Home > Data Lab > Data Set
  • O2O_Coupon_Usage_Prediction

    Providers : Ant Financial Services

    Posted : 2017.08.03

    #Participants : 117

Data Set Description

Document (You can download after you login)

Format

ccf_offline_stage1_test_revised.csv

.csv (3MB)

ccf_offline_stage1_train.csv

.csv (68MB)

ccf_online_stage1_train.csv

.csv (460MB)

sample_submission.csv

.csv (364B)

Coupon Usage Prediction in Online2Offline Retail

With the popularity of smart phone, the‘O2O market(Online to Offline)’become extremely attractive. According to recent report, there are at least 10 startups valued at more than 1 billion Yuan. The O2O marketing is typically based on a large amount of users’ behavior and location information recorded by various APPs, thus admits an ideal application of big data and machine learning.
In this challenge, we consider the coupon promotion conducted by retail stores, of which the core is customer targeting via personal taste. People often like coupons issued by their favorite stores and feel disturbed while receiving non-related ads. Motivated by this observation, the task is to predict the probability of coupon usage in 15 days, based on their online and offline shopping data in the past. Specially, the training set are collected from Jan. 1st 2016 to June 30th 2016 while the test set from July. 1st 2016 to July 15th 2016. For privacy and commmercial consideration, data are anonymized and sampled biasely.

Evaluation metric: As coupons from different merchants are of varying popularity, we evaluate the average of the area under roc curve (AUC) of each coupon.

Table 1: Offline Transaction Record: train_offline_stage2

Field

Description

User_id

User ID

Merchant_id

Merchant ID

Coupon_id

Coupon ID`null’ for purchase without coupon. While Coupon_id is   null, Discount_rate and Date_received are not applicable

Discount_rate

Discount_rateA real number x \in [0,1] denotes the discount rate,   while `x:y’represents the discount is CNY `y’Yuan only if the transaction   costs at least CNY `x’ Yuan.

Distance

x\in[0,10],   which denotes the distance between the shop and the nearest location that   the  user often presents is x*500m. ‘null’   means not applicable0 for less   than 500mand 10 for father than 5000m

Date_received

Date received   the coupon

Date

Date the transaction   occurred. If Data=null & Coupon_id != nullthe user gets a coupon without usage, admitting a   negative sample. If Date!=null & Coupon_id = nullit is a transaction without coupon, while Date!=null   & Coupon_id != null denotes a transaction with coupon, representing a   positive sample.

Table 2: Online Transaction Record: train_online_stage2

Field

Description

User_id

User ID

Merchant_id

Merchant ID

Action

0 for click 1 for purchase2 for getting coupon

Coupon_id

Coupon ID.   `Fixed’ denotes a fixed and discounted price transaction.

Discount_rate

Discount_rateA real number x \in [0,1] denotes the discount rate,   while `x:y’represents the discount is CNY `y’Yuan only if the transaction   costs at least CNY `x’ Yuan. `Fixed’ denotes a fixed and discounted price   transaction.

Date_received

Date received   the coupon.

Date

Date the transaction   occurred. If Data=null & Coupon_id != nullthe user gets a coupon without usage. If Date!=null   & Coupon_id = nullit is a transaction   without coupon, while Date!=null & Coupon_id != null denotes a transaction   with coupon.

Table 3Offline Coupon Usage to Predict  prediction_stage2

Field

Description

User_id

User ID

Merchant_id

Merchant ID

Coupon_id

Coupon ID.

Discount_rate

As description   in Table 1.

Distance

As description   in Table 1.

Date_received

Date received   the coupon.