Application of Machine Learning Techniques to Classify Intention to Pay for Forest Ecosystem Services

Capturing the ability to take part in the payment of forest ecosystem services by beneficiaries is the result that policy-making agencies are always concerned. This research selects several machine learning techniques, including single classifiers (Multilayer Perceptron, Naive Bayes, SMO) and ensemble classifiers (LogitBoost, Random Forest, Bagging) to evaluate and classify willingness-to-pay intention for mangrove ecosystem services of people in PhuLong commune, Vietnam. Research data is inherited from a previous contingent valuation survey, with a sample size of 235. The results show that the machine learning algorithms are workt with small sample-size data sets with feasibility prediction results in behavioral intent classification. The LogitBoost model achieves the best classification performance compared to the remaining models. Besides, socio-psychological factors are ranked as important factors in classifying behavioral intentions related to payment for forest ecosystem services.