algorithm is proposed in this paper.Compared with traditional KNN algorithm,this improved KNN algorithm is improved from two aspects: classification efficiency and classification effect.In order to improving the classification efficiency,DBSCAN algorithm is used to cluster the training data sets and reduce the number of samples of the training data sets by reducing the non-core samples in each cluster,thus indirectly improving the classification efficiency.In order to improve the classification effect,we use the dual voting mechanism of high weight and distance weight of the remaining representative samples to classify the voting,which makes the classification voting more reasonable.The improved KNN algorithm finally obtained the expected results in the experiment of the sample data set of the actual takeout order,and basically achieved the expected results in the experiment of the UCI public data set.