Prediction of hotel booking cancellation using machine learning algorithms

In data mining, prediction models estimate continuously valued functions, whereas classification models forecast based on categorical class labels. For example, a classification model can classify bank loan applications as either safe or risky, whereas a prediction model can anticipate how much money potential customers will spend on computer equipment depending on their work and income. Unexpected booking cancellations have a detrimental impact on hotels' reputation, relationships with customers, and revenue. Predictive analysis can assist hotels in estimating cancellation probability to help mitigate this issue. Hotels can examine their own data sets to find patterns and predict future cancellations by utilizing classification models. In this research, a hotel dataset is used to predict booking cancellations with the help of four different classification algorithms i.e., J48 Tree, Naive Bayes, Logistic Function and Random Forest Tree algorithms using Weka workbench. After studying the results and comparing the time taken to build models, accuracy percentage, error rates, precision, sensitivity, specificity and lift of all four algorithms, the Random Forest Tree Algorithm was found to be most suitable one to predict booking cancellations.