Order Brushing in Ecommerce with R Example
What is Order Brushing?
On e-commerce platforms, some sellers want to boost their rating scores, and therefore they recruit somebody to place the fake orders in a short time.
Order schemes will cause the decreasing accuracy of eCommerce recommendation systems and inflation of the sales amount on financial reports.
Order brushing is a technique to identify the spam orders happening on which shops and who are suspected spam users.
Below I will use the dataset from shopee code league 2020 to demonstrate.
Order Brushing Definition
How to define which shops and users are suspected and the order brushing applies to?
The concentration rate = number of orders within an hour on a shop/ number of unique purchasers within an hour
If the concentration rate is ≥ 3, we recognize the shop needs the order brushing.
About Data
This dataset is from shopee code league 2020.
A glimpse of the dataset:
Volume: 222,750 observevations
Columns: orderid, shopid, userid and event_time (the order placed time)
The target result:
Columns: shopid, userid (spam users)
If there is no spam user for the shop, place 0 in the ‘userid’ column.
If there are multiple spam users, use ‘&’ to concatenate.
R Example
library(dplyr)# loading data and a glimpse of data
data <- read.csv('/Users/chenlyn/Desktop/Order Brushing/order_brush_order.csv')
head(data)
length(unique(data$shopid)) # num of unique shop is 18770# How many orders on each shopshop_order <- data %>% group_by(shopid)%>%count()# If the shop has less than 3 orders, the shop is not suspected.no_brushing_shop <- shop_order[which(shop_order[,'n'] < 3),]# Remain shop having more than 2 ordersbrushing_shop_data <- subset(data, !(data$shopid %in% no_brushing_shop$shopid))
#sort the data by shopid, userid and event_time
brushing_shop_data <- brushing_shop_data %>% arrange(shopid, userid, event_time)# To find users who placed multiple orders on the same shopshop_orders_uniuser <- brushing_shop_data %>% group_by(shopid) %>%
summarise(orders = n(), unique_user = n_distinct(userid), total_centrate_rate = orders / unique_user ) # If shops' orders are from unique user, the shop is not suspectedno_brushing_shop_v2 <- shop_orders_uniuser[which(shop_orders_uniuser$total_centrate_rate ==1), ]# Remain shop with users placed the mutiple ordersbrushing_shop_data <- subset(brushing_shop_data, !(brushing_shop_data$shopid %in% no_brushing_shop_v2$shopid))#sort the data by shopid, userid and event_timebrushing_shop_data <- brushing_shop_data %>% arrange(shopid, userid, event_time)# To find users who placed multiple orders on the same shopshop_orders_uniuser <- brushing_shop_data %>% group_by(shopid) %>%
summarise(orders = n(), unique_user = n_distinct(userid), total_centrate_rate = orders / unique_user )
# If shops' orders are from unique user, the shop is not suspectedno_brushing_shop_v2 <- shop_orders_uniuser[which(shop_orders_uniuser$total_centrate_rate ==1), ]
# Remain shop with users placed the mutiple ordersbrushing_shop_data <- subset(brushing_shop_data, !(brushing_shop_data$shopid %in% no_brushing_shop_v2$shopid))# Calculate the period between the first order and the third order placed by the same user on the same shopbrushing_shop_data <- brushing_shop_data %>% group_by(shopid, userid) %>%
mutate(next_2_event_time = lead(event_time, 2), time_diff = next_2_event_time - event_time)# If the period between the first order and the third order is less than one hour, it means that the user placed >= 3 orders on the same shop within an hour.# The concentrate rate is >= 3spam_user <- brushing_shop_data[which(brushing_shop_data$time_diff <= 3600),]
# The shops without any spam user are not suspected.no_brushing_shop_v3 <- subset(brushing_shop_data, !(brushing_shop_data$shopid %in% spam_user$shopid))
no_brushing_shop_v3 <- as.data.frame(unique(no_brushing_shop_v3$shopid))
colnames(no_brushing_shop_v3) <- 'shopid'# Make the spam user lists for shops with order brushingbrushing_shop <- spam_user %>% group_by(shopid) %>%
summarise(userid = paste(unique(userid), collapse = "&"))# Combind all shops without order brushing, and insert 0 in the 'userid' column.no_brushing_shop_total <- rbind(list(no_brushing_shop$shopid, no_brushing_shop_v2$shopid, no_brushing_shop_v3$shopid))
no_brushing_shop_total <- as.data.frame(unlist(no_brushing_shop_total))
no_brushing_shop_total$userid <- 0
colnames(no_brushing_shop_total) <- c('shopid', 'userid')# Combind all shops and exportsubmission <- rbind(brushing_shop, no_brushing_shop_total)
write.csv(submission, '/Users/chenlyn/Desktop/Order Brushing/after_competion_final.csv', row.names = FALSE)
Thanks for reading this article.
If there is any suggestion, please feel free to inform me.
