A Comparative Study of Machine Learning Models for Airbnb Booking Likelihood Prediction in Singapore with GridSearchCV Optimization

Abstract

The rapid development of digitization has driven a fundamental shift in the travel industry, particularly through the emergence of shared economy platforms like Airbnb. Thus, the requirement to have a prediction model for travel industry is really crucial point and will give huge benefit to travel industry to solve their problem to finding host to rent their building for the industry and guest to rent for property rented from the platform. From this study we have several prediction models which developed from 10 machine learning algorithm. Each model has a distinct of its own accuracy f1 score and ROC AUC which each score would reveal how good is each model to be utilized for booking likelihood prediction. Some Algorithm like XGboost, Random Forest, Logistic Regression, SVM has huge of accuracy for each score to extend from 90% accuracy after we did a hyperparameter tunning to boost each model’s performances. However, these three score (accuracy, f1 score, and roc auc) aren’t sufficient to make the model work efficient and to be reliable to be utilized for our prediction model. Hence, more analytical methods are required to make sure our models are perform well for our aim to create a reliable prediction model and less bias on output we desired. In this study we shall commit curve learning analysist to make our models surely become a more reliable models which give numerous benefits to model accuracy and which many studies are still overturned the methods that very unfortunate. We shall dismantle our discovery and reveal which machine learning algorithm, optimization technique and how to analyst learning curve for each model on its own chapter hopefully our research is useful for everyone to gain new knowledge to developing prediction model using machine learning algorithm.