pricing recommendation system for airbnb host-gist

6
Note: As a part of my Portfolio this is the gist of my Master’s Project that I have worked. This is a real time industrial project that I have worked on with Airbnb Data and Analytics team for 6 months. Pricing Recommendation For Airbnb Host’s Summary Rates on Airbnb are set by hosts, who are normally individual homeowners. Some of the factors that impact the rates are location of the accommodation, type of apartment, number of guests, amenities provided and also nearby events. Such services have large competition with the hotel deals, prices of different apartment and services provided. The hosts do not have much idea on what their prices should range as they do not take into account the above factors. In this project a recommendation system for the hosts as well as the guests will be developed which gives best prices for the room/apartment. As a primary feature, the system will have an interactive visualization interface which would suggest the host which factors are affecting the current rates in that specific location and suggests price range for their apartment. In the advanced phase, this recommendation model will help guests get price range for future dates where hosts are not available yet. The project will use Apache Spark as the framework and hybrid recommendation model which includes Content based approach and Collaborative Information filtering. Primary model is written in R and Python. This recommendation system has a wide future scope and various data analysis can be performed to get more effective results. The Bigger Picture A major part of any recommendation system lies behind the machine learning algorithm on which it based on. Machine learning algorithm trains the system and makes it more reliable. In this project our recommendation system would be based on two

Upload: jay-vora

Post on 12-Apr-2017

9 views

Category:

Engineering


1 download

TRANSCRIPT

Page 1: Pricing recommendation system for Airbnb Host-Gist

Note: As a part of my Portfolio this is the gist of my Master’s Project that I have worked. This is a real time industrial project that I have worked on with Airbnb Data and Analytics team for 6 months.

Pricing Recommendation For Airbnb Host’s

Summary

Rates on Airbnb are set by hosts, who are normally individual homeowners. Some of the factors that impact the rates are location of the accommodation, type of apartment, number of guests, amenities provided and also nearby events. Such services have large competition with the hotel deals, prices of different apartment and services provided. The hosts do not have much idea on what their prices should range as they do not take into account the above factors. In this project a recommendation system for the hosts as well as the guests will be developed which gives best prices for the room/apartment. As a primary feature, the system will have an interactive visualization interface which would suggest the host which factors are affecting the current rates in that specific location and suggests price range for their apartment. In the advanced phase, this recommendation model will help guests get price range for future dates where hosts are not available yet. The project will use Apache Spark as the framework and hybrid recommendation model which includes Content based approach and Collaborative Information filtering. Primary model is written in R and Python. This recommendation system has a wide future scope and various data analysis can be performed to get more effective results.

The Bigger Picture

A major part of any recommendation system lies behind the machine learning algorithm on which it based on. Machine learning algorithm trains the system and makes it more reliable. In this project our recommendation system would be based on two major algorithms. Collaborative based filtering Algorithm and Content based filtering Algorithm.

Content based algorithms and collaborative filtering algorithm exploit the full database to produce a forecast. There are measurable methods to locate an arrangement of different hosts, known as neighbors, that have a past filled with concurring with the objective (i.e., they either rate distinctive things likewise or they tend to purchase comparable arrangements of things). Once an area of hosts is shaped, these algorithms use diverse calculations to consolidate the inclinations of neighbors to deliver an expectation or top-N suggestion for the dynamic host. As discussed above. The basic requirements to full fill this we need proper attributes to calculate. Few of the successfully mined data and then converted into structured data is given below.

Jay Vora/Airbnb Recommendation system

Page 2: Pricing recommendation system for Airbnb Host-Gist

Days of the week

Location of the host

Type of Apartment/Rooms

Check in/check out date

Past price range of nearby location

Though AirBNB provides pricing recommendation for host as discussed over various forums and tech talks there are actually many loopholes which widens the gaps of giving an accurate price. We are trying to optimize the classifier model used for pricing and through our initiative we are trying to lessen this gap and build more accurate model.

Jay Vora/Airbnb Recommendation system

Page 3: Pricing recommendation system for Airbnb Host-Gist

Future scope of Optimized Pricing model

The framework of pricing would be divided into three parts

Value Based: Value based is where you compute the value of the customer and price your product accordingly.Cost Based: Cost based is where you look at all costs associated (mortgage data in this case) and then add your margin on top to recommend price.Competition based: To look at competition like nearby hosts\ hotel rates in the areas and then price accordingly.

Jay Vora/Airbnb Recommendation system