Table 6 List of 80 selected feature set and finally selected 53 features
From: Machine learning based customer churn prediction in home appliance rental business
Feature group | Feature list | Selected | Format | Feature descriptions |
|---|---|---|---|---|
1. Rental fee | Monthly discount amount | o | Numeric | Monthly rate discount |
Sale price | o | Numeric | Sale price | |
Sale charge | o | Numeric | Sales fee | |
Total rental fee | o | Numeric | Total rental amount | |
1st Year Rental Fee, 2nd Year Rental Fee, 3rd Year Rental Fee, 4th Year Rental Fee, 5th Year Rental Fee | Numeric | Rental fee of 1st year (to 5th year) | ||
2. Contact history (call count, call type history, etc.) | Call Count | o | Numeric | Number of inbound calls in DATE_DIFF(CURRENT_DATE(),CONTACT_DATE, Year) < 1)) |
Call Type 004 | o | Numeric | Number of calls of type (GB, TYPE, GROUP) = (“ARS,” “Maintenance Service,” “Others”) in DATE_DIFF(CURRENT_DATE(),CONTACT_DATE, Year) < 1)) | |
Call Type 008 | o | Numeric | Number of calls of type (GB, TYPE, GROUP) = (“I/B,” “Contract related,” “Contract information”) in DATE_DIFF(CURRENT_DATE(),CONTACT_DATE, Year) < 1)) | |
Call Type 020 | o | Numeric | Number of calls of type (GB, TYPE, GROUP) = (“I/B,” “Fee collection related,” “Fee collection”) in DATE_DIFF(CURRENT_DATE(),CONTACT_DATE, Year) < 1)) | |
Call Type 048 | o | Numeric | Number of calls of type (GB, TYPE, GROUP) = (“I/B,” “contract termination Related,” “consultation on termination”) in DATE_DIFF(CURRENT_DATE(),CONTACT_DATE, Year) < 1)) | |
3. Discount (applied amounts, reason category, type, etc.) | Discount application Amount_min | o | Numeric | Minimum value of discount application amount |
Discount application Amount_max | o | Numeric | Maximum value of discount application amount | |
Discount application Amount_avg | Numeric | Mean value of discount application amount | ||
Discount reason code(40) Count | Numeric | Number of times “Reason for discount” is code “promotion” | ||
Discount reason code(77) Count | Numeric | Number of times “Reason for discount” is code “commitment discount” | ||
Discount reason code(90) Count | Numeric | Number of times “Reason for discount” is code “settlement of churn receipt balance” | ||
Discount reason code(10) count | Numeric | Number of times “Reason for discount” is code “customer dissatisfaction” | ||
Discount reason code(88) count | Numeric | Number of times “Reason for discount” is code “churn defense discount” | ||
Discount reason code_40 | o | Yes/No | Whether or not the code value “promotion” of “Reason for discount” exists | |
Discount reason code_77 | Yes/No | Whether or not the code value “commitment discount” of “Reason for discount” exists | ||
Discount reason code_90 | o | Yes/No | Whether or not the code value “settlement of churn receipt balance” of “Reason for discount” exists | |
Discount reason code_10 | o | Yes/No | Whether or not the code value “customer dissatisfaction” of “Reason for discount” exists | |
Discount reason code_88 | o | Yes/No | Whether or not the code value “churn defense discount” of “Reason for discount” exists | |
Discount type ST count | Numeric | Number of times “discount type” is code “ST (normal discount)” | ||
Discount type SP count | Numeric | Number of times “discount type” is code “SP (special discount)” | ||
Discount TYPE SC Count | Numeric | Number of times “discount type” is code “SC (commitment discount)” | ||
Discount Type ST | o | Yes/No | Whether or not the code value “ST (normal discount)” of “Discount Type” exists | |
Discount Type SP | o | Yes/No | Whether or not the code value “SP (special discount)” of “Discount Type” exists | |
Discount Type SC | o | Yes/No | Whether or not the code value “SC (commitment discount)” of “Discount Type” exists | |
Discount date diff from start | o | Numeric | The number of days between the current date and the maximum discount start date | |
4. Commitment (sequence, amount, date, etc.) | Commitment max amounts | Numeric | The maximum value of commitment amounts | |
Commitment max sequences | o | Numeric | The maximum value of commitment sequences | |
Commitment date diff from creation to start | o | Numeric | The number of days between the minimum commitment start date and the minimum creation date | |
Commitment date diff from creation | Numeric | The number of days between the current date and the minimum creation date | ||
Commitment date diff from start | Numeric | The number of days between the current date and the minimum commitment start date | ||
5. Usage days | Usage days | o | Numeric | The number of days between the current date and contract creation date |
6. Visit history (scheduling style.) | Non period visit | o | Numeric | Number of irregular visits |
Period visit | o | Numeric | Number of regular visits | |
Visit change | Numeric | Number of rescheduled visits | ||
Not visit | o | Numeric | Number of cancelled visits | |
Weekend visit | o | Numeric | Number of weekend visits | |
7. Manage age (Rental door-to-door service manager age) | Manager Age_min | Numeric | Minimum age of manager the customer encountered | |
Manager Age_max | Numeric | Maximum age of manager the customer encountered | ||
Manager Age_avg | o | Numeric | Average age of manager the customer encountered | |
Manager Age_range | Numeric | Range in age of manager the customer encountered | ||
Manager Age_std | o | Numeric | Standard deviation in age of manager the customer encountered | |
Last Manager Age | o | Numeric | The age of the last manager the customer encountered | |
8. Repair history | Normal repaiR | o | Numeric | Number of “general type” repairs received |
Shibang repair | o | Numeric | Number of “shibang type” repairs received | |
Heavy repair | o | Numeric | Number of “heavy type” repairs received | |
9. Rental model category | Rental model category | o | Categoric | Categories such as “GUBUN1_ SLIM(16’)_cold and hot,” “GUBUN1_GTOP,” “GUBUN1_SLIM(16’)_cold and purification,” “GUBUN1_CIEL,” “GUBUN1_SLIM(16’)_purification,” “GUBUN1_ECO(RO water tank),” “GUBUN1_STAND,” “GUBUN1_JINTEK(RO water tank)” |
10. Sales type | Sales type | Categoric | Sales channels such as home appliance store, personal sales, electronic land, commercial specialty store, hi-mart, hi-care, home shopping, etc | |
11. Rental model color | Rental Model Color | o | Categoric | Colors such as “PROGRAM.COLOR_luxury white,” “PROGRAM.COLOR_white,” “PROGRAM.COLOR_silver,” “PROGRAM.COLOR_black,” “PROGRAM.COLOR_red,” “PROGRAM.COLOR_luxury black,” “PROGRAM.COLOR_shine,” “PROGRAM.COLOR_gray” |
12. Rental model function | Rental model function | o | Categoric | Functions such as “PROGRAM.FUNCTION_hot/cold/purification,” “PROGRAM.FUNCTION_purification,” “PROGRAM.FUNCTION_cold/purification,” “PROGRAM.FUNCTION_hot/purification,” “PROGRAM.FUNCTION_hot/cold/week cooling” |
13. Delivery goods type | Order type (e.g. DELIVERY_GOODS_TYPE_TypeA) | o | Categoric (15 values) | Delivery type such as re-rental after 3 years of use (additional discount), home appliance package, re-rental for 5 years (before expiration/additional discount), plus (50,000 won /prepayment exemption), plus (100,000/5000 won discount), water tank type replacement (re-rent for less than 3 years), etc (Select the values (SELECTED_VALUES) in which the ratio of the value count of the DELIVERY_GOODS_TYPE variable is 1% or more and create a boolean variable (DELIVERY_GOODS_TYPE_TypeA) for each cont_line_seq (e.g. “TypeA” in SELECTED_VALUES)) |
14. Transfer type | Transfer Type Transfer Change Route | Categoric categoric | Payment methods such as bank transfer, cash payment, deposit without bankbook, credit card transfer, etc | |
15. Receiver address | Address_1st offset Address_2nd offset | Categoric categoric | Address of first offset of full address Address of second offset of full address |