Many people around the globe live in areas that have unhealthy levels of air pollution. Such air pollution raises the risks of health problems. Current travel guidance systems such as Google Maps give recommendations to users mainly based on short travel time. Driven by the idea of improving air quality, mobility and quality of life for city residents, we contribute smart travel recommender system, which uses a recommendation engine to suggest the optimum commute mode based on variables such as air quality, distance, weather, and users' preferences. To make it more realistic and adaptive, our algorithm gives more weight to newer commute observations and preferences while suppressing the impact of the older ones. The recommendations are influenced by different predictor variables such as 'Distance', 'Air Quality', 'Weather' etc. Each variable is analyzed separately using a parameter called 'Phi-Correlation' coefficient to determine its impact on the commute mode choice. We proposed a ranking algorithm to calculate a score based upon the weighted impact of predictor variables and choose the one with highest score. We simulated different user profiles to determine the recommendations accuracy. The results proved that the recommender system quickly learns about user's travel patterns and generates relevant suggestions.