Voluntary Turnover Prediction System for Tourism Industry with Special Reference to Hotel Industry in Sri Lanka

Abstract

"Sri Lanka is a popular tourist destination which attracts a high number of tourists. Tourism sector directly contributes 4.3% of Sri Lanka's Gross Domestic Product and ranks third in terms of country's foreign exchange earnings. The hotels and restaurant sector is one of the four sub-sectors which form the tourism and travel sector and the main role in the tourism sector is played by hotels by offering various guest amenities to visitors. In addition, 81% of all direct employment associated with tourism in Sri Lanka are found in hotels and restaurants. Furthermore, as the hotel industry is a service industry, it relies heavily upon human resources. One of the key challenges faced by the Sri Lankan tourism industry is high voluntary turnover, particularly among lower-level workforce. Since the highest productivity of the hotel industry is contributed by lower-level hotel employees as they directly engage with customers, retaining them is vital. When voluntary turnover occurs, hotels experience inability to deliver a quality service to customers and increase in expenses. Even though the need to gain an in depth understanding of employee turnover exists, carrying out research on this issue has been heavily neglected. Therefore, currently no proper systematic approaches are in place for hotel human resources personnel to utilise to address hotel employee turnover effectively . The author implemented machine learning models to accurately predict the turnover rate and the possible retention period of lower-level hotel employees. A classification model was developed to predict the turnover rate and a regression model was developed to predict the possible retention period of employees. Pre-processing tasks such as feature scaling, class imbalance handling was incorporated and hyper parameter tuning was also incorporated to enhance the performance of the models. Out of the classification models, ExtraTreesClassifier exhibited the highest F1 score of 98.8%, whilst XGBRFRegressor produced the highest R2 score of 96.33% out of the selected regression models. Through these models, the author sought to provide turnover insights pertaining to each individual employee and enable hotel Human Resources personnel to successfully address the issue of voluntary turnover among lower-level employees."