The Research Of Image Classification Based On Improved KNN Algorithm

Machine learning is an important method and means in current big data processing,it is a hotspot of current scholars to make use of machine learning to process image classification.Nowadays many mature and effective machine learning algorithms have been spawned in image classification processing technology.Among them,the K-nearest neighbor algorithm(KNN)is simple,intuitive,and theoretically mature,which is one of the simplest machine learning methods,however,there are also shortcomings such as insufficient classification accuracy and low calculation efficiency when processing image classification.Aiming to improve the classification accuracy and operation efficiency of the image classification algorithm,we put forward three image classification algorithms based on the improved KNN algorithm.First,an improved KNN algorithm based on K-value selection strategy is proposed.Two classification decision rules have been added to the improved algorithm to achieve the effect of improving classification accuracy.Secondly,a KNN algorithm based on K-means clustering model is proposed.Before the classification processing,the training samples are optimized,and then the samples are clustered into different subclasses by K-means clustering method.Each subclass uses the cluster center as a new sample point to construct a new data set to achieve the effect of compressing the sample size,which reduces the amount of calculation and improves the calculation efficiency.Finally,an improved KNN algorithm based on parallel computing model is proposed.The algorithm builds a parallel computing mode,allows the entire classification algorithm to be distributed in parallel,and greatly improves the calculation speed while maintaining or improving the classification accuracy.