Collabrative-filtering based recommender system
The dataset used for making recommendations to users is taken from FOURSQUARE. The data is processed and each tupple for every user is of the form :
user = (userid , locid , countTotal , rating , distance , Y-Value)
To recommend locations or points of Interest(POI's) ‘classification’ algorithms to classify the locations based on the clusters have been used. For this, I compare support vector machines (SVM), radial basis function (RBF) neural network, and probabilistic neural network (PNN), which are the state-of-the-art algorithms currently used for classification. In addition, Adaptive Boosting-Gradient (more commonly known as AdaBoost) algorithm, which is a meta-algorithm that combines the result from many weaker decision-tree algorithms into a single strong classifier is introduced.
For every user to whom locations have to be recommended , the complete data for the user should be there in the form specified above. On the basis of that data recommendations are made.
Processed Foursquare dataset used : https://drive.google.com/open?id=0BxBHOsPsYg5DRW9xN3c0bWVNV2M.