Tourism Footfall Forecasting and Recommendation System

Tourism industry is crucial factor in the growth of an economy as it boosts the revenue and creates jobs for the local population. For the development of infrastructure to support the tourist influx, an accurate forecasting of tourist footfall is essential. The uncertainty regarding tourist arrivals because of inaccurate forecasting models directly and indirectly affect the tourism industry. In this study we have discussed a possible solution of building an ML model that forecasts arrivals based on time series data and google search index. The solution operates on the time series data as the basis of prediction and optimizes the results as per the search frequency of various queries listed in the search index. These results further can be interpreted and tweaked by adjusting the parametric weights used for calculations. For analysis, we have used the publicly available Manali tourist data as well as Google Trends to import worldwide search query data since 2008. The analysis has been done using the ARIMA and LSTM algorithms. Normalised Root Mean Squared Error (NRMSE) has been used as an evaluation criteria of the different models being tested. The results suggest that the LSTM model outperforms the traditional ARIMA model. A recommendation system which provides information regarding nearby hotels based on the real-time location has also been built to enhance the user experience.