knn.rar_R语言_R语言knn预测_knn算法

########################### Chapter Three KNN ################################### ## Advantage of K-Nearest Neighbor: 1. simple and effective ## 2. no other requirement of sample distribution ## 3. training fast ## Disadvantage of K-Nearest Neighbor: 1. hard to comprehend how to classify ## 2. proper K ## 3. slow in classficition step install.packages("class") install.packages("gmodels") ############################# PRE-REVIEW OF DATA SET ############################### setwd("C:/Users/Shijie Wang/Desktop/R/Machine-Learning-with-R-datasets-master") # working path wbcd = read.csv("wisc_bc_data.csv", stringsAsFactors = FALSE) # import data str(wbcd) # summary of data structure wbcd = wbcd[-1] # DELETE ID IN CASE OF OVER-FIT table(wbcd$diagnosis) # DIAGNOSIS FACTOR - THE CLASSFICAITON FACTOR( BENIGN/MALIGNANCE) wbcd$diagnosis = factor(wbcd$diagnosis, levels = c("B", "M"), labels = c("Benign", "Malignant")) # FACTORIZATION OF CLASS IN ML round(prop.table(table(wbcd$diagnosis)) * 100, digits = 1) summary(wbcd[c("radius_mean", "area_mean", "smoothness_mean")]) # THREE CHARACTER # NOTE: CALCULATION OF DISTANCE: UNIT INFLUENCE SHOULD BE ELMINATED BY *NORMALIZATION* ############################### PRE-PROCESSS OF DATA ################################# normaliza = function(x){ return((x-min(x))/(max(x)-min(x))) } # NORMALIZE DATA TO ELIMATE UNIT IMPACT wbcd_n = as.data.frame(lapply(wbcd[2:31], normaliza)) # APPLY TO EVERY COLUMN summary(wbcd_n$area_mean) ############################# ESTABLISH MODEL(TRAIN DATA AND TEST DATA) wbcd_train = wbcd_n[1:469, ] wbcd_test = wbcd_n[470:569, ] # NOTE DATA SET IS ALREADY "READOM" NO NEED TO SAMPLE wbcd_train_labels = wbcd[1:469,1] wbcd_test_labels = wbcd[470:569,1] # CLASS FACTORS OF DATA TO TRAIN library(class) wbcd_test_pred = knn(train = wbcd_train, test = wbcd_test, cl = wbcd_train_labels, k = 21) # NOTE: NORMALLY, K VALUE EQUALS Squr DATA_SIZE ## PROBLEM: HOW TO DEFINE K IN A PROPER WAY ????????????????????????? ############################# EVULATE AND TEST MODEL ABILITY ########################### library(gmodels) CrossTable(x = wbcd_test_labels, y = wbcd_test_pred, prop.chisq = FALSE) # INTERPOLATE THE OUTCOME, ESPECIALLY THE FALSE NEGATIVE AND FALSE POSITIVE