Tourist-Demand-Forecasting-For-TTD

Our tourism demand forecasting model to forecast arrivals for the tourist place Tirumala Tirupati Devasthanam(TTD). We have considered TTD to our model because it maintains historic data of pilgrim arrivals which we access through TTD.News and TTD is popular tourist destination attracting large number of tourists every year. We have considered attributes Day Speciality(weekend or weekday), Weather condition(Based on temperature), Google Trends(frequency of ttd related keywords searched on google), Twitter trend(whether opinion of twitter is +ve,-ve or neutral). Our model uses structured variables to build a tourism demand forecasting model based on Light Gradient Boosting Machine Regressor. LGBMRegresssion uses a leaf-wise tree growth strategy which differs from level-wise strategy employed by many other gradient boosting implementations. In leaf-wise growth, algorithm selects the leaf node with maximum delta loss as the next node to grow. The ensembling algorithm at last forms a improved model. In this approach the google trends, day speciality contributed more information in recognizing the patterns of tourist arrivals. Another three model multiple linear regression, decision tree regression, xgboost(extreme gradient boosting algorithm) also tried to train the model, they have performed considerably but LGBMR gives better results over those three models.