Recommender systems belp people find items of interest by making personalibed recotamendations accotding to their preferences. For example, a recosnmender systrm can male personalised recomamendations of iterns such as movies, boolss, hotels, es music to people. In order to model noers' preferences their past interactions such as product views, ratings, and purchves are used. In the recotumender systems area of research (which is a sabtield of machine learning and information retrietal) different algorithms have been developed which are currently being used by many large coenpanies. In this project you will implement neighborhood based collaborative filtering (NBCF) algorithms in order to make predictions for motie ratings of people. There are two main types of NBCF algorithms: user-based (UBCF) and item-based (IBCF). Let ws docribe each with an example dataset. 2 UBCF and IBCF One of the fundamental problems in recommender systems is to predict the rating of a nser for a partienlar item. For example, given the dataset in Table I, what maight be the rating of User 2 for Movie 3? If we can predict this rating then, if it is a high value lilee 4 or 5 , we can decide to recommend Movie 3 to User 2. Table 1: An example datasct. In its simplest form, UBCF works as follows in order to predict the rating of wer w to item i, we first find the most similar k users to u (who rated i ) asd predict the rating as the average ratings of these most similar k users on item 1 . For finding the similarity between two users different methods can be ased. In collaborative filtering we look at the ratings of other users and find users which have similar ratings. One popular method to find sach a similarity is called cosine similarity which is given below: cosin(A,B)=ABBAB=i=1nAi2i=1nBi2i=1nA,Bi For example, given two vectors A=2,4,5 and B=3,2,4, their coeine similarity is given by; cosine(A,B)=22+42+5232+22+4223+42+54