To attract customers, the hotel chain has added to its website the ability to book a room without prepayment. However, if the customer cancelled the booking, the company suffered losses. The hotel staff could, for example, buy groceries for the arrival of a guest or simply not have time to find another client.
To solve this problem, we need to develop a system that predicts the rejection of armor. If the model shows that the reservation will be canceled, the client is invited to make a deposit. The deposit amount is 80% of the room rate for one day and the cost of a one-time cleaning. The money will be debited from the client's account if he still cancels the reservation.
The hotel_train and hotel_test tables contain the same columns:
- id - record number;
- adults - number of adult guests;
- arrival_date_year - year of arrival;
- arrival_date_month - month of arrival;
- arrival_date_week_number - arrival week;
- arrival_date_day_of_month - day of arrival;
- babies - number of babies;
- booking_changes:
- the number of changes to the order parameters;
- children:
- number of children from 3 to 14 years old;
- country - citizenship of the guest;
- customer_type - type of the customer:
- Contract - contract with a legal entity;
- Group - group check-in:
- Transient - not related to a contract or a group check-in;
- Transient-party - is not related to a contract or a group check-in, but is related to a booking of the Transient type.
- days_in_waiting_list - how many days the order was waiting for confirmation;
- distribution_channel - order distribution channel;
- is_canceled - order cancellation;
- is_repeated_guest - indicates that the guest is booking a room for the second time;
- lead_time - the number of days between the booking date and
- the arrival date;
- meal - order options:
- sc - no additional options;
- bb - breakfast included;
- Hb - breakfast and lunch included;
- breakfast, lunch and dinner are included.
- previous_bookings_not_canceled -
- the number of confirmed orders from the client;
- previous_cancellations - the number of cancelled orders from the client;
- required_car_parking_spaces - the need for a place for the car;
- reserved_room_type - type of the reserved room;
- stays_in_weekend_nights - number of nights on weekends;
- stays_in_week_nights - number of nights on weekdays;
- total_nights - total number of nights;
- total_of_special_requests - the number of special marks.
The main business metric for any hotel chain is its profit. The profit of the hotel is the difference between the cost of a room for all nights and the cost of service: both during the preparation of the room and during the stay of the guest.
The hotel has several types of rooms. Depending on the type of room, the cost per night is assigned. There are also cleaning costs. If the client has rented a room for a long time, then they are cleaned every two days.
The cost of hotel rooms:
- category A: per night - 1,000, one-time service - 400;
- category B: per night - 800, one-time service - 350;
- category C: per night - 600, one-time service - 350;
- category D: per night - 550, one-time service - 150;
- category E: 500 per night, one-time service - 150;
- category F: per night - 450, one-time service - 150;
- category G: per night - 350, one-time service - 150.
The hotel's pricing policy uses seasonal coefficients: in spring and autumn prices increase by 20%, in summer - by 40%.
The budget for the development of the forecasting system is 400,000. At the same time, it should be taken into account that the implementation of the model should pay off in a year. Development costs should be less than the revenue that the system will bring to the company.