Hackntu_x_Cathay_2017

• Thanks for everyone's attending!

introduction(pdf)

• 資料介紹
• 方向建議與實作
• 商業分析

dataset (for preview purpose)

profile

field	type	Description
customer_id	String	客戶 ID
birth_time	Long	出生時間戳記
gender	String	性別
contact_loc	String	居住地區
contact_code	String	居住地區郵遞區號代碼
register_loc	String	戶籍地區
register_code	String	戶籍地區郵遞區號代碼
start_time	Long	起始往來時間戳記
aum	Double	近三個月平均的總資產餘額
net_profit	Double	近一年淨收益
credit_card_flag	String	是否有信用卡產品往來
loan_flag	String	是否有貸款產品往來
deposit_flag	String	是否有存款產品往來
wealth_flag	String	是否有理財產品往來
partition_time	Long	snapshot時間戳記

sample

1,1871111837,F,D,181903,F,373933,3320438400,1217.1769900503969,528.7309192220683,Y,Y,N,N,3696969600

event

themes

• cctxn
• atm
• cti
• mybank

field	type	description
actor_type	String	值皆是"customer_id"
actor_id	String	行為者ID
action_type	String	行為類型
action_time	Long	行為發生時間戳記
object_type	String	影響物件類型
object_id	String	影響物件ID
channel_type	String	通路類型
channel_id	String	通路ID
attrs	String	補充資料 - JSON string
partition_time	Long	最近上一區段時間戳記 (可串連profile)
theme	String	主題 - cc_txn, atm, cti, mybank

sample

customer_id,17,inquire,3696970704,saving_acct,MJQXLDJMQQOYAPFBBPBS,ATM,NCZHYDSTABCUAXLYJLQZ,"{\"action\": {\"txn_amt\": 0.0, \"txn_fee_amt\": 0.0, \"trans_type\": \"trans_out\"}, \"object\": {\"target_bank_code\": \"B_bank\", \"target_acct_nbr\": \"UMOTEPKACILAPBIKOYHZ\"}, \"channel\": {\"address_zipcode\": 601401, \"machine_bank_code\": \"C_bank\"}}",3696969600,atm

example

• Easy Profiling
• Wrangling data with Apache Spark

Already installed libraries in VM

Python 2.7

sklearn, numpy, pandas, jupyter, networkx, keras, tensorflow, flask, django, xgboost

R 3.3.3

xgboost

Java 1.7

Spark 2.1

node.js

IDE

sublime, pycharm, Rstudio

other reference

Credit Card

• MCC codes

Spark

• Python+Spark 2.0+Hadoop 機器學習與大數據分析實戰

• Learning Spark

• Spark RDD minibatch for Deep learning - Welcome for refining

• R on Spark

Python

• Hands-On Machine Learning with Scikit-Learn and TensorFlow