The Research On KNN Classification Algorithm And Its Applications In Poison Diagnose Process

Data classification technology is a kind of powerful analysis tool. It aims at generating a classification function or a classification model, shines upon the data item in database into a certain given classification by this model. Existing data classification algorithms may be mainly divided into two kinds: Active learning methods and lazy learning methods. In those lazy learning algorithms most extensively used is nearest neighbor classification (NN) algorithm. Because lazy method uses many different local linear functions to form implicitly the overall situation for that goal, it has more rich assumption space than positive method. Therefore, research on lazy method and its application is a very important field.Firstly, this paper analyzes the theoretical base and realization method of K nearest neighboring (kNN) algorithm. KNN algorithm is a kind of efficiently sum up inference method. Secondly, the paper analyzes the relating features of kNN algorithm, including such problems as complex degree, classification accuracy and stocking expense.Focusing on existing problems in data classification of nearest neighbor algorithm, this paper puts forward a kind of pre-clustering weighted kNN classification algorithm model. Namely, through carrying out pretreatment to training data set, the features of which is analyzed, training data set is clustering processed and then a classification model is established. Experiments prove that new algorithm can not only efficiently reduce original kNN algorithm's process expense in the calculation procedure, but determine automatically the best k value, and accordingly, it can achieve higher accuracy than classical kNN algorithm.To satisfy the needs of poisoning classification system, the P-trees kNN classification algorithm is put forward on the basis of P-tree data structure algorithm. It means, composing a "symptom weighted vector table "according to different clinical symptoms corresponding different weight value vectors of poison forms, which is used to be property value of data set and create P tree structure of symptom weighted vector table, meanwhile, selecting HOBBit distance as distance measurement and utilizing P-trees kNN classification algorithm to carry out poison classification. The...