twitter的AnomalyDetection 官網效果圖以下:git
嘗試寫了下面這個R程序:github
get_specify_df <- function(start_ts,stop_ts,category='totaluploadspeed'){ library(httr) library(rlist) library(jsonlite) base <- "http://8.8.8.8/serverdata/chartobjects?type=dcache_charge&unit=m&begin_time=" url <- paste(base,start_ts,'&end_time=',stop_ts,sep="") response <-GET(url) result <- fromJSON(content(response, as="text",encoding='utf-8')) if(1 == result$status) { # Fix here in the future... # specify_df <- list.stack(list.select(result$result,result$result$datetime,result$result$category)) specify_df <- list.stack(list.select(result$result,result$result$datetime,result$result$'totaluploadspeed')) return(specify_df) } return(NULL) } specify_df <- get_specify_df('153386640','1533870000','totaluploadspeed') library(AnomalyDetection) data(specify_df) res = AnomalyDetectionTs(specify_df, max_anoms=0.02, direction='both', plot=TRUE) res$plot
想利用Twitter開源的這個異常檢測模塊,可是遇到的問題不少,R語言自己可參考資料很少,而且目前貌似已經沒人維護了...json
因此暫時不想把過多的時間放在這上面,仍是改用Python吧...url